fix(chat): sort messages appropriately

Add a running-session view to the sidebar, including header controls, running counts, empty states, and row-level processing indicators so active provider work is visible outside the current chat.

Hydrate running state after refresh through a status-only /api/providers/sessions/running endpoint backed by chatRunRegistry.listRunningRuns, then sync and poll the frontend processingSessions map from AppContent without attaching to chat streams or replaying messages.

Preserve fresh local processing entries during sync so newly sent messages are not cleared before the backend registry catches up, and clear completed sessions once the status endpoint no longer reports them.

Thread active session state through sidebar project/session components, show rotating loaders for processing sessions, and keep the running search mode expanded and filterable.

Fix optimistic local user-message dedupe so repeated prompts are only collapsed when a matching server echo appears from the same send window, preventing sent messages from disappearing until assistant completion.

Add registry test coverage for listing currently running app sessions.

Tests: npx eslint on changed files; npx tsc --noEmit -p tsconfig.json; npx tsc --noEmit -p server/tsconfig.json; npx tsx --tsconfig server/tsconfig.json --test server/modules/websocket/tests/chat-run-registry.test.ts.

CHANGELOG.md

@@ -3,6 +3,74 @@
 All notable changes to CloudCLI UI will be documented in this file.
 
 
+## [](https://github.com/siteboon/claudecodeui/compare/v1.33.3...vnull) (2026-06-09)
+
+### New Features
+
+* adding Fable 5 in claude code ([ce327b6](https://github.com/siteboon/claudecodeui/commit/ce327b6fa9329aa3e9a3a1da7225ca01d3b06ac5))
+
+## [1.33.3](https://github.com/siteboon/claudecodeui/compare/v1.33.2...v1.33.3) (2026-06-09)
+
+### New Features
+
+* add file tree upload progress ([c235b05](https://github.com/siteboon/claudecodeui/commit/c235b05e1d3b626667dba4043b685512e3cd3d5d))
+* signal when chat runs complete ([d70dc07](https://github.com/siteboon/claudecodeui/commit/d70dc077bfbbfcf2ff4fa5514fabf7b4485861fa))
+
+### Bug Fixes
+
+* address notification review feedback ([602e6ad](https://github.com/siteboon/claudecodeui/commit/602e6ad4acba612a7ea66fb3bc7485054f5675ee))
+* align prism plugin name and id with manifest.json ([ca8fd0e](https://github.com/siteboon/claudecodeui/commit/ca8fd0ee235b6a3210157bd0d9af83024d4a2248))
+* **chat:** re-anchor initial scroll across lazy content reflow ([33a4e72](https://github.com/siteboon/claudecodeui/commit/33a4e72ca4f84df60aadfc4ff3f3467d6f5ae948))
+* keep editor toolbar in view on long unwrapped lines ([beae8c6](https://github.com/siteboon/claudecodeui/commit/beae8c6513daa7518b9de40d8bfde3bf08e7bc87))
+* **sandbox:** prevent server SIGHUP on sbx exec exit ([#792](https://github.com/siteboon/claudecodeui/issues/792)) ([f4a1614](https://github.com/siteboon/claudecodeui/commit/f4a1614a0a4ab4b65e8368d5e4221f015cb7555d)), closes [#791](https://github.com/siteboon/claudecodeui/issues/791)
+* slash command suggestions trigger at any / in input, not only at start ([#843](https://github.com/siteboon/claudecodeui/issues/843)) ([f7c0024](https://github.com/siteboon/claudecodeui/commit/f7c0024fe15057ad049c71e15e88adb482a4497f))
+* update naming convention ([3cd8995](https://github.com/siteboon/claudecodeui/commit/3cd89956ba06f0fc3e17d349b0c50baab4012658))
+
+### Maintenance
+
+* add prism plugin ([01dbe2a](https://github.com/siteboon/claudecodeui/commit/01dbe2a8bfcb3b265995f01f905b218d5f576f7b))
+
+## [1.33.2](https://github.com/siteboon/claudecodeui/compare/v1.33.1...v1.33.2) (2026-06-08)
+
+### New Features
+
+* **chat:** open cost modal from token usage ([f238050](https://github.com/siteboon/claudecodeui/commit/f238050b85c3b99a702a8635059735e1a3b3a4f4))
+* **i18n:** add Traditional Chinese (zh-TW) locale ([#773](https://github.com/siteboon/claudecodeui/issues/773)) ([c21a9f4](https://github.com/siteboon/claudecodeui/commit/c21a9f45610eb1eeb650d8e6cf8650e798f77f6f))
+
+### Bug Fixes
+
+* do not show model description in chat view ([d638a89](https://github.com/siteboon/claudecodeui/commit/d638a8982c7f75b08fc7f65f01d6d54989c790d1))
+* include Claude cache tokens in usage ([ed9cdf0](https://github.com/siteboon/claudecodeui/commit/ed9cdf01145fa0d063580bb76d30cfa7ee67af86))
+
+## [1.33.1](https://github.com/siteboon/claudecodeui/compare/v1.33.0...v1.33.1) (2026-06-05)
+
+### New Features
+
+* **chat:** auto-detect text direction for RTL languages ([#729](https://github.com/siteboon/claudecodeui/issues/729)) ([fa9eaf5](https://github.com/siteboon/claudecodeui/commit/fa9eaf5573a6f870a19fb62ab430ffd87c466582))
+
+### Bug Fixes
+
+* file tree concurrency ([#828](https://github.com/siteboon/claudecodeui/issues/828)) ([ebb0e59](https://github.com/siteboon/claudecodeui/commit/ebb0e59e8023c0a8040d168a5adffb7102e80561))
+* load claude models directly from provider ([cdcac18](https://github.com/siteboon/claudecodeui/commit/cdcac182d458a24908777568979c8e756f94428c))
+* plugin svg icon sanitization ([#817](https://github.com/siteboon/claudecodeui/issues/817)) ([d9e9df1](https://github.com/siteboon/claudecodeui/commit/d9e9df183f462c88c3b60975eb8254faa9168717))
+* recognize claude auth token env ([#818](https://github.com/siteboon/claudecodeui/issues/818)) ([43c33d5](https://github.com/siteboon/claudecodeui/commit/43c33d5cb1b41835dfe3bccd450c5a9c2441509b))
+* redact websocket auth token in logs ([#827](https://github.com/siteboon/claudecodeui/issues/827)) ([14ddbc7](https://github.com/siteboon/claudecodeui/commit/14ddbc7c57a01da9fb65fd87d8588532b11833fa))
+* remove thinking mode ([#835](https://github.com/siteboon/claudecodeui/issues/835)) ([2149b87](https://github.com/siteboon/claudecodeui/commit/2149b8776b7ebfec0eace413f4fc527ccb2324c0))
+* **shell:** disconnect and restart buttons ([#831](https://github.com/siteboon/claudecodeui/issues/831)) ([ef2fd48](https://github.com/siteboon/claudecodeui/commit/ef2fd48b46452d4b9e2bf1f5e3c30fafe19f27f2))
+* show Claude tool result errors ([bb8db58](https://github.com/siteboon/claudecodeui/commit/bb8db5815c2d20ee4fbfa02d14c886a56ef352e0))
+* **vite:** proxy /plugin-ws WebSocket requests to the backend in dev ([#757](https://github.com/siteboon/claudecodeui/issues/757)) ([96b16b4](https://github.com/siteboon/claudecodeui/commit/96b16b42e4f807d04ec743a5a4117a37a3f5e0d9))
+* **websocket:** add 30s server-side heartbeat to prevent proxy idle disconnects ([#770](https://github.com/siteboon/claudecodeui/issues/770)) ([2edfef2](https://github.com/siteboon/claudecodeui/commit/2edfef2e3f4271c29ae8670df9dd382a9eef7c3c)), closes [#769](https://github.com/siteboon/claudecodeui/issues/769)
+* **websocket:** reset unmountedRef on each effect re-run so token refresh reconnects ([#721](https://github.com/siteboon/claudecodeui/issues/721)) ([f082cdc](https://github.com/siteboon/claudecodeui/commit/f082cdc63bd0de90f8b3da1df6071e91ab545831))
+
+### Documentation
+
+* add nginx subpath deployment template ([#820](https://github.com/siteboon/claudecodeui/issues/820)) ([3ec76b5](https://github.com/siteboon/claudecodeui/commit/3ec76b5bb15a13cec41056f4c9b9c425195022fa))
+
+### Maintenance
+
+* update Claude fallback models ([94785bf](https://github.com/siteboon/claudecodeui/commit/94785bfa579d1f39a2bee0f9dd0f09fd0243bc79))
+* update package-lock.json ([c90b341](https://github.com/siteboon/claudecodeui/commit/c90b34108e86a3effdb5c6979ea7b1692d2b9da0))
+
 ## [](https://github.com/siteboon/claudecodeui/compare/v1.32.0...vnull) (2026-06-01)
 
 ### New Features

README.de.md

@@ -15,7 +15,7 @@
   <a href="https://trendshift.io/repositories/15586" target="_blank"><img src="https://trendshift.io/api/badge/repositories/15586" alt="siteboon%2Fclaudecodeui | Trendshift" style="width: 250px; height: 55px;" width="250" height="55"/></a>
 </p>
 
-<div align="right"><i><a href="./README.md">English</a> · <a href="./README.ru.md">Русский</a> · <b>Deutsch</b> · <a href="./README.ko.md">한국어</a> · <a href="./README.zh-CN.md">中文</a> · <a href="./README.ja.md">日本語</a> · <a href="./README.tr.md">Türkçe</a></i></div>
+<div align="right"><i><a href="./README.md">English</a> · <a href="./README.ru.md">Русский</a> · <b>Deutsch</b> · <a href="./README.ko.md">한국어</a> · <a href="./README.zh-CN.md">简体中文</a> · <a href="./README.zh-TW.md">繁體中文</a> · <a href="./README.ja.md">日本語</a> · <a href="./README.tr.md">Türkçe</a></i></div>
 
 ---
 
@@ -62,7 +62,7 @@
 - **Sitzungsverwaltung** – Gespräche fortsetzen, mehrere Sitzungen verwalten und Verlauf nachverfolgen
 - **Plugin-System** – CloudCLI mit eigenen Plugins erweitern – neue Tabs, Backend-Dienste und Integrationen hinzufügen. [Eigenes Plugin erstellen →](https://github.com/cloudcli-ai/cloudcli-plugin-starter)
 - **TaskMaster AI Integration** *(Optional)* – Erweitertes Projektmanagement mit KI-gestützter Aufgabenplanung, PRD-Parsing und Workflow-Automatisierung
-- **Modell-Kompatibilität** – Funktioniert mit Claude, GPT und Gemini (vollständige Liste unterstützter Modelle in [`public/modelConstants.js`](public/modelConstants.js))
+- **Modell-Kompatibilität** – Funktioniert mit Claude, GPT und Gemini (vollständige Liste unterstützter Modelle zur Laufzeit über `GET /api/providers/:provider/models`)
 
 
 ## Schnellstart

README.ja.md

@@ -15,7 +15,7 @@
   <a href="https://trendshift.io/repositories/15586" target="_blank"><img src="https://trendshift.io/api/badge/repositories/15586" alt="siteboon%2Fclaudecodeui | Trendshift" style="width: 250px; height: 55px;" width="250" height="55"/></a>
 </p>
 
-<div align="right"><i><a href="./README.md">English</a> · <a href="./README.ru.md">Русский</a> · <a href="./README.de.md">Deutsch</a> · <a href="./README.ko.md">한국어</a> · <a href="./README.zh-CN.md">中文</a> · <b>日本語</b> · <a href="./README.tr.md">Türkçe</a></i></div>
+<div align="right"><i><a href="./README.md">English</a> · <a href="./README.ru.md">Русский</a> · <a href="./README.de.md">Deutsch</a> · <a href="./README.ko.md">한국어</a> · <a href="./README.zh-CN.md">简体中文</a> · <a href="./README.zh-TW.md">繁體中文</a> · <b>日本語</b> · <a href="./README.tr.md">Türkçe</a></i></div>
 
 ---
 

README.ko.md

@@ -15,7 +15,7 @@
   <a href="https://trendshift.io/repositories/15586" target="_blank"><img src="https://trendshift.io/api/badge/repositories/15586" alt="siteboon%2Fclaudecodeui | Trendshift" style="width: 250px; height: 55px;" width="250" height="55"/></a>
 </p>
 
-<div align="right"><i><a href="./README.md">English</a> · <a href="./README.ru.md">Русский</a> · <a href="./README.de.md">Deutsch</a> · <b>한국어</b> · <a href="./README.zh-CN.md">中文</a> · <a href="./README.ja.md">日本語</a> · <a href="./README.tr.md">Türkçe</a></i></div>
+<div align="right"><i><a href="./README.md">English</a> · <a href="./README.ru.md">Русский</a> · <a href="./README.de.md">Deutsch</a> · <b>한국어</b> · <a href="./README.zh-CN.md">简体中文</a> · <a href="./README.zh-TW.md">繁體中文</a> · <a href="./README.ja.md">日本語</a> · <a href="./README.tr.md">Türkçe</a></i></div>
 
 ---
 
@@ -60,7 +60,7 @@
 - **세션 관리** - 대화를 재개하고, 여러 세션을 관리하며 기록을 추적
 - **플러그인 시스템** - 커스텀 탭, 백엔드 서비스, 통합을 추가하여 CloudCLI 확장. [직접 빌드 →](https://github.com/cloudcli-ai/cloudcli-plugin-starter)
 - **TaskMaster AI 통합** *(선택사항)* - AI 중심의 작업 계획, PRD 파싱, 워크플로 자동화를 통한 고급 프로젝트 관리
-- **모델 호환성** - Claude, GPT, Gemini 모델 계열에서 작동 (`public/modelConstants.js`에서 전체 지원 모델 확인)
+- **모델 호환성** - Claude, GPT, Gemini 모델 계열에서 작동 (`GET /api/providers/:provider/models` API에서 전체 지원 모델 확인)
 
 ## 빠른 시작
 

README.md

@@ -15,7 +15,7 @@
   <a href="https://trendshift.io/repositories/15586" target="_blank"><img src="https://trendshift.io/api/badge/repositories/15586" alt="siteboon%2Fclaudecodeui | Trendshift" style="width: 250px; height: 55px;" width="250" height="55"/></a>
 </p>
 
-<div align="right"><i><b>English</b> · <a href="./README.ru.md">Русский</a> · <a href="./README.de.md">Deutsch</a> · <a href="./README.ko.md">한국어</a> · <a href="./README.zh-CN.md">中文</a> · <a href="./README.ja.md">日本語</a> · <a href="./README.tr.md">Türkçe</a></i></div>
+<div align="right"><i><b>English</b> · <a href="./README.ru.md">Русский</a> · <a href="./README.de.md">Deutsch</a> · <a href="./README.ko.md">한국어</a> · <a href="./README.zh-CN.md">简体中文</a> · <a href="./README.zh-TW.md">繁體中文</a> · <a href="./README.ja.md">日本語</a> · <a href="./README.tr.md">Türkçe</a></i></div>
 
 ---
 
@@ -62,7 +62,7 @@
 - **Session Management** - Resume conversations, manage multiple sessions, and track history
 - **Plugin System** - Extend CloudCLI with custom plugins — add new tabs, backend services, and integrations. [Build your own →](https://github.com/cloudcli-ai/cloudcli-plugin-starter)
 - **TaskMaster AI Integration** *(Optional)* - Advanced project management with AI-powered task planning, PRD parsing, and workflow automation
-- **Model Compatibility** - Works with Claude, GPT, and Gemini model families (see [`public/modelConstants.js`](public/modelConstants.js) for the full list of supported models)
+- **Model Compatibility** - Works with Claude, GPT, and Gemini model families (the full list of supported models is available at runtime via `GET /api/providers/:provider/models`)
 
 
 ## Quick Start

README.ru.md

@@ -15,7 +15,7 @@
   <a href="https://trendshift.io/repositories/15586" target="_blank"><img src="https://trendshift.io/api/badge/repositories/15586" alt="siteboon%2Fclaudecodeui | Trendshift" style="width: 250px; height: 55px;" width="250" height="55"/></a>
 </p>
 
-<div align="right"><i><a href="./README.md">English</a> · <b>Русский</b> · <a href="./README.de.md">Deutsch</a> · <a href="./README.ko.md">한국어</a> · <a href="./README.zh-CN.md">中文</a> · <a href="./README.ja.md">日本語</a> · <a href="./README.tr.md">Türkçe</a></i></div>
+<div align="right"><i><a href="./README.md">English</a> · <b>Русский</b> · <a href="./README.de.md">Deutsch</a> · <a href="./README.ko.md">한국어</a> · <a href="./README.zh-CN.md">简体中文</a> · <a href="./README.zh-TW.md">繁體中文</a> · <a href="./README.ja.md">日本語</a> · <a href="./README.tr.md">Türkçe</a></i></div>
 
 ---
 
@@ -62,7 +62,7 @@
 - **Управление сессиями** - возобновляйте диалоги, управляйте несколькими сессиями и отслеживайте историю
 - **Система плагинов** - расширяйте CloudCLI кастомными плагинами — добавляйте новые вкладки, бэкенд-сервисы и интеграции. [Создать свой →](https://github.com/cloudcli-ai/cloudcli-plugin-starter)
 - **Интеграция с TaskMaster AI** *(опционально)* - продвинутое управление проектами с планированием задач на базе AI, разбором PRD и автоматизацией workflow
-- **Совместимость с моделями** - работает с семействами моделей Claude, GPT и Gemini (см. [`public/modelConstants.js`](public/modelConstants.js) для полного списка поддерживаемых моделей)
+- **Совместимость с моделями** - работает с семействами моделей Claude, GPT и Gemini (полный список поддерживаемых моделей доступен через `GET /api/providers/:provider/models`)
 
 
 ## Быстрый старт

README.tr.md

@@ -15,7 +15,7 @@
   <a href="https://trendshift.io/repositories/15586" target="_blank"><img src="https://trendshift.io/api/badge/repositories/15586" alt="siteboon%2Fclaudecodeui | Trendshift" style="width: 250px; height: 55px;" width="250" height="55"/></a>
 </p>
 
-<div align="right"><i><a href="./README.md">English</a> · <a href="./README.ru.md">Русский</a> · <a href="./README.de.md">Deutsch</a> · <a href="./README.ko.md">한국어</a> · <a href="./README.zh-CN.md">中文</a> · <a href="./README.ja.md">日本語</a> · <b>Türkçe</b></i></div>
+<div align="right"><i><a href="./README.md">English</a> · <a href="./README.ru.md">Русский</a> · <a href="./README.de.md">Deutsch</a> · <a href="./README.ko.md">한국어</a> · <a href="./README.zh-CN.md">简体中文</a> · <a href="./README.zh-TW.md">繁體中文</a> · <a href="./README.ja.md">日本語</a> · <b>Türkçe</b></i></div>
 
 ---
 
@@ -62,7 +62,7 @@
 - **Oturum Yönetimi** — Konuşmalara devam et, birden fazla oturumu yönet ve geçmişi takip et
 - **Eklenti Sistemi** — CloudCLI'ı özel eklentilerle genişlet: yeni sekmeler, arka uç servisleri ve entegrasyonlar ekle. [Kendi eklentini yaz →](https://github.com/cloudcli-ai/cloudcli-plugin-starter)
 - **TaskMaster AI Entegrasyonu** *(İsteğe Bağlı)* — AI destekli görev planlama, PRD ayrıştırma ve iş akışı otomasyonu ile gelişmiş proje yönetimi
-- **Model Uyumluluğu** — Claude, GPT ve Gemini model aileleriyle çalışır (desteklenen tüm modeller için [`public/modelConstants.js`](public/modelConstants.js) dosyasına bak)
+- **Model Uyumluluğu** — Claude, GPT ve Gemini model aileleriyle çalışır (desteklenen tüm modeller için `GET /api/providers/:provider/models` API'sine bak)
 
 
 ## Hızlı Başlangıç

README.zh-CN.md

@@ -15,7 +15,7 @@
   <a href="https://trendshift.io/repositories/15586" target="_blank"><img src="https://trendshift.io/api/badge/repositories/15586" alt="siteboon%2Fclaudecodeui | Trendshift" style="width: 250px; height: 55px;" width="250" height="55"/></a>
 </p>
 
-<div align="right"><i><a href="./README.md">English</a> · <a href="./README.ru.md">Русский</a> · <a href="./README.de.md">Deutsch</a> · <a href="./README.ko.md">한국어</a> · <b>中文</b> · <a href="./README.ja.md">日本語</a> · <a href="./README.tr.md">Türkçe</a></i></div>
+<div align="right"><i><a href="./README.md">English</a> · <a href="./README.ru.md">Русский</a> · <a href="./README.de.md">Deutsch</a> · <a href="./README.ko.md">한국어</a> · <b>简体中文</b> · <a href="./README.zh-TW.md">繁體中文</a> · <a href="./README.ja.md">日本語</a> · <a href="./README.tr.md">Türkçe</a></i></div>
 
 ---
 
@@ -60,7 +60,7 @@
 - **会话管理** - 恢复对话、管理多个会话并跟踪历史记录
 - **插件系统** - 通过自定义选项卡、后端服务与集成扩展 CloudCLI。 [开始构建 →](https://github.com/cloudcli-ai/cloudcli-plugin-starter)
 - **TaskMaster AI 集成** *(可选)* - 结合 AI 任务规划、PRD 分析与工作流自动化，实现高级项目管理
-- **模型兼容性** - 支持 Claude、GPT、Gemini 模型家族（完整支持列表见 [`public/modelConstants.js`](public/modelConstants.js)）
+- **模型兼容性** - 支持 Claude、GPT、Gemini 模型家族（完整支持列表可通过 `GET /api/providers/:provider/models` 接口获取）
 
 ## 快速开始
 

README.zh-TW.md

@@ -0,0 +1,242 @@
+<div align="center">
+  <img src="public/logo.svg" alt="CloudCLI UI" width="64" height="64">
+  <h1>Cloud CLI（又名 Claude Code UI）</h1>
+  <p><a href="https://docs.anthropic.com/en/docs/claude-code">Claude Code</a>、<a href="https://docs.cursor.com/en/cli/overview">Cursor CLI</a>、<a href="https://developers.openai.com/codex">Codex</a> 和 <a href="https://geminicli.com/">Gemini-CLI</a> 的桌面和行動裝置 UI。可在本機或遠端使用，從任何地方查看您的專案與工作階段。</p>
+</div>
+
+<p align="center">
+  <a href="https://cloudcli.ai">CloudCLI Cloud</a> · <a href="https://cloudcli.ai/docs">文件</a> · <a href="https://discord.gg/buxwujPNRE">Discord</a> · <a href="https://github.com/siteboon/claudecodeui/issues">Bug 回報</a> · <a href="CONTRIBUTING.md">貢獻指南</a>
+</p>
+
+<p align="center">
+  <a href="https://cloudcli.ai"><img src="https://img.shields.io/badge/☁️_CloudCLI_Cloud-Try_Now-0066FF?style=for-the-badge" alt="CloudCLI Cloud"></a>
+  <a href="https://discord.gg/buxwujPNRE"><img src="https://img.shields.io/badge/Discord-Join%20Community-5865F2?style=for-the-badge&logo=discord&logoColor=white" alt="加入 Discord 社群"></a>
+  <br><br>
+  <a href="https://trendshift.io/repositories/15586" target="_blank"><img src="https://trendshift.io/api/badge/repositories/15586" alt="siteboon%2Fclaudecodeui | Trendshift" style="width: 250px; height: 55px;" width="250" height="55"/></a>
+</p>
+
+<div align="right"><i><a href="./README.md">English</a> · <a href="./README.ru.md">Русский</a> · <a href="./README.de.md">Deutsch</a> · <a href="./README.ko.md">한국어</a> · <a href="./README.zh-CN.md">简体中文</a> · <b>繁體中文</b> · <a href="./README.ja.md">日本語</a> · <a href="./README.tr.md">Türkçe</a></i></div>
+
+---
+
+## 截圖
+
+<div align="center">
+
+<table>
+<tr>
+<td align="center">
+<h3>桌面檢視</h3>
+<img src="public/screenshots/desktop-main.png" alt="桌面介面" width="400">
+<br>
+<em>顯示專案總覽和聊天的主介面</em>
+</td>
+<td align="center">
+<h3>行動裝置體驗</h3>
+<img src="public/screenshots/mobile-chat.png" alt="行動裝置介面" width="250">
+<br>
+<em>具有觸控導覽的響應式行動裝置設計</em>
+</td>
+</tr>
+<tr>
+<td align="center" colspan="2">
+<h3>CLI 選擇</h3>
+<img src="public/screenshots/cli-selection.png" alt="CLI 選擇" width="400">
+<br>
+<em>在 Claude Code、Gemini、Cursor CLI 與 Codex 之間進行選擇</em>
+</td>
+</tr>
+</table>
+
+</div>
+
+## 功能
+
+- **響應式設計** — 在桌面、平板和行動裝置上無縫運作，讓您隨時隨地使用 Agents
+- **互動聊天介面** — 內建聊天 UI，輕鬆與 Agents 交流
+- **整合 Shell 終端機** — 透過內建 shell 功能直接存取 Agents CLI
+- **檔案瀏覽器** — 互動式檔案樹，支援語法醒目提示與即時編輯
+- **Git 瀏覽器** — 檢視、暫存並提交變更，還可切換分支
+- **工作階段管理** — 恢復對話、管理多個工作階段並追蹤歷史紀錄
+- **外掛系統** — 透過自訂分頁、後端服務與整合來擴充 CloudCLI。[開始建構 →](https://github.com/cloudcli-ai/cloudcli-plugin-starter)
+- **TaskMaster AI 整合** *(選用)* — 結合 AI 任務規劃、PRD 分析與工作流程自動化，實現進階專案管理
+- **模型相容性** — 支援 Claude、GPT、Gemini 模型家族（完整支援列表可透過 `GET /api/providers/:provider/models` 介面取得）
+
+## 快速開始
+
+### CloudCLI Cloud（推薦）
+
+無需本機設定即可快速啟動。提供可透過網路瀏覽器、行動應用程式、API 或慣用的 IDE 存取的完全容器化託管開發環境。
+
+**[立即開始 CloudCLI Cloud](https://cloudcli.ai)**
+
+### 自架（開源）
+
+#### npm
+
+啟動 CloudCLI UI，只需一行 `npx`（需要 Node.js v22+）：
+
+```bash
+npx @cloudcli-ai/cloudcli
+```
+
+或進行全域安裝，便於日常使用：
+
+```bash
+npm install -g @cloudcli-ai/cloudcli
+cloudcli
+```
+
+開啟 `http://localhost:3001`，系統會自動發現所有現有工作階段。
+
+更多設定選項、PM2、遠端伺服器設定等，請參閱 **[文件 →](https://cloudcli.ai/docs)**。
+
+#### Docker Sandboxes（實驗性）
+
+在隔離的沙箱中執行代理，具有虛擬機管理程式等級的隔離。預設啟動 Claude Code。需要 [`sbx` CLI](https://docs.docker.com/ai/sandboxes/get-started/)。
+
+```bash
+npx @cloudcli-ai/cloudcli@latest sandbox ~/my-project
+```
+
+支援 Claude Code、Codex 和 Gemini CLI。詳情請參閱[沙箱文件](docker/)。
+
+---
+
+## 哪個選項更適合你？
+
+CloudCLI UI 是 CloudCLI Cloud 的開源 UI 層。你可以在本機上自架它，也可以使用提供團隊功能與深入整合的 CloudCLI Cloud。
+
+| | CloudCLI UI（自架） | CloudCLI Cloud |
+|---|---|---|
+| **適合對象** | 需要為本機代理工作階段提供完整 UI 的開發者 | 需要部署在雲端，隨時從任何地方存取代理的團隊與開發者 |
+| **存取方式** | 透過 `[yourip]:port` 在瀏覽器中存取 | 瀏覽器、任意 IDE、REST API、n8n |
+| **設定** | `npx @cloudcli-ai/cloudcli` | 無需設定 |
+| **機器需保持開機嗎** | 是 | 否 |
+| **行動裝置存取** | 網路內任意瀏覽器 | 任意裝置（原生應用程式即將推出） |
+| **可用工作階段** | 自動發現 `~/.claude` 中的所有工作階段 | 雲端環境內的工作階段 |
+| **支援的 Agents** | Claude Code、Cursor CLI、Codex、Gemini CLI | Claude Code、Cursor CLI、Codex、Gemini CLI |
+| **檔案瀏覽與 Git** | 內建於 UI | 內建於 UI |
+| **MCP 設定** | UI 管理，與本機 `~/.claude` 設定同步 | UI 管理 |
+| **IDE 存取** | 本機 IDE | 任何連線到雲端環境的 IDE |
+| **REST API** | 是 | 是 |
+| **n8n 節點** | 否 | 是 |
+| **團隊共享** | 否 | 是 |
+| **平台費用** | 免費開源 | 起價 $7/月 |
+
+> 兩種方式都使用你自己的 AI 訂閱（Claude、Cursor 等）— CloudCLI 提供環境，而非 AI。
+
+---
+
+## 安全與工具設定
+
+**🔒 重要提示**：所有 Claude Code 工具預設**停用**，可防止潛在的有害操作自動執行。
+
+### 啟用工具
+
+1. **開啟工具設定** — 點擊側邊欄齒輪圖示
+2. **選擇性啟用** — 僅啟用所需工具
+3. **套用設定** — 偏好設定儲存在本機
+
+<div align="center">
+
+![工具設定彈出視窗](public/screenshots/tools-modal.png)
+*工具設定介面 — 只啟用你需要的內容*
+
+</div>
+
+**建議做法**：先啟用基礎工具，再根據需要新增其他工具。隨時可以調整。
+
+---
+
+## 外掛
+
+CloudCLI 配備外掛系統，允許你新增帶有自訂前端 UI 和選用 Node.js 後端的分頁。在 Settings > Plugins 中直接從 Git 儲存庫安裝外掛，或自行開發。
+
+### 可用外掛
+
+| 外掛 | 描述 |
+|---|---|
+| **[Project Stats](https://github.com/cloudcli-ai/cloudcli-plugin-starter)** | 展示目前專案的檔案數、程式碼行數、檔案類型分佈、最大檔案以及最近修改的檔案 |
+
+### 自行建構
+
+**[Plugin Starter Template →](https://github.com/cloudcli-ai/cloudcli-plugin-starter)** — Fork 該儲存庫以建構自己的外掛。範例包括前端渲染、即時上下文更新和 RPC 通訊。
+
+**[外掛文件 →](https://cloudcli.ai/docs/plugin-overview)** — 提供外掛 API、清單格式、安全模型等完整指南。
+
+---
+
+## 常見問題
+
+<details>
+<summary>與 Claude Code Remote Control 有何不同？</summary>
+
+Claude Code Remote Control 讓你傳送訊息到本機終端機中已經執行的工作階段。該方式要求你的機器保持開機，終端機保持開啟，中斷網路後約 10 分鐘工作階段會逾時。
+
+CloudCLI UI 與 CloudCLI Cloud 是對 Claude Code 的擴充，而非旁觀 — MCP 伺服器、權限、設定、工作階段與 Claude Code 完全一致。
+
+- **涵蓋全部工作階段** — CloudCLI UI 會自動掃描 `~/.claude` 資料夾中的每個工作階段。Remote Control 只暴露目前活動的工作階段。
+- **設定統一** — 在 CloudCLI UI 中修改的 MCP、工具權限等設定會立即寫入 Claude Code。
+- **支援更多 Agents** — Claude Code、Cursor CLI、Codex、Gemini CLI。
+- **完整 UI** — 除了聊天介面，還包括檔案瀏覽器、Git 整合、MCP 管理和 Shell 終端機。
+- **CloudCLI Cloud 持續運作於雲端** — 關閉本機裝置也不會中斷代理執行，無需監控終端機。
+
+</details>
+
+<details>
+<summary>需要額外購買 AI 訂閱嗎？</summary>
+
+需要。CloudCLI 只提供環境。你仍需自行取得 Claude、Cursor、Codex 或 Gemini 訂閱。CloudCLI Cloud 從 $7/月起提供託管環境。
+
+</details>
+
+<details>
+<summary>能在手機上使用 CloudCLI UI 嗎？</summary>
+
+可以。自架時，在你的裝置上執行伺服器，然後在網路中的任意瀏覽器開啟 `[yourip]:port`。CloudCLI Cloud 可從任意裝置存取，內建原生應用程式也在開發中。
+
+</details>
+
+<details>
+<summary>UI 中的變更會影響本機 Claude Code 設定嗎？</summary>
+
+會的。自架模式下，CloudCLI UI 讀取並寫入 Claude Code 使用的 `~/.claude` 設定。透過 UI 新增的 MCP 伺服器會立即在 Claude Code 中可見。
+
+</details>
+
+---
+
+## 社群與支援
+
+- **[文件](https://cloudcli.ai/docs)** — 安裝、設定、功能與疑難排解指南
+- **[Discord](https://discord.gg/buxwujPNRE)** — 取得協助並與社群交流
+- **[GitHub Issues](https://github.com/siteboon/claudecodeui/issues)** — 回報 Bug 與建議功能
+- **[貢獻指南](CONTRIBUTING.md)** — 如何參與專案貢獻
+
+## 授權條款
+
+GNU 通用公共授權條款 v3.0 — 詳見 [LICENSE](LICENSE) 檔案。
+
+該專案為開源軟體，在 GPL v3 授權條款下可自由使用、修改與散布。
+
+## 致謝
+
+### 使用技術
+- **[Claude Code](https://docs.anthropic.com/en/docs/claude-code)** — Anthropic 官方 CLI
+- **[Cursor CLI](https://docs.cursor.com/en/cli/overview)** — Cursor 官方 CLI
+- **[Codex](https://developers.openai.com/codex)** — OpenAI Codex
+- **[Gemini-CLI](https://geminicli.com/)** — Google Gemini CLI
+- **[React](https://react.dev/)** — 使用者介面函式庫
+- **[Vite](https://vitejs.dev/)** — 快速建構工具與開發伺服器
+- **[Tailwind CSS](https://tailwindcss.com/)** — 實用優先 CSS 框架
+- **[CodeMirror](https://codemirror.net/)** — 進階程式碼編輯器
+- **[TaskMaster AI](https://github.com/eyaltoledano/claude-task-master)** *(選用)* — AI 驅動的專案管理與任務規劃
+
+### 贊助商
+- [Siteboon - AI powered website builder](https://siteboon.ai)
+---
+
+<div align="center">
+  <strong>為 Claude Code、Cursor 和 Codex 社群精心打造。</strong>
+</div>

docs/nginx-subpath-template.conf

@@ -0,0 +1,218 @@
+# CloudCLI UI Nginx subpath deployment template.
+#
+# Purpose:
+#   Serve CloudCLI UI from a path prefix such as:
+#     http://localhost/ai/
+#     https://example.com/ai/
+#
+# CloudCLI itself still runs at the root of its own HTTP server, for example:
+#     http://127.0.0.1:3001/
+#
+# Nginx receives public requests under /ai, strips that prefix, and forwards the
+# remaining path to CloudCLI. For example:
+#     /ai/                  ->  /
+#     /ai/session/abc       ->  /session/abc
+#     /ai/assets/index.js   ->  /assets/index.js
+#
+# Important Nginx limitation:
+#   Nginx does not allow variables in `location` matchers or `rewrite` regexes.
+#   The configurable variables below are still useful for proxy/filter values,
+#   but if you change /ai to a different subpath, also update every line marked:
+#     [SUBPATH LITERAL]
+#
+# To use a different subpath, replace these literal matchers:
+#     location = /ai
+#     location ^~ /ai/
+#     rewrite ^/ai(?<cloudcli_path>/.*)$ ...
+#
+# Recommended deployment shape:
+#   CloudCLI is the only app using /ai, while root paths /api, /ws, and /shell
+#   are also proxied because the current frontend still calls those endpoints
+#   with root-relative URLs.
+
+worker_processes 1;
+
+events {
+    # Maximum simultaneous connections handled by each worker process.
+    # The default is enough for local testing and small self-hosted deployments.
+    worker_connections 1024;
+}
+
+http {
+    # WebSocket requests include an Upgrade header. Normal HTTP requests do not.
+    # This map gives us the right Connection header for both cases:
+    #   Upgrade present -> "upgrade"
+    #   Upgrade absent  -> "close"
+    map $http_upgrade $connection_upgrade {
+        default upgrade;
+        '' close;
+    }
+
+    server {
+        # For HTTPS deployments, replace this with `listen 443 ssl http2;` and
+        # add ssl_certificate / ssl_certificate_key lines.
+        listen 80 default_server;
+
+        # Use your real hostname in production, for example:
+        #   server_name cloudcli.example.com;
+        server_name localhost 127.0.0.1;
+
+        # ---- User settings -------------------------------------------------
+        #
+        # Public path prefix where users access CloudCLI.
+        # Do not add a trailing slash.
+        #
+        # This variable can be used in redirects and response rewrites. It
+        # cannot be used in `location` matchers, so update the [SUBPATH LITERAL]
+        # lines too if you change it.
+        set $cloudcli_subpath /ai;
+
+        # Private upstream URL where the CloudCLI server is listening.
+        # For a default local server this is usually http://127.0.0.1:3001.
+        set $cloudcli_upstream http://127.0.0.1:3001;
+
+        # Allow larger file uploads through the code editor/project file APIs.
+        client_max_body_size 200m;
+
+        # Redirect /ai to /ai/ so relative browser URL resolution is stable.
+        # [SUBPATH LITERAL] Change `/ai` if you change $cloudcli_subpath.
+        location = /ai {
+            return 301 $cloudcli_subpath/;
+        }
+
+        # Main prefixed CloudCLI UI route.
+        #
+        # [SUBPATH LITERAL] Change `/ai/` and the `^/ai` rewrite if you change
+        # $cloudcli_subpath.
+        location ^~ /ai/ {
+            # Strip the public subpath before proxying. CloudCLI expects to see
+            # root paths such as /, /session/:id, /assets/..., /manifest.json.
+            rewrite ^/ai(?<cloudcli_path>/.*)$ $cloudcli_path break;
+
+            # Forward the rewritten request to the private CloudCLI server.
+            proxy_pass $cloudcli_upstream;
+
+            # Use HTTP/1.1 so WebSocket upgrade requests can pass through if a
+            # browser reaches a socket endpoint under the subpath.
+            proxy_http_version 1.1;
+
+            # Preserve useful request metadata for logs and future app support.
+            proxy_set_header Host $host;
+            proxy_set_header X-Real-IP $remote_addr;
+            proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
+            proxy_set_header X-Forwarded-Proto $scheme;
+            proxy_set_header X-Forwarded-Prefix $cloudcli_subpath;
+
+            # WebSocket upgrade headers. Harmless for normal HTTP requests.
+            proxy_set_header Upgrade $http_upgrade;
+            proxy_set_header Connection $connection_upgrade;
+
+            # Long-running agent and terminal sessions can stay open for a long
+            # time, so avoid closing idle proxied connections too aggressively.
+            proxy_read_timeout 3600s;
+            proxy_send_timeout 3600s;
+
+            # Disable gzip from the upstream response so sub_filter can inspect
+            # and rewrite HTML/JSON/JS response bodies.
+            proxy_set_header Accept-Encoding "";
+
+            # Rewrite browser-visible root-relative URLs so the runtime can
+            # discover that the app is mounted under the subpath.
+            #
+            # Examples:
+            #   href="/manifest.json" -> href="/ai/manifest.json"
+            #   src="/assets/app.js"  -> src="/ai/assets/app.js"
+            #
+            # These rewrites are important for React Router basename detection.
+            sub_filter_once off;
+            sub_filter_types
+                application/json
+                application/manifest+json
+                application/javascript
+                text/javascript;
+
+            sub_filter 'href="/' 'href="$cloudcli_subpath/';
+            sub_filter 'src="/' 'src="$cloudcli_subpath/';
+
+            # The production HTML and JS register the service worker at /sw.js.
+            # Rewrite that registration so the worker is served from /ai/sw.js.
+            sub_filter "register('/sw.js')" "register('$cloudcli_subpath/sw.js')";
+            sub_filter 'register("/sw.js")' 'register("$cloudcli_subpath/sw.js")';
+
+            # The manifest and service worker contain root-relative paths too.
+            # Rewriting them keeps PWA metadata and cached manifest requests
+            # under the same public subpath.
+            sub_filter '"start_url": "/"' '"start_url": "$cloudcli_subpath/"';
+            sub_filter '"scope": "/"' '"scope": "$cloudcli_subpath/"';
+            sub_filter '"src": "/' '"src": "$cloudcli_subpath/';
+            sub_filter "'/manifest.json'" "'$cloudcli_subpath/manifest.json'";
+            sub_filter '"/manifest.json"' '"$cloudcli_subpath/manifest.json"';
+        }
+
+        # Root API proxy.
+        #
+        # The current CloudCLI frontend calls APIs with root-relative URLs such
+        # as /api/auth/login. Keep this location unless the frontend becomes
+        # fully prefix-aware for API requests.
+        location ^~ /api/ {
+            proxy_pass $cloudcli_upstream;
+
+            proxy_http_version 1.1;
+            proxy_set_header Host $host;
+            proxy_set_header X-Real-IP $remote_addr;
+            proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
+            proxy_set_header X-Forwarded-Proto $scheme;
+            proxy_set_header X-Forwarded-Prefix $cloudcli_subpath;
+
+            proxy_read_timeout 3600s;
+            proxy_send_timeout 3600s;
+        }
+
+        # Main app WebSocket proxy.
+        #
+        # The frontend opens /ws for realtime chat/session/task updates.
+        location /ws {
+            proxy_pass $cloudcli_upstream;
+
+            proxy_http_version 1.1;
+            proxy_set_header Host $host;
+            proxy_set_header Upgrade $http_upgrade;
+            proxy_set_header Connection $connection_upgrade;
+            proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
+            proxy_set_header X-Forwarded-Proto $scheme;
+            proxy_set_header X-Forwarded-Prefix $cloudcli_subpath;
+
+            proxy_read_timeout 3600s;
+            proxy_send_timeout 3600s;
+        }
+
+        # Shell WebSocket proxy.
+        #
+        # The browser terminal uses /shell. It requires the same WebSocket
+        # upgrade handling as /ws.
+        location /shell {
+            proxy_pass $cloudcli_upstream;
+
+            proxy_http_version 1.1;
+            proxy_set_header Host $host;
+            proxy_set_header Upgrade $http_upgrade;
+            proxy_set_header Connection $connection_upgrade;
+            proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
+            proxy_set_header X-Forwarded-Proto $scheme;
+            proxy_set_header X-Forwarded-Prefix $cloudcli_subpath;
+
+            proxy_read_timeout 3600s;
+            proxy_send_timeout 3600s;
+        }
+
+        # Optional health endpoint proxy used by the frontend version checker.
+        location = /health {
+            proxy_pass $cloudcli_upstream;
+
+            proxy_set_header Host $host;
+            proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
+            proxy_set_header X-Forwarded-Proto $scheme;
+            proxy_set_header X-Forwarded-Prefix $cloudcli_subpath;
+        }
+    }
+}

package-lock.json

@@ -1,16 +1,16 @@
 {
   "name": "@cloudcli-ai/cloudcli",
-  "version": "1.33.0",
+  "version": "1.34.0",
   "lockfileVersion": 3,
   "requires": true,
   "packages": {
     "": {
       "name": "@cloudcli-ai/cloudcli",
-      "version": "1.33.0",
+      "version": "1.34.0",
       "hasInstallScript": true,
       "license": "AGPL-3.0-or-later",
       "dependencies": {
-        "@anthropic-ai/claude-agent-sdk": "^0.2.116",
+        "@anthropic-ai/claude-agent-sdk": "^0.3.165",
         "@codemirror/lang-css": "^6.3.1",
         "@codemirror/lang-html": "^6.4.9",
         "@codemirror/lang-javascript": "^6.2.4",
@@ -139,35 +139,33 @@
       }
     },
     "node_modules/@anthropic-ai/claude-agent-sdk": {
-      "version": "0.2.116",
-      "resolved": "https://registry.npmjs.org/@anthropic-ai/claude-agent-sdk/-/claude-agent-sdk-0.2.116.tgz",
-      "integrity": "sha512-5NKpgaOZkzNCGCvLxJZUVGimf5IcYmpQ2x2XrR9ilK+2UkWrnnwcUfIWo8bBz9e7lSYcUf9XleGigq2eOOF7aw==",
+      "version": "0.3.165",
+      "resolved": "https://registry.npmjs.org/@anthropic-ai/claude-agent-sdk/-/claude-agent-sdk-0.3.165.tgz",
+      "integrity": "sha512-wEUJNTAWkE6KMV35abqGi30lwhZz+jQLMtLh4SuTN2Hllzsysq8kmQFgcWulza3FLHG/GHzGHPi0+Sp2fb8xlw==",
       "license": "SEE LICENSE IN README.md",
-      "dependencies": {
-        "@anthropic-ai/sdk": "^0.81.0",
-        "@modelcontextprotocol/sdk": "^1.29.0"
-      },
       "engines": {
         "node": ">=18.0.0"
       },
       "optionalDependencies": {
-        "@anthropic-ai/claude-agent-sdk-darwin-arm64": "0.2.116",
-        "@anthropic-ai/claude-agent-sdk-darwin-x64": "0.2.116",
-        "@anthropic-ai/claude-agent-sdk-linux-arm64": "0.2.116",
-        "@anthropic-ai/claude-agent-sdk-linux-arm64-musl": "0.2.116",
-        "@anthropic-ai/claude-agent-sdk-linux-x64": "0.2.116",
-        "@anthropic-ai/claude-agent-sdk-linux-x64-musl": "0.2.116",
-        "@anthropic-ai/claude-agent-sdk-win32-arm64": "0.2.116",
-        "@anthropic-ai/claude-agent-sdk-win32-x64": "0.2.116"
+        "@anthropic-ai/claude-agent-sdk-darwin-arm64": "0.3.165",
+        "@anthropic-ai/claude-agent-sdk-darwin-x64": "0.3.165",
+        "@anthropic-ai/claude-agent-sdk-linux-arm64": "0.3.165",
+        "@anthropic-ai/claude-agent-sdk-linux-arm64-musl": "0.3.165",
+        "@anthropic-ai/claude-agent-sdk-linux-x64": "0.3.165",
+        "@anthropic-ai/claude-agent-sdk-linux-x64-musl": "0.3.165",
+        "@anthropic-ai/claude-agent-sdk-win32-arm64": "0.3.165",
+        "@anthropic-ai/claude-agent-sdk-win32-x64": "0.3.165"
       },
       "peerDependencies": {
+        "@anthropic-ai/sdk": ">=0.93.0",
+        "@modelcontextprotocol/sdk": "^1.29.0",
         "zod": "^4.0.0"
       }
     },
     "node_modules/@anthropic-ai/claude-agent-sdk-darwin-arm64": {
-      "version": "0.2.116",
-      "resolved": "https://registry.npmjs.org/@anthropic-ai/claude-agent-sdk-darwin-arm64/-/claude-agent-sdk-darwin-arm64-0.2.116.tgz",
-      "integrity": "sha512-mG19ovtXCpETmd5KmTU1JO2iIHZBG09IP8DmgZjLA3wLmTzpgn9Au9veRaeJeXb1EqiHiFZU+z+mNB79+w5v9g==",
+      "version": "0.3.165",
+      "resolved": "https://registry.npmjs.org/@anthropic-ai/claude-agent-sdk-darwin-arm64/-/claude-agent-sdk-darwin-arm64-0.3.165.tgz",
+      "integrity": "sha512-obVodJmppNc6lgcM6Y5y3VCQLrYO2curOXrRaziKtjxYbuZP7kYsUhnonMvGoVAQh3uHKz2tivQDeztvWe3f9w==",
       "cpu": [
         "arm64"
       ],
@@ -178,9 +176,9 @@
       ]
     },
     "node_modules/@anthropic-ai/claude-agent-sdk-darwin-x64": {
-      "version": "0.2.116",
-      "resolved": "https://registry.npmjs.org/@anthropic-ai/claude-agent-sdk-darwin-x64/-/claude-agent-sdk-darwin-x64-0.2.116.tgz",
-      "integrity": "sha512-qC25N0HRM8IXbM4Qi4svH9f51Y6DciDvjLV+oNYnxkdPgDG8p/+b7vQirN7qPxytIQb2TPdoFgUeCsSe7lrQyw==",
+      "version": "0.3.165",
+      "resolved": "https://registry.npmjs.org/@anthropic-ai/claude-agent-sdk-darwin-x64/-/claude-agent-sdk-darwin-x64-0.3.165.tgz",
+      "integrity": "sha512-0jc1tlYLXzPvZIkHKGHzsEEKq2YqTS8oHSNFroqLgbhrIk1Zy05ZXbciI289VDAe1Fq2a+qcUhkXct8Parx1Rg==",
       "cpu": [
         "x64"
       ],
@@ -191,9 +189,9 @@
       ]
     },
     "node_modules/@anthropic-ai/claude-agent-sdk-linux-arm64": {
-      "version": "0.2.116",
-      "resolved": "https://registry.npmjs.org/@anthropic-ai/claude-agent-sdk-linux-arm64/-/claude-agent-sdk-linux-arm64-0.2.116.tgz",
-      "integrity": "sha512-MQIcJhhPM+RPJ7kMQdOQarkJ2FlJqOiu953c08YyJOoWdHykd3DIiHws3mf1Mwl/dfFeIyshOVpNND3hyIy5Dg==",
+      "version": "0.3.165",
+      "resolved": "https://registry.npmjs.org/@anthropic-ai/claude-agent-sdk-linux-arm64/-/claude-agent-sdk-linux-arm64-0.3.165.tgz",
+      "integrity": "sha512-t87HgDPPaRYMTTB5cqA0M36Fyq4DOny89yk71BMgA8hAzhOjV9bla8pMVZTuX3xYYPjsa/TOmxSzwI8GZLf4Aw==",
       "cpu": [
         "arm64"
       ],
@@ -204,9 +202,9 @@
       ]
     },
     "node_modules/@anthropic-ai/claude-agent-sdk-linux-arm64-musl": {
-      "version": "0.2.116",
-      "resolved": "https://registry.npmjs.org/@anthropic-ai/claude-agent-sdk-linux-arm64-musl/-/claude-agent-sdk-linux-arm64-musl-0.2.116.tgz",
-      "integrity": "sha512-Dg/T3NkSp35ODiwdhj0KquvC6Xu+DMbyWFNkfepA3bz4oF2SVSgyOPYwVmfoJerzEUnYDldP4YhOxRrhbt0vXA==",
+      "version": "0.3.165",
+      "resolved": "https://registry.npmjs.org/@anthropic-ai/claude-agent-sdk-linux-arm64-musl/-/claude-agent-sdk-linux-arm64-musl-0.3.165.tgz",
+      "integrity": "sha512-Rccmr5chZdZJVRvoB0nildB5PTKX+amatUho9JIcNOf1iX/6ej39fwf8q9W1MRHYP7AEc4t9GrSAGLcn7/JO4w==",
       "cpu": [
         "arm64"
       ],
@@ -217,9 +215,9 @@
       ]
     },
     "node_modules/@anthropic-ai/claude-agent-sdk-linux-x64": {
-      "version": "0.2.116",
-      "resolved": "https://registry.npmjs.org/@anthropic-ai/claude-agent-sdk-linux-x64/-/claude-agent-sdk-linux-x64-0.2.116.tgz",
-      "integrity": "sha512-Bww1fzQB+vcF0tRhmCAlwSsN4wR2HgX7pBT9AWuwzJj6DKsVC23N54Ea80lsnM7dTUtUTrGYMTwVUHTWqfYnfQ==",
+      "version": "0.3.165",
+      "resolved": "https://registry.npmjs.org/@anthropic-ai/claude-agent-sdk-linux-x64/-/claude-agent-sdk-linux-x64-0.3.165.tgz",
+      "integrity": "sha512-Y8fEW0zKBn0XZI5AOQWHep0Srz0qsCauynTWkhsC6J2vSPxkTiOxv2hmb7qdfiNlFn0k1etCWVFoRkhhFJzGfg==",
       "cpu": [
         "x64"
       ],
@@ -230,9 +228,9 @@
       ]
     },
     "node_modules/@anthropic-ai/claude-agent-sdk-linux-x64-musl": {
-      "version": "0.2.116",
-      "resolved": "https://registry.npmjs.org/@anthropic-ai/claude-agent-sdk-linux-x64-musl/-/claude-agent-sdk-linux-x64-musl-0.2.116.tgz",
-      "integrity": "sha512-LMYxUMa1nK4N9BPRJdcGBAvl9rjTI4ZHo+kfAKrJ3MlfB6VFF1tRIubwsWOaOtkuNazMdAYovsZJg4bdzOBBTQ==",
+      "version": "0.3.165",
+      "resolved": "https://registry.npmjs.org/@anthropic-ai/claude-agent-sdk-linux-x64-musl/-/claude-agent-sdk-linux-x64-musl-0.3.165.tgz",
+      "integrity": "sha512-Y9Acr1RmydfEX+t+3mFn0K9VOx6nfyo08QuQH9R6ap1YYZWuobze++pNUY/rzwbQjXqcbjORtPKbO/kLQtSr9w==",
       "cpu": [
         "x64"
       ],
@@ -243,9 +241,9 @@
       ]
     },
     "node_modules/@anthropic-ai/claude-agent-sdk-win32-arm64": {
-      "version": "0.2.116",
-      "resolved": "https://registry.npmjs.org/@anthropic-ai/claude-agent-sdk-win32-arm64/-/claude-agent-sdk-win32-arm64-0.2.116.tgz",
-      "integrity": "sha512-h0YO1vkTIeUtffQhONrYbNC1pXmk1yjb1xxMEw7bAwucqtFoFpLDWe+q4+RhxaQr8ZOj6LtRE/U3dzPWHOlshA==",
+      "version": "0.3.165",
+      "resolved": "https://registry.npmjs.org/@anthropic-ai/claude-agent-sdk-win32-arm64/-/claude-agent-sdk-win32-arm64-0.3.165.tgz",
+      "integrity": "sha512-4Q01L3xaDDCvlOhABf2MnO7v7yJxKwwDyiMr+DaneUSvuh1qH0YE7qErSYLf6D9VfH8TdRwKZXwQplVVwCoHWw==",
       "cpu": [
         "arm64"
       ],
@@ -256,9 +254,9 @@
       ]
     },
     "node_modules/@anthropic-ai/claude-agent-sdk-win32-x64": {
-      "version": "0.2.116",
-      "resolved": "https://registry.npmjs.org/@anthropic-ai/claude-agent-sdk-win32-x64/-/claude-agent-sdk-win32-x64-0.2.116.tgz",
-      "integrity": "sha512-3lllmtDFHgpW0ZM3iNvxsEjblrgRzF9Qm1lxTOtunP3hIn+pA/IkWMtKlN1ixxWiaBguLVQkJ90V6JHsvJJIvw==",
+      "version": "0.3.165",
+      "resolved": "https://registry.npmjs.org/@anthropic-ai/claude-agent-sdk-win32-x64/-/claude-agent-sdk-win32-x64-0.3.165.tgz",
+      "integrity": "sha512-Y0uOx7b7ZnkguvFFI5T5fSLnRA/e0uvMC++gSnyz6XMpNekgWc3+Mny7Dv2NO22nKbV2YiFsj6MkYYFEd51BDw==",
       "cpu": [
         "x64"
       ],
@@ -269,12 +267,14 @@
       ]
     },
     "node_modules/@anthropic-ai/sdk": {
-      "version": "0.81.0",
-      "resolved": "https://registry.npmjs.org/@anthropic-ai/sdk/-/sdk-0.81.0.tgz",
-      "integrity": "sha512-D4K5PvEV6wPiRtVlVsJHIUhHAmOZ6IT/I9rKlTf84gR7GyyAurPJK7z9BOf/AZqC5d1DhYQGJNKRmV+q8dGhgw==",
+      "version": "0.100.1",
+      "resolved": "https://registry.npmjs.org/@anthropic-ai/sdk/-/sdk-0.100.1.tgz",
+      "integrity": "sha512-RANcEe7LpiLczkKGOwoXOTuFdPhuubS0i4xaAKOMpcqc55YO0mukgxppV7eygx3DXNjxWT6RYOLPyOy0aIAmwg==",
       "license": "MIT",
+      "peer": true,
       "dependencies": {
-        "json-schema-to-ts": "^3.1.1"
+        "json-schema-to-ts": "^3.1.1",
+        "standardwebhooks": "^1.0.0"
       },
       "bin": {
         "anthropic-ai-sdk": "bin/cli"
@@ -1837,6 +1837,7 @@
       "resolved": "https://registry.npmjs.org/@hono/node-server/-/node-server-1.19.14.tgz",
       "integrity": "sha512-GwtvgtXxnWsucXvbQXkRgqksiH2Qed37H9xHZocE5sA3N8O8O8/8FA3uclQXxXVzc9XBZuEOMK7+r02FmSpHtw==",
       "license": "MIT",
+      "peer": true,
       "engines": {
         "node": ">=18.14.1"
       },
@@ -2601,6 +2602,7 @@
       "resolved": "https://registry.npmjs.org/@modelcontextprotocol/sdk/-/sdk-1.29.0.tgz",
       "integrity": "sha512-zo37mZA9hJWpULgkRpowewez1y6ML5GsXJPY8FI0tBBCd77HEvza4jDqRKOXgHNn867PVGCyTdzqpz0izu5ZjQ==",
       "license": "MIT",
+      "peer": true,
       "dependencies": {
         "@hono/node-server": "^1.19.9",
         "ajv": "^8.17.1",
@@ -2641,6 +2643,7 @@
       "resolved": "https://registry.npmjs.org/accepts/-/accepts-2.0.0.tgz",
       "integrity": "sha512-5cvg6CtKwfgdmVqY1WIiXKc3Q1bkRqGLi+2W/6ao+6Y7gu/RCwRuAhGEzh5B4KlszSuTLgZYuqFqo5bImjNKng==",
       "license": "MIT",
+      "peer": true,
       "dependencies": {
         "mime-types": "^3.0.0",
         "negotiator": "^1.0.0"
@@ -2650,10 +2653,11 @@
       }
     },
     "node_modules/@modelcontextprotocol/sdk/node_modules/ajv": {
-      "version": "8.18.0",
-      "resolved": "https://registry.npmjs.org/ajv/-/ajv-8.18.0.tgz",
-      "integrity": "sha512-PlXPeEWMXMZ7sPYOHqmDyCJzcfNrUr3fGNKtezX14ykXOEIvyK81d+qydx89KY5O71FKMPaQ2vBfBFI5NHR63A==",
+      "version": "8.20.0",
+      "resolved": "https://registry.npmjs.org/ajv/-/ajv-8.20.0.tgz",
+      "integrity": "sha512-Thbli+OlOj+iMPYFBVBfJ3OmCAnaSyNn4M1vz9T6Gka5Jt9ba/HIR56joy65tY6kx/FCF5VXNB819Y7/GUrBGA==",
       "license": "MIT",
+      "peer": true,
       "dependencies": {
         "fast-deep-equal": "^3.1.3",
         "fast-uri": "^3.0.1",
@@ -2670,6 +2674,7 @@
       "resolved": "https://registry.npmjs.org/body-parser/-/body-parser-2.2.2.tgz",
       "integrity": "sha512-oP5VkATKlNwcgvxi0vM0p/D3n2C3EReYVX+DNYs5TjZFn/oQt2j+4sVJtSMr18pdRr8wjTcBl6LoV+FUwzPmNA==",
       "license": "MIT",
+      "peer": true,
       "dependencies": {
         "bytes": "^3.1.2",
         "content-type": "^1.0.5",
@@ -2694,6 +2699,7 @@
       "resolved": "https://registry.npmjs.org/content-disposition/-/content-disposition-1.1.0.tgz",
       "integrity": "sha512-5jRCH9Z/+DRP7rkvY83B+yGIGX96OYdJmzngqnw2SBSxqCFPd0w2km3s5iawpGX8krnwSGmF0FW5Nhr0Hfai3g==",
       "license": "MIT",
+      "peer": true,
       "engines": {
         "node": ">=18"
       },
@@ -2707,6 +2713,7 @@
       "resolved": "https://registry.npmjs.org/cookie-signature/-/cookie-signature-1.2.2.tgz",
       "integrity": "sha512-D76uU73ulSXrD1UXF4KE2TMxVVwhsnCgfAyTg9k8P6KGZjlXKrOLe4dJQKI3Bxi5wjesZoFXJWElNWBjPZMbhg==",
       "license": "MIT",
+      "peer": true,
       "engines": {
         "node": ">=6.6.0"
       }
@@ -2716,6 +2723,7 @@
       "resolved": "https://registry.npmjs.org/express/-/express-5.2.1.tgz",
       "integrity": "sha512-hIS4idWWai69NezIdRt2xFVofaF4j+6INOpJlVOLDO8zXGpUVEVzIYk12UUi2JzjEzWL3IOAxcTubgz9Po0yXw==",
       "license": "MIT",
+      "peer": true,
       "dependencies": {
         "accepts": "^2.0.0",
         "body-parser": "^2.2.1",
@@ -2759,6 +2767,7 @@
       "resolved": "https://registry.npmjs.org/finalhandler/-/finalhandler-2.1.1.tgz",
       "integrity": "sha512-S8KoZgRZN+a5rNwqTxlZZePjT/4cnm0ROV70LedRHZ0p8u9fRID0hJUZQpkKLzro8LfmC8sx23bY6tVNxv8pQA==",
       "license": "MIT",
+      "peer": true,
       "dependencies": {
         "debug": "^4.4.0",
         "encodeurl": "^2.0.0",
@@ -2780,6 +2789,7 @@
       "resolved": "https://registry.npmjs.org/fresh/-/fresh-2.0.0.tgz",
       "integrity": "sha512-Rx/WycZ60HOaqLKAi6cHRKKI7zxWbJ31MhntmtwMoaTeF7XFH9hhBp8vITaMidfljRQ6eYWCKkaTK+ykVJHP2A==",
       "license": "MIT",
+      "peer": true,
       "engines": {
         "node": ">= 0.8"
       }
@@ -2789,6 +2799,7 @@
       "resolved": "https://registry.npmjs.org/http-errors/-/http-errors-2.0.1.tgz",
       "integrity": "sha512-4FbRdAX+bSdmo4AUFuS0WNiPz8NgFt+r8ThgNWmlrjQjt1Q7ZR9+zTlce2859x4KSXrwIsaeTqDoKQmtP8pLmQ==",
       "license": "MIT",
+      "peer": true,
       "dependencies": {
         "depd": "~2.0.0",
         "inherits": "~2.0.4",
@@ -2809,6 +2820,7 @@
       "resolved": "https://registry.npmjs.org/iconv-lite/-/iconv-lite-0.7.2.tgz",
       "integrity": "sha512-im9DjEDQ55s9fL4EYzOAv0yMqmMBSZp6G0VvFyTMPKWxiSBHUj9NW/qqLmXUwXrrM7AvqSlTCfvqRb0cM8yYqw==",
       "license": "MIT",
+      "peer": true,
       "dependencies": {
         "safer-buffer": ">= 2.1.2 < 3.0.0"
       },
@@ -2824,13 +2836,15 @@
       "version": "1.0.0",
       "resolved": "https://registry.npmjs.org/json-schema-traverse/-/json-schema-traverse-1.0.0.tgz",
       "integrity": "sha512-NM8/P9n3XjXhIZn1lLhkFaACTOURQXjWhV4BA/RnOv8xvgqtqpAX9IO4mRQxSx1Rlo4tqzeqb0sOlruaOy3dug==",
-      "license": "MIT"
+      "license": "MIT",
+      "peer": true
     },
     "node_modules/@modelcontextprotocol/sdk/node_modules/media-typer": {
       "version": "1.1.0",
       "resolved": "https://registry.npmjs.org/media-typer/-/media-typer-1.1.0.tgz",
       "integrity": "sha512-aisnrDP4GNe06UcKFnV5bfMNPBUw4jsLGaWwWfnH3v02GnBuXX2MCVn5RbrWo0j3pczUilYblq7fQ7Nw2t5XKw==",
       "license": "MIT",
+      "peer": true,
       "engines": {
         "node": ">= 0.8"
       }
@@ -2840,6 +2854,7 @@
       "resolved": "https://registry.npmjs.org/merge-descriptors/-/merge-descriptors-2.0.0.tgz",
       "integrity": "sha512-Snk314V5ayFLhp3fkUREub6WtjBfPdCPY1Ln8/8munuLuiYhsABgBVWsozAG+MWMbVEvcdcpbi9R7ww22l9Q3g==",
       "license": "MIT",
+      "peer": true,
       "engines": {
         "node": ">=18"
       },
@@ -2852,6 +2867,7 @@
       "resolved": "https://registry.npmjs.org/mime-types/-/mime-types-3.0.2.tgz",
       "integrity": "sha512-Lbgzdk0h4juoQ9fCKXW4by0UJqj+nOOrI9MJ1sSj4nI8aI2eo1qmvQEie4VD1glsS250n15LsWsYtCugiStS5A==",
       "license": "MIT",
+      "peer": true,
       "dependencies": {
         "mime-db": "^1.54.0"
       },
@@ -2868,15 +2884,17 @@
       "resolved": "https://registry.npmjs.org/negotiator/-/negotiator-1.0.0.tgz",
       "integrity": "sha512-8Ofs/AUQh8MaEcrlq5xOX0CQ9ypTF5dl78mjlMNfOK08fzpgTHQRQPBxcPlEtIw0yRpws+Zo/3r+5WRby7u3Gg==",
       "license": "MIT",
+      "peer": true,
       "engines": {
         "node": ">= 0.6"
       }
     },
     "node_modules/@modelcontextprotocol/sdk/node_modules/qs": {
-      "version": "6.15.1",
-      "resolved": "https://registry.npmjs.org/qs/-/qs-6.15.1.tgz",
-      "integrity": "sha512-6YHEFRL9mfgcAvql/XhwTvf5jKcOiiupt2FiJxHkiX1z4j7WL8J/jRHYLluORvc1XxB5rV20KoeK00gVJamspg==",
+      "version": "6.15.2",
+      "resolved": "https://registry.npmjs.org/qs/-/qs-6.15.2.tgz",
+      "integrity": "sha512-Rzq0KEyX/w/tEybncDgdkZrJgVUsUMk3xjh3t5bv3S1HTAtg+uOYt72+ZfwiQwKdysThkTBdL/rTi6HDmX9Ddw==",
       "license": "BSD-3-Clause",
+      "peer": true,
       "dependencies": {
         "side-channel": "^1.1.0"
       },
@@ -2892,6 +2910,7 @@
       "resolved": "https://registry.npmjs.org/raw-body/-/raw-body-3.0.2.tgz",
       "integrity": "sha512-K5zQjDllxWkf7Z5xJdV0/B0WTNqx6vxG70zJE4N0kBs4LovmEYWJzQGxC9bS9RAKu3bgM40lrd5zoLJ12MQ5BA==",
       "license": "MIT",
+      "peer": true,
       "dependencies": {
         "bytes": "~3.1.2",
         "http-errors": "~2.0.1",
@@ -2907,6 +2926,7 @@
       "resolved": "https://registry.npmjs.org/send/-/send-1.2.1.tgz",
       "integrity": "sha512-1gnZf7DFcoIcajTjTwjwuDjzuz4PPcY2StKPlsGAQ1+YH20IRVrBaXSWmdjowTJ6u8Rc01PoYOGHXfP1mYcZNQ==",
       "license": "MIT",
+      "peer": true,
       "dependencies": {
         "debug": "^4.4.3",
         "encodeurl": "^2.0.0",
@@ -2933,6 +2953,7 @@
       "resolved": "https://registry.npmjs.org/serve-static/-/serve-static-2.2.1.tgz",
       "integrity": "sha512-xRXBn0pPqQTVQiC8wyQrKs2MOlX24zQ0POGaj0kultvoOCstBQM5yvOhAVSUwOMjQtTvsPWoNCHfPGwaaQJhTw==",
       "license": "MIT",
+      "peer": true,
       "dependencies": {
         "encodeurl": "^2.0.0",
         "escape-html": "^1.0.3",
@@ -2952,22 +2973,42 @@
       "resolved": "https://registry.npmjs.org/statuses/-/statuses-2.0.2.tgz",
       "integrity": "sha512-DvEy55V3DB7uknRo+4iOGT5fP1slR8wQohVdknigZPMpMstaKJQWhwiYBACJE3Ul2pTnATihhBYnRhZQHGBiRw==",
       "license": "MIT",
+      "peer": true,
       "engines": {
         "node": ">= 0.8"
       }
     },
     "node_modules/@modelcontextprotocol/sdk/node_modules/type-is": {
-      "version": "2.0.1",
-      "resolved": "https://registry.npmjs.org/type-is/-/type-is-2.0.1.tgz",
-      "integrity": "sha512-OZs6gsjF4vMp32qrCbiVSkrFmXtG/AZhY3t0iAMrMBiAZyV9oALtXO8hsrHbMXF9x6L3grlFuwW2oAz7cav+Gw==",
+      "version": "2.1.0",
+      "resolved": "https://registry.npmjs.org/type-is/-/type-is-2.1.0.tgz",
+      "integrity": "sha512-faYHw0anBbc/kWF3zFTEnxSFOAGUX9GFbOBthvDdLsIlEoWOFOtS0zgCiQYwIskL9iGXZL3kAXD8OoZ4GmMATA==",
       "license": "MIT",
+      "peer": true,
       "dependencies": {
-        "content-type": "^1.0.5",
+        "content-type": "^2.0.0",
         "media-typer": "^1.1.0",
         "mime-types": "^3.0.0"
       },
       "engines": {
-        "node": ">= 0.6"
+        "node": ">= 18"
+      },
+      "funding": {
+        "type": "opencollective",
+        "url": "https://opencollective.com/express"
+      }
+    },
+    "node_modules/@modelcontextprotocol/sdk/node_modules/type-is/node_modules/content-type": {
+      "version": "2.0.0",
+      "resolved": "https://registry.npmjs.org/content-type/-/content-type-2.0.0.tgz",
+      "integrity": "sha512-j/O/d7GcZCyNl7/hwZAb606rzqkyvaDctLmckbxLzHvFBzTJHuGEdodATcP3yIRoDrLHkIATJuvzbFlp/ki2cQ==",
+      "license": "MIT",
+      "peer": true,
+      "engines": {
+        "node": ">=18"
+      },
+      "funding": {
+        "type": "opencollective",
+        "url": "https://opencollective.com/express"
       }
     },
     "node_modules/@napi-rs/wasm-runtime": {
@@ -4284,6 +4325,13 @@
         "url": "https://ko-fi.com/dangreen"
       }
     },
+    "node_modules/@stablelib/base64": {
+      "version": "1.0.1",
+      "resolved": "https://registry.npmjs.org/@stablelib/base64/-/base64-1.0.1.tgz",
+      "integrity": "sha512-1bnPQqSxSuc3Ii6MhBysoWCg58j97aUjuCSZrGSmDxNqtytIi0k8utUenAwTZN4V5mXXYGsVUI9zeBqy+jBOSQ==",
+      "license": "MIT",
+      "peer": true
+    },
     "node_modules/@tailwindcss/typography": {
       "version": "0.5.16",
       "resolved": "https://registry.npmjs.org/@tailwindcss/typography/-/typography-0.5.16.tgz",
@@ -5417,6 +5465,7 @@
       "resolved": "https://registry.npmjs.org/ajv-formats/-/ajv-formats-3.0.1.tgz",
       "integrity": "sha512-8iUql50EUR+uUcdRQ3HDqa6EVyo3docL8g5WJ3FNcWmu62IbkGUue/pEyLBW8VGKKucTPgqeks4fIU1DA4yowQ==",
       "license": "MIT",
+      "peer": true,
       "dependencies": {
         "ajv": "^8.0.0"
       },
@@ -5430,10 +5479,11 @@
       }
     },
     "node_modules/ajv-formats/node_modules/ajv": {
-      "version": "8.18.0",
-      "resolved": "https://registry.npmjs.org/ajv/-/ajv-8.18.0.tgz",
-      "integrity": "sha512-PlXPeEWMXMZ7sPYOHqmDyCJzcfNrUr3fGNKtezX14ykXOEIvyK81d+qydx89KY5O71FKMPaQ2vBfBFI5NHR63A==",
+      "version": "8.20.0",
+      "resolved": "https://registry.npmjs.org/ajv/-/ajv-8.20.0.tgz",
+      "integrity": "sha512-Thbli+OlOj+iMPYFBVBfJ3OmCAnaSyNn4M1vz9T6Gka5Jt9ba/HIR56joy65tY6kx/FCF5VXNB819Y7/GUrBGA==",
       "license": "MIT",
+      "peer": true,
       "dependencies": {
         "fast-deep-equal": "^3.1.3",
         "fast-uri": "^3.0.1",
@@ -5449,7 +5499,8 @@
       "version": "1.0.0",
       "resolved": "https://registry.npmjs.org/json-schema-traverse/-/json-schema-traverse-1.0.0.tgz",
       "integrity": "sha512-NM8/P9n3XjXhIZn1lLhkFaACTOURQXjWhV4BA/RnOv8xvgqtqpAX9IO4mRQxSx1Rlo4tqzeqb0sOlruaOy3dug==",
-      "license": "MIT"
+      "license": "MIT",
+      "peer": true
     },
     "node_modules/ansi-escapes": {
       "version": "7.3.0",
@@ -8515,6 +8566,7 @@
       "resolved": "https://registry.npmjs.org/eventsource/-/eventsource-3.0.7.tgz",
       "integrity": "sha512-CRT1WTyuQoD771GW56XEZFQ/ZoSfWid1alKGDYMmkt2yl8UXrVR4pspqWNEcqKvVIzg6PAltWjxcSSPrboA4iA==",
       "license": "MIT",
+      "peer": true,
       "dependencies": {
         "eventsource-parser": "^3.0.1"
       },
@@ -8523,10 +8575,11 @@
       }
     },
     "node_modules/eventsource-parser": {
-      "version": "3.0.8",
-      "resolved": "https://registry.npmjs.org/eventsource-parser/-/eventsource-parser-3.0.8.tgz",
-      "integrity": "sha512-70QWGkr4snxr0OXLRWsFLeRBIRPuQOvt4s8QYjmUlmlkyTZkRqS7EDVRZtzU3TiyDbXSzaOeF0XUKy8PchzukQ==",
+      "version": "3.1.0",
+      "resolved": "https://registry.npmjs.org/eventsource-parser/-/eventsource-parser-3.1.0.tgz",
+      "integrity": "sha512-kJezFj9YFAMLeORyi7aCLxLbD5/qWMQnoMVlVPyHIll7lgRJCc3JVln9Vgl9nwQi0YkMnhdGTMNn7CkRRAptMg==",
       "license": "MIT",
+      "peer": true,
       "engines": {
         "node": ">=18.0.0"
       }
@@ -8634,12 +8687,13 @@
       }
     },
     "node_modules/express-rate-limit": {
-      "version": "8.3.2",
-      "resolved": "https://registry.npmjs.org/express-rate-limit/-/express-rate-limit-8.3.2.tgz",
-      "integrity": "sha512-77VmFeJkO0/rvimEDuUC5H30oqUC4EyOhyGccfqoLebB0oiEYfM7nwPrsDsBL1gsTpwfzX8SFy2MT3TDyRq+bg==",
+      "version": "8.5.2",
+      "resolved": "https://registry.npmjs.org/express-rate-limit/-/express-rate-limit-8.5.2.tgz",
+      "integrity": "sha512-5Kb34ipNX694DH48vN9irak1Qx30nb0PLYHXfJgw4YEjiC3ZEmZJhwOp+VfiCYwFzvFTdB9QkArYS5kXa2cx2A==",
       "license": "MIT",
+      "peer": true,
       "dependencies": {
-        "ip-address": "10.1.0"
+        "ip-address": "^10.2.0"
       },
       "engines": {
         "node": ">= 16"
@@ -8755,6 +8809,13 @@
       "dev": true,
       "license": "MIT"
     },
+    "node_modules/fast-sha256": {
+      "version": "1.3.0",
+      "resolved": "https://registry.npmjs.org/fast-sha256/-/fast-sha256-1.3.0.tgz",
+      "integrity": "sha512-n11RGP/lrWEFI/bWdygLxhI+pVeo1ZYIVwvvPkW7azl/rOy+F3HYRZ2K5zeE9mmkhQppyv9sQFx0JM9UabnpPQ==",
+      "license": "Unlicense",
+      "peer": true
+    },
     "node_modules/fast-uri": {
       "version": "3.1.0",
       "resolved": "https://registry.npmjs.org/fast-uri/-/fast-uri-3.1.0.tgz",
@@ -9812,10 +9873,11 @@
       "license": "CC0-1.0"
     },
     "node_modules/hono": {
-      "version": "4.12.14",
-      "resolved": "https://registry.npmjs.org/hono/-/hono-4.12.14.tgz",
-      "integrity": "sha512-am5zfg3yu6sqn5yjKBNqhnTX7Cv+m00ox+7jbaKkrLMRJ4rAdldd1xPd/JzbBWspqaQv6RSTrgFN95EsfhC+7w==",
+      "version": "4.12.23",
+      "resolved": "https://registry.npmjs.org/hono/-/hono-4.12.23.tgz",
+      "integrity": "sha512-eIaZ9qDgu7XV0pxOCrg7/WhnQ6Ivm22UcxhXx/A3dcbqbbYgBEkc6e/J/s7j2tS96zoB0S9VBdLwQNCWwUo4LA==",
       "license": "MIT",
+      "peer": true,
       "engines": {
         "node": ">=16.9.0"
       }
@@ -10187,9 +10249,9 @@
       }
     },
     "node_modules/ip-address": {
-      "version": "10.1.0",
-      "resolved": "https://registry.npmjs.org/ip-address/-/ip-address-10.1.0.tgz",
-      "integrity": "sha512-XXADHxXmvT9+CRxhXg56LJovE+bmWnEWB78LB83VZTprKTmaC5QfruXocxzTZ2Kl0DNwKuBdlIhjL8LeY8Sf8Q==",
+      "version": "10.2.0",
+      "resolved": "https://registry.npmjs.org/ip-address/-/ip-address-10.2.0.tgz",
+      "integrity": "sha512-/+S6j4E9AHvW9SWMSEY9Xfy66O5PWvVEJ08O0y5JGyEKQpojb0K0GKpz/v5HJ/G0vi3D2sjGK78119oXZeE0qA==",
       "license": "MIT",
       "engines": {
         "node": ">= 12"
@@ -10632,7 +10694,8 @@
       "version": "4.0.0",
       "resolved": "https://registry.npmjs.org/is-promise/-/is-promise-4.0.0.tgz",
       "integrity": "sha512-hvpoI6korhJMnej285dSg6nu1+e6uxs7zG3BYAm5byqDsgJNWwxzM6z6iZiAgQR4TJ30JmBTOwqZUw3WlyH3AQ==",
-      "license": "MIT"
+      "license": "MIT",
+      "peer": true
     },
     "node_modules/is-regex": {
       "version": "1.2.1",
@@ -10903,10 +10966,11 @@
       }
     },
     "node_modules/jose": {
-      "version": "6.2.2",
-      "resolved": "https://registry.npmjs.org/jose/-/jose-6.2.2.tgz",
-      "integrity": "sha512-d7kPDd34KO/YnzaDOlikGpOurfF0ByC2sEV4cANCtdqLlTfBlw2p14O/5d/zv40gJPbIQxfES3nSx1/oYNyuZQ==",
+      "version": "6.2.3",
+      "resolved": "https://registry.npmjs.org/jose/-/jose-6.2.3.tgz",
+      "integrity": "sha512-YYVDInQKFJfR/xa3ojUTl8c2KoTwiL1R5Wg9YCydwH0x0B9grbzlg5HC7mMjCtUJjbQ/YnGEZIhI5tCgfTb4Hw==",
       "license": "MIT",
+      "peer": true,
       "funding": {
         "url": "https://github.com/sponsors/panva"
       }
@@ -10968,6 +11032,7 @@
       "resolved": "https://registry.npmjs.org/json-schema-to-ts/-/json-schema-to-ts-3.1.1.tgz",
       "integrity": "sha512-+DWg8jCJG2TEnpy7kOm/7/AxaYoaRbjVB4LFZLySZlWn8exGs3A4OLJR966cVvU26N7X9TWxl+Jsw7dzAqKT6g==",
       "license": "MIT",
+      "peer": true,
       "dependencies": {
         "@babel/runtime": "^7.18.3",
         "ts-algebra": "^2.0.0"
@@ -10987,7 +11052,8 @@
       "version": "8.0.2",
       "resolved": "https://registry.npmjs.org/json-schema-typed/-/json-schema-typed-8.0.2.tgz",
       "integrity": "sha512-fQhoXdcvc3V28x7C7BMs4P5+kNlgUURe2jmUT1T//oBRMDrqy1QPelJimwZGo7Hg9VPV3EQV5Bnq4hbFy2vetA==",
-      "license": "BSD-2-Clause"
+      "license": "BSD-2-Clause",
+      "peer": true
     },
     "node_modules/json-stable-stringify-without-jsonify": {
       "version": "1.0.1",
@@ -13979,6 +14045,7 @@
       "resolved": "https://registry.npmjs.org/pkce-challenge/-/pkce-challenge-5.0.1.tgz",
       "integrity": "sha512-wQ0b/W4Fr01qtpHlqSqspcj3EhBvimsdh0KlHhH8HRZnMsEa0ea2fTULOXOS9ccQr3om+GcGRk4e+isrZWV8qQ==",
       "license": "MIT",
+      "peer": true,
       "engines": {
         "node": ">=16.20.0"
       }
@@ -15279,6 +15346,7 @@
       "resolved": "https://registry.npmjs.org/router/-/router-2.2.0.tgz",
       "integrity": "sha512-nLTrUKm2UyiL7rlhapu/Zl45FwNgkZGaCpZbIHajDYgwlJCOzLSk+cIPAnsEqV955GjILJnKbdQC1nVPz+gAYQ==",
       "license": "MIT",
+      "peer": true,
       "dependencies": {
         "debug": "^4.4.0",
         "depd": "^2.0.0",
@@ -15295,6 +15363,7 @@
       "resolved": "https://registry.npmjs.org/path-to-regexp/-/path-to-regexp-8.4.2.tgz",
       "integrity": "sha512-qRcuIdP69NPm4qbACK+aDogI5CBDMi1jKe0ry5rSQJz8JVLsC7jV8XpiJjGRLLol3N+R5ihGYcrPLTno6pAdBA==",
       "license": "MIT",
+      "peer": true,
       "funding": {
         "type": "opencollective",
         "url": "https://opencollective.com/express"
@@ -16353,6 +16422,17 @@
         "node": ">=12.0.0"
       }
     },
+    "node_modules/standardwebhooks": {
+      "version": "1.0.0",
+      "resolved": "https://registry.npmjs.org/standardwebhooks/-/standardwebhooks-1.0.0.tgz",
+      "integrity": "sha512-BbHGOQK9olHPMvQNHWul6MYlrRTAOKn03rOe4A8O3CLWhNf4YHBqq2HJKKC+sfqpxiBY52pNeesD6jIiLDz8jg==",
+      "license": "MIT",
+      "peer": true,
+      "dependencies": {
+        "@stablelib/base64": "^1.0.0",
+        "fast-sha256": "^1.3.0"
+      }
+    },
     "node_modules/statuses": {
       "version": "2.0.1",
       "resolved": "https://registry.npmjs.org/statuses/-/statuses-2.0.1.tgz",
@@ -17063,7 +17143,8 @@
       "version": "2.0.0",
       "resolved": "https://registry.npmjs.org/ts-algebra/-/ts-algebra-2.0.0.tgz",
       "integrity": "sha512-FPAhNPFMrkwz76P7cdjdmiShwMynZYN6SgOujD1urY4oNm80Ou9oMdmbR45LotcKOXoy7wSmHkRFE6Mxbrhefw==",
-      "license": "MIT"
+      "license": "MIT",
+      "peer": true
     },
     "node_modules/ts-api-utils": {
       "version": "2.4.0",
@@ -18949,6 +19030,7 @@
       "resolved": "https://registry.npmjs.org/zod-to-json-schema/-/zod-to-json-schema-3.25.2.tgz",
       "integrity": "sha512-O/PgfnpT1xKSDeQYSCfRI5Gy3hPf91mKVDuYLUHZJMiDFptvP41MSnWofm8dnCm0256ZNfZIM7DSzuSMAFnjHA==",
       "license": "ISC",
+      "peer": true,
       "peerDependencies": {
         "zod": "^3.25.28 || ^4"
       }

package.json

@@ -1,6 +1,6 @@
 {
   "name": "@cloudcli-ai/cloudcli",
-  "version": "1.33.0",
+  "version": "1.34.0",
   "description": "A web-based UI for Claude Code CLI",
   "type": "module",
   "main": "dist-server/server/index.js",
@@ -11,7 +11,6 @@
     "server/",
     "shared/",
     "public/api-docs.html",
-    "public/modelConstants.js",
     "dist/",
     "dist-server/",
     "scripts/",
@@ -67,7 +66,7 @@
   "author": "CloudCLI UI Contributors",
   "license": "AGPL-3.0-or-later",
   "dependencies": {
-    "@anthropic-ai/claude-agent-sdk": "^0.2.116",
+    "@anthropic-ai/claude-agent-sdk": "^0.3.165",
     "@codemirror/lang-css": "^6.3.1",
     "@codemirror/lang-html": "^6.4.9",
     "@codemirror/lang-javascript": "^6.2.4",

public/api-docs.html

@@ -820,31 +820,49 @@ data: {"type":"done"}</code></pre>
         </div>
     </div>
 
-    <script type="module">
-        // Import model constants
-        import { PROVIDERS } from './modelConstants.js';
-
+    <script>
         // Dynamic URL replacement
         const apiUrl = window.location.origin;
         document.querySelectorAll('.api-url').forEach(el => {
             el.textContent = apiUrl;
         });
 
-        // Dynamically populate model documentation
-        window.addEventListener('DOMContentLoaded', () => {
-            const modelCell = document.getElementById('model-options-cell');
-            if (modelCell) {
-                const providerModels = PROVIDERS.map(provider => {
-                    const models = provider.models.OPTIONS.map(m => `<code>${m.value}</code>`).join(', ');
-                    return `<strong>${provider.name}:</strong> ${models} (default: <code>${provider.models.DEFAULT}</code>)`;
-                }).join('<br><br>');
+        // Populate model documentation from the live provider API
+        const PROVIDER_ORDER = [
+            { id: 'claude',   name: 'Anthropic' },
+            { id: 'codex',    name: 'OpenAI' },
+            { id: 'gemini',   name: 'Google' },
+            { id: 'cursor',   name: 'Cursor' },
+            { id: 'opencode', name: 'OpenCode' },
+        ];
 
-                modelCell.innerHTML = `
-                    Model identifier for the AI provider:<br><br>
-                    ${providerModels}
-                `;
-            }
-        });
+        async function populateModels() {
+            const modelCell = document.getElementById('model-options-cell');
+            if (!modelCell) return;
+
+            const token = localStorage.getItem('auth-token');
+            const headers = token ? { Authorization: `Bearer ${token}` } : {};
+
+            const results = await Promise.allSettled(
+                PROVIDER_ORDER.map(({ id }) =>
+                    fetch(`/api/providers/${id}/models`, { headers }).then(r => r.json())
+                )
+            );
+
+            const providerModels = results.map((result, i) => {
+                const { name } = PROVIDER_ORDER[i];
+                if (result.status === 'rejected' || !result.value?.data?.models) {
+                    return `<strong>${name}:</strong> <em>unavailable</em>`;
+                }
+                const { OPTIONS, DEFAULT } = result.value.data.models;
+                const models = OPTIONS.map(m => `<code>${m.value}</code>`).join(', ');
+                return `<strong>${name}:</strong> ${models} (default: <code>${DEFAULT}</code>)`;
+            }).join('<br><br>');
+
+            modelCell.innerHTML = `Model identifier for the AI provider:<br><br>${providerModels}`;
+        }
+
+        document.addEventListener('DOMContentLoaded', populateModels);
 
         // Tab switching
         window.showTab = function(tabName) {

public/modelConstants.js

@@ -1,841 +0,0 @@
-/**
- * Documentation Model Definitions
- * Used by README links and the public API docs.
- */
-
-/**
- * Claude (Anthropic) Models
- */
-export const CLAUDE_MODELS = {
-  OPTIONS: [
-    {
-      value: "default",
-      label: "Default (recommended)",
-      description: "Use the default model (currently Opus 4.7 (1M context)) · $5/$25 per Mtok",
-    },
-    {
-      value: "sonnet",
-      label: "Sonnet",
-      description: "Sonnet 4.6 · Best for everyday tasks · $3/$15 per Mtok",
-    },
-    {
-      value: "sonnet[1m]",
-      label: "Sonnet (1M context)",
-      description: "Sonnet 4.6 for long sessions · $3/$15 per Mtok",
-    },
-    {
-      value: "haiku",
-      label: "Haiku",
-      description: "Haiku 4.5 · Fastest for quick answers · $1/$5 per Mtok",
-    },
-  ],
-
-  DEFAULT: "default",
-};
-
-/**
- * Cursor Models
- */
-export const CURSOR_MODELS = {
-  OPTIONS: [
-    { value: "auto", label: "auto", description: "Auto" },
-    {
-      value: "composer-2-fast",
-      label: "composer-2-fast",
-      description: "Composer 2 Fast",
-    },
-    {
-      value: "composer-2",
-      label: "composer-2",
-      description: "Composer 2",
-    },
-    {
-      value: "gpt-5.3-codex-low",
-      label: "gpt-5.3-codex-low",
-      description: "Codex 5.3 Low",
-    },
-    {
-      value: "gpt-5.3-codex-low-fast",
-      label: "gpt-5.3-codex-low-fast",
-      description: "Codex 5.3 Low Fast",
-    },
-    {
-      value: "gpt-5.3-codex",
-      label: "gpt-5.3-codex",
-      description: "Codex 5.3",
-    },
-    {
-      value: "gpt-5.3-codex-fast",
-      label: "gpt-5.3-codex-fast",
-      description: "Codex 5.3 Fast",
-    },
-    {
-      value: "gpt-5.3-codex-high",
-      label: "gpt-5.3-codex-high",
-      description: "Codex 5.3 High",
-    },
-    {
-      value: "gpt-5.3-codex-high-fast",
-      label: "gpt-5.3-codex-high-fast",
-      description: "Codex 5.3 High Fast",
-    },
-    {
-      value: "gpt-5.3-codex-xhigh",
-      label: "gpt-5.3-codex-xhigh",
-      description: "Codex 5.3 Extra High",
-    },
-    {
-      value: "gpt-5.3-codex-xhigh-fast",
-      label: "gpt-5.3-codex-xhigh-fast",
-      description: "Codex 5.3 Extra High Fast",
-    },
-    { value: "gpt-5.2", label: "gpt-5.2", description: "GPT-5.2" },
-    {
-      value: "gpt-5.2-codex-low",
-      label: "gpt-5.2-codex-low",
-      description: "Codex 5.2 Low",
-    },
-    {
-      value: "gpt-5.2-codex-low-fast",
-      label: "gpt-5.2-codex-low-fast",
-      description: "Codex 5.2 Low Fast",
-    },
-    {
-      value: "gpt-5.2-codex",
-      label: "gpt-5.2-codex",
-      description: "Codex 5.2",
-    },
-    {
-      value: "gpt-5.2-codex-fast",
-      label: "gpt-5.2-codex-fast",
-      description: "Codex 5.2 Fast",
-    },
-    {
-      value: "gpt-5.2-codex-high",
-      label: "gpt-5.2-codex-high",
-      description: "Codex 5.2 High",
-    },
-    {
-      value: "gpt-5.2-codex-high-fast",
-      label: "gpt-5.2-codex-high-fast",
-      description: "Codex 5.2 High Fast",
-    },
-    {
-      value: "gpt-5.2-codex-xhigh",
-      label: "gpt-5.2-codex-xhigh",
-      description: "Codex 5.2 Extra High",
-    },
-    {
-      value: "gpt-5.2-codex-xhigh-fast",
-      label: "gpt-5.2-codex-xhigh-fast",
-      description: "Codex 5.2 Extra High Fast",
-    },
-    {
-      value: "gpt-5.1-codex-max-low",
-      label: "gpt-5.1-codex-max-low",
-      description: "Codex 5.1 Max Low",
-    },
-    {
-      value: "gpt-5.1-codex-max-low-fast",
-      label: "gpt-5.1-codex-max-low-fast",
-      description: "Codex 5.1 Max Low Fast",
-    },
-    {
-      value: "gpt-5.1-codex-max-medium",
-      label: "gpt-5.1-codex-max-medium",
-      description: "Codex 5.1 Max",
-    },
-    {
-      value: "gpt-5.1-codex-max-medium-fast",
-      label: "gpt-5.1-codex-max-medium-fast",
-      description: "Codex 5.1 Max Medium Fast",
-    },
-    {
-      value: "gpt-5.1-codex-max-high",
-      label: "gpt-5.1-codex-max-high",
-      description: "Codex 5.1 Max High",
-    },
-    {
-      value: "gpt-5.1-codex-max-high-fast",
-      label: "gpt-5.1-codex-max-high-fast",
-      description: "Codex 5.1 Max High Fast",
-    },
-    {
-      value: "gpt-5.1-codex-max-xhigh",
-      label: "gpt-5.1-codex-max-xhigh",
-      description: "Codex 5.1 Max Extra High",
-    },
-    {
-      value: "gpt-5.1-codex-max-xhigh-fast",
-      label: "gpt-5.1-codex-max-xhigh-fast",
-      description: "Codex 5.1 Max Extra High Fast",
-    },
-    {
-      value: "composer-2.5",
-      label: "composer-2.5",
-      description: "Composer 2.5",
-    },
-    {
-      value: "gpt-5.5-high",
-      label: "gpt-5.5-high",
-      description: "GPT-5.5 1M High",
-    },
-    {
-      value: "gpt-5.5-high-fast",
-      label: "gpt-5.5-high-fast",
-      description: "GPT-5.5 High Fast",
-    },
-    {
-      value: "claude-opus-4-7-thinking-high",
-      label: "claude-opus-4-7-thinking-high",
-      description: "Opus 4.7 1M High Thinking",
-    },
-    {
-      value: "gpt-5.4-high",
-      label: "gpt-5.4-high",
-      description: "GPT-5.4 1M High",
-    },
-    {
-      value: "gpt-5.4-high-fast",
-      label: "gpt-5.4-high-fast",
-      description: "GPT-5.4 High Fast",
-    },
-    {
-      value: "claude-4.6-opus-high-thinking",
-      label: "claude-4.6-opus-high-thinking",
-      description: "Opus 4.6 1M Thinking",
-    },
-    {
-      value: "claude-4.6-opus-high-thinking-fast",
-      label: "claude-4.6-opus-high-thinking-fast",
-      description: "Opus 4.6 1M Thinking Fast",
-    },
-    {
-      value: "composer-2.5-fast",
-      label: "composer-2.5-fast",
-      description: "Composer 2.5 Fast",
-    },
-    {
-      value: "gpt-5.5-none",
-      label: "gpt-5.5-none",
-      description: "GPT-5.5 1M None",
-    },
-    {
-      value: "gpt-5.5-none-fast",
-      label: "gpt-5.5-none-fast",
-      description: "GPT-5.5 None Fast",
-    },
-    {
-      value: "gpt-5.5-low",
-      label: "gpt-5.5-low",
-      description: "GPT-5.5 1M Low",
-    },
-    {
-      value: "gpt-5.5-low-fast",
-      label: "gpt-5.5-low-fast",
-      description: "GPT-5.5 Low Fast",
-    },
-    {
-      value: "gpt-5.5-medium",
-      label: "gpt-5.5-medium",
-      description: "GPT-5.5 1M",
-    },
-    {
-      value: "gpt-5.5-medium-fast",
-      label: "gpt-5.5-medium-fast",
-      description: "GPT-5.5 Fast",
-    },
-    {
-      value: "gpt-5.5-extra-high",
-      label: "gpt-5.5-extra-high",
-      description: "GPT-5.5 1M Extra High",
-    },
-    {
-      value: "gpt-5.5-extra-high-fast",
-      label: "gpt-5.5-extra-high-fast",
-      description: "GPT-5.5 Extra High Fast",
-    },
-    {
-      value: "claude-4.6-sonnet-medium",
-      label: "claude-4.6-sonnet-medium",
-      description: "Sonnet 4.6 1M",
-    },
-    {
-      value: "claude-4.6-sonnet-medium-thinking",
-      label: "claude-4.6-sonnet-medium-thinking",
-      description: "Sonnet 4.6 1M Thinking",
-    },
-    {
-      value: "claude-opus-4-7-low",
-      label: "claude-opus-4-7-low",
-      description: "Opus 4.7 1M Low",
-    },
-    {
-      value: "claude-opus-4-7-low-fast",
-      label: "claude-opus-4-7-low-fast",
-      description: "Opus 4.7 1M Low Fast",
-    },
-    {
-      value: "claude-opus-4-7-medium",
-      label: "claude-opus-4-7-medium",
-      description: "Opus 4.7 1M Medium",
-    },
-    {
-      value: "claude-opus-4-7-medium-fast",
-      label: "claude-opus-4-7-medium-fast",
-      description: "Opus 4.7 1M Medium Fast",
-    },
-    {
-      value: "claude-opus-4-7-high",
-      label: "claude-opus-4-7-high",
-      description: "Opus 4.7 1M High",
-    },
-    {
-      value: "claude-opus-4-7-high-fast",
-      label: "claude-opus-4-7-high-fast",
-      description: "Opus 4.7 1M High Fast",
-    },
-    {
-      value: "claude-opus-4-7-xhigh",
-      label: "claude-opus-4-7-xhigh",
-      description: "Opus 4.7 1M",
-    },
-    {
-      value: "claude-opus-4-7-xhigh-fast",
-      label: "claude-opus-4-7-xhigh-fast",
-      description: "Opus 4.7 1M Fast",
-    },
-    {
-      value: "claude-opus-4-7-max",
-      label: "claude-opus-4-7-max",
-      description: "Opus 4.7 1M Max",
-    },
-    {
-      value: "claude-opus-4-7-max-fast",
-      label: "claude-opus-4-7-max-fast",
-      description: "Opus 4.7 1M Max Fast",
-    },
-    {
-      value: "claude-opus-4-7-thinking-low",
-      label: "claude-opus-4-7-thinking-low",
-      description: "Opus 4.7 1M Low Thinking",
-    },
-    {
-      value: "claude-opus-4-7-thinking-low-fast",
-      label: "claude-opus-4-7-thinking-low-fast",
-      description: "Opus 4.7 1M Low Thinking Fast",
-    },
-    {
-      value: "claude-opus-4-7-thinking-medium",
-      label: "claude-opus-4-7-thinking-medium",
-      description: "Opus 4.7 1M Medium Thinking",
-    },
-    {
-      value: "claude-opus-4-7-thinking-medium-fast",
-      label: "claude-opus-4-7-thinking-medium-fast",
-      description: "Opus 4.7 1M Medium Thinking Fast",
-    },
-    {
-      value: "claude-opus-4-7-thinking-high-fast",
-      label: "claude-opus-4-7-thinking-high-fast",
-      description: "Opus 4.7 1M High Thinking Fast",
-    },
-    {
-      value: "claude-opus-4-7-thinking-xhigh",
-      label: "claude-opus-4-7-thinking-xhigh",
-      description: "Opus 4.7 1M Thinking",
-    },
-    {
-      value: "claude-opus-4-7-thinking-xhigh-fast",
-      label: "claude-opus-4-7-thinking-xhigh-fast",
-      description: "Opus 4.7 1M Thinking Fast",
-    },
-    {
-      value: "claude-opus-4-7-thinking-max",
-      label: "claude-opus-4-7-thinking-max",
-      description: "Opus 4.7 1M Max Thinking",
-    },
-    {
-      value: "claude-opus-4-7-thinking-max-fast",
-      label: "claude-opus-4-7-thinking-max-fast",
-      description: "Opus 4.7 1M Max Thinking Fast",
-    },
-    {
-      value: "grok-build-0.1",
-      label: "grok-build-0.1",
-      description: "Grok Build 0.1 1M",
-    },
-    {
-      value: "gpt-5.4-low",
-      label: "gpt-5.4-low",
-      description: "GPT-5.4 1M Low",
-    },
-    {
-      value: "gpt-5.4-medium",
-      label: "gpt-5.4-medium",
-      description: "GPT-5.4 1M",
-    },
-    {
-      value: "gpt-5.4-medium-fast",
-      label: "gpt-5.4-medium-fast",
-      description: "GPT-5.4 Fast",
-    },
-    {
-      value: "gpt-5.4-xhigh",
-      label: "gpt-5.4-xhigh",
-      description: "GPT-5.4 1M Extra High",
-    },
-    {
-      value: "gpt-5.4-xhigh-fast",
-      label: "gpt-5.4-xhigh-fast",
-      description: "GPT-5.4 Extra High Fast",
-    },
-    {
-      value: "claude-4.6-opus-high",
-      label: "claude-4.6-opus-high",
-      description: "Opus 4.6 1M",
-    },
-    {
-      value: "claude-4.6-opus-max",
-      label: "claude-4.6-opus-max",
-      description: "Opus 4.6 1M Max",
-    },
-    {
-      value: "claude-4.6-opus-max-thinking",
-      label: "claude-4.6-opus-max-thinking",
-      description: "Opus 4.6 1M Max Thinking",
-    },
-    {
-      value: "claude-4.6-opus-max-thinking-fast",
-      label: "claude-4.6-opus-max-thinking-fast",
-      description: "Opus 4.6 1M Max Thinking Fast",
-    },
-    {
-      value: "claude-4.5-opus-high",
-      label: "claude-4.5-opus-high",
-      description: "Opus 4.5",
-    },
-    {
-      value: "claude-4.5-opus-high-thinking",
-      label: "claude-4.5-opus-high-thinking",
-      description: "Opus 4.5 Thinking",
-    },
-    {
-      value: "gpt-5.2-low",
-      label: "gpt-5.2-low",
-      description: "GPT-5.2 Low",
-    },
-    {
-      value: "gpt-5.2-low-fast",
-      label: "gpt-5.2-low-fast",
-      description: "GPT-5.2 Low Fast",
-    },
-    {
-      value: "gpt-5.2-fast",
-      label: "gpt-5.2-fast",
-      description: "GPT-5.2 Fast",
-    },
-    {
-      value: "gpt-5.2-high",
-      label: "gpt-5.2-high",
-      description: "GPT-5.2 High",
-    },
-    {
-      value: "gpt-5.2-high-fast",
-      label: "gpt-5.2-high-fast",
-      description: "GPT-5.2 High Fast",
-    },
-    {
-      value: "gpt-5.2-xhigh",
-      label: "gpt-5.2-xhigh",
-      description: "GPT-5.2 Extra High",
-    },
-    {
-      value: "gpt-5.2-xhigh-fast",
-      label: "gpt-5.2-xhigh-fast",
-      description: "GPT-5.2 Extra High Fast",
-    },
-    {
-      value: "gemini-3.1-pro",
-      label: "gemini-3.1-pro",
-      description: "Gemini 3.1 Pro",
-    },
-    {
-      value: "gpt-5.4-mini-none",
-      label: "gpt-5.4-mini-none",
-      description: "GPT-5.4 Mini None",
-    },
-    {
-      value: "gpt-5.4-mini-low",
-      label: "gpt-5.4-mini-low",
-      description: "GPT-5.4 Mini Low",
-    },
-    {
-      value: "gpt-5.4-mini-medium",
-      label: "gpt-5.4-mini-medium",
-      description: "GPT-5.4 Mini",
-    },
-    {
-      value: "gpt-5.4-mini-high",
-      label: "gpt-5.4-mini-high",
-      description: "GPT-5.4 Mini High",
-    },
-    {
-      value: "gpt-5.4-mini-xhigh",
-      label: "gpt-5.4-mini-xhigh",
-      description: "GPT-5.4 Mini Extra High",
-    },
-    {
-      value: "gpt-5.4-nano-none",
-      label: "gpt-5.4-nano-none",
-      description: "GPT-5.4 Nano None",
-    },
-    {
-      value: "gpt-5.4-nano-low",
-      label: "gpt-5.4-nano-low",
-      description: "GPT-5.4 Nano Low",
-    },
-    {
-      value: "gpt-5.4-nano-medium",
-      label: "gpt-5.4-nano-medium",
-      description: "GPT-5.4 Nano",
-    },
-    {
-      value: "gpt-5.4-nano-high",
-      label: "gpt-5.4-nano-high",
-      description: "GPT-5.4 Nano High",
-    },
-    {
-      value: "gpt-5.4-nano-xhigh",
-      label: "gpt-5.4-nano-xhigh",
-      description: "GPT-5.4 Nano Extra High",
-    },
-    {
-      value: "grok-4.3",
-      label: "grok-4.3",
-      description: "Grok 4.3 1M",
-    },
-    {
-      value: "claude-4.5-sonnet",
-      label: "claude-4.5-sonnet",
-      description: "Sonnet 4.5",
-    },
-    {
-      value: "claude-4.5-sonnet-thinking",
-      label: "claude-4.5-sonnet-thinking",
-      description: "Sonnet 4.5 Thinking",
-    },
-    {
-      value: "gpt-5.1-low",
-      label: "gpt-5.1-low",
-      description: "GPT-5.1 Low",
-    },
-    {
-      value: "gpt-5.1",
-      label: "gpt-5.1",
-      description: "GPT-5.1",
-    },
-    {
-      value: "gpt-5.1-high",
-      label: "gpt-5.1-high",
-      description: "GPT-5.1 High",
-    },
-    {
-      value: "gemini-3-flash",
-      label: "gemini-3-flash",
-      description: "Gemini 3 Flash",
-    },
-    {
-      value: "gemini-3.5-flash",
-      label: "gemini-3.5-flash",
-      description: "Gemini 3.5 Flash",
-    },
-    {
-      value: "gpt-5.1-codex-mini-low",
-      label: "gpt-5.1-codex-mini-low",
-      description: "Codex 5.1 Mini Low",
-    },
-    {
-      value: "gpt-5.1-codex-mini",
-      label: "gpt-5.1-codex-mini",
-      description: "Codex 5.1 Mini",
-    },
-    {
-      value: "gpt-5.1-codex-mini-high",
-      label: "gpt-5.1-codex-mini-high",
-      description: "Codex 5.1 Mini High",
-    },
-    {
-      value: "claude-4-sonnet",
-      label: "claude-4-sonnet",
-      description: "Sonnet 4",
-    },
-    {
-      value: "claude-4-sonnet-thinking",
-      label: "claude-4-sonnet-thinking",
-      description: "Sonnet 4 Thinking",
-    },
-    {
-      value: "gpt-5-mini",
-      label: "gpt-5-mini",
-      description: "GPT-5 Mini",
-    },
-    {
-      value: "kimi-k2.5",
-      label: "kimi-k2.5",
-      description: "Kimi K2.5",
-    },
-  ],
-
-  DEFAULT: "composer-2.5-fast",
-};
-
-/**
- * Codex (OpenAI) Models
- */
-export const CODEX_MODELS = {
-  OPTIONS: [
-    { value: "gpt-5.5", label: "gpt-5.5" },
-    { value: "gpt-5.4", label: "gpt-5.4" },
-    { value: "gpt-5.4-mini", label: "gpt-5.4-mini" },
-    { value: "gpt-5.3-codex", label: "gpt-5.3-codex" },
-    { value: "gpt-5.2", label: "gpt-5.2" },
-  ],
-
-  DEFAULT: "gpt-5.4",
-};
-
-/**
- * Gemini Models
- */
-export const GEMINI_MODELS = {
-  OPTIONS: [
-    { value: "gemini-3.1-pro-preview", label: "Gemini 3.1 Pro Preview" },
-    { value: "gemini-3-pro-preview", label: "Gemini 3 Pro Preview" },
-    { value: "gemini-3-flash-preview", label: "Gemini 3 Flash Preview" },
-    { value: "gemini-2.5-flash", label: "Gemini 2.5 Flash" },
-    { value: "gemini-2.5-pro", label: "Gemini 2.5 Pro" },
-    { value: "gemini-2.0-flash-lite", label: "Gemini 2.0 Flash Lite" },
-    { value: "gemini-2.0-flash", label: "Gemini 2.0 Flash" },
-    { value: "gemini-2.0-pro-exp", label: "Gemini 2.0 Pro Experimental" },
-    {
-      value: "gemini-2.0-flash-thinking-exp",
-      label: "Gemini 2.0 Flash Thinking",
-    },
-  ],
-
-  DEFAULT: "gemini-3.1-pro-preview",
-};
-
-/**
- * OpenCode Models
- *
- * OpenCode model ids include the upstream provider prefix.
- */
-export const OPENCODE_MODELS = {
-  OPTIONS: [
-    {
-      value: "opencode/big-pickle",
-      label: "Big Pickle",
-      description: "opencode - opencode/big-pickle",
-    },
-    {
-      value: "opencode/deepseek-v4-flash-free",
-      label: "Deepseek V4 Flash Free",
-      description: "opencode - opencode/deepseek-v4-flash-free",
-    },
-    {
-      value: "opencode/nemotron-3-super-free",
-      label: "Nemotron 3 Super Free",
-      description: "opencode - opencode/nemotron-3-super-free",
-    },
-    {
-      value: "anthropic/claude-3-5-haiku-20241022",
-      label: "Claude 3.5 Haiku (2024-10-22)",
-      description: "anthropic - anthropic/claude-3-5-haiku-20241022",
-    },
-    {
-      value: "anthropic/claude-3-5-haiku-latest",
-      label: "Claude 3.5 Haiku Latest",
-      description: "anthropic - anthropic/claude-3-5-haiku-latest",
-    },
-    {
-      value: "anthropic/claude-3-5-sonnet-20240620",
-      label: "Claude 3.5 Sonnet (2024-06-20)",
-      description: "anthropic - anthropic/claude-3-5-sonnet-20240620",
-    },
-    {
-      value: "anthropic/claude-3-5-sonnet-20241022",
-      label: "Claude 3.5 Sonnet (2024-10-22)",
-      description: "anthropic - anthropic/claude-3-5-sonnet-20241022",
-    },
-    {
-      value: "anthropic/claude-3-7-sonnet-20250219",
-      label: "Claude 3.7 Sonnet (2025-02-19)",
-      description: "anthropic - anthropic/claude-3-7-sonnet-20250219",
-    },
-    {
-      value: "anthropic/claude-3-haiku-20240307",
-      label: "Claude 3 Haiku (2024-03-07)",
-      description: "anthropic - anthropic/claude-3-haiku-20240307",
-    },
-    {
-      value: "anthropic/claude-3-opus-20240229",
-      label: "Claude 3 Opus (2024-02-29)",
-      description: "anthropic - anthropic/claude-3-opus-20240229",
-    },
-    {
-      value: "anthropic/claude-3-sonnet-20240229",
-      label: "Claude 3 Sonnet (2024-02-29)",
-      description: "anthropic - anthropic/claude-3-sonnet-20240229",
-    },
-    {
-      value: "anthropic/claude-haiku-4-5",
-      label: "Claude Haiku 4.5",
-      description: "anthropic - anthropic/claude-haiku-4-5",
-    },
-    {
-      value: "anthropic/claude-haiku-4-5-20251001",
-      label: "Claude Haiku 4.5 (2025-10-01)",
-      description: "anthropic - anthropic/claude-haiku-4-5-20251001",
-    },
-    {
-      value: "anthropic/claude-opus-4-0",
-      label: "Claude Opus 4.0",
-      description: "anthropic - anthropic/claude-opus-4-0",
-    },
-    {
-      value: "anthropic/claude-opus-4-1",
-      label: "Claude Opus 4.1",
-      description: "anthropic - anthropic/claude-opus-4-1",
-    },
-    {
-      value: "anthropic/claude-opus-4-1-20250805",
-      label: "Claude Opus 4.1 (2025-08-05)",
-      description: "anthropic - anthropic/claude-opus-4-1-20250805",
-    },
-    {
-      value: "anthropic/claude-opus-4-20250514",
-      label: "Claude Opus 4 (2025-05-14)",
-      description: "anthropic - anthropic/claude-opus-4-20250514",
-    },
-    {
-      value: "anthropic/claude-opus-4-5",
-      label: "Claude Opus 4.5",
-      description: "anthropic - anthropic/claude-opus-4-5",
-    },
-    {
-      value: "anthropic/claude-opus-4-5-20251101",
-      label: "Claude Opus 4.5 (2025-11-01)",
-      description: "anthropic - anthropic/claude-opus-4-5-20251101",
-    },
-    {
-      value: "anthropic/claude-opus-4-6",
-      label: "Claude Opus 4.6",
-      description: "anthropic - anthropic/claude-opus-4-6",
-    },
-    {
-      value: "anthropic/claude-opus-4-6-fast",
-      label: "Claude Opus 4.6 Fast",
-      description: "anthropic - anthropic/claude-opus-4-6-fast",
-    },
-    {
-      value: "anthropic/claude-opus-4-7",
-      label: "Claude Opus 4.7",
-      description: "anthropic - anthropic/claude-opus-4-7",
-    },
-    {
-      value: "anthropic/claude-opus-4-7-fast",
-      label: "Claude Opus 4.7 Fast",
-      description: "anthropic - anthropic/claude-opus-4-7-fast",
-    },
-    {
-      value: "anthropic/claude-sonnet-4-0",
-      label: "Claude Sonnet 4.0",
-      description: "anthropic - anthropic/claude-sonnet-4-0",
-    },
-    {
-      value: "anthropic/claude-sonnet-4-20250514",
-      label: "Claude Sonnet 4 (2025-05-14)",
-      description: "anthropic - anthropic/claude-sonnet-4-20250514",
-    },
-    {
-      value: "anthropic/claude-sonnet-4-5",
-      label: "Claude Sonnet 4.5",
-      description: "anthropic - anthropic/claude-sonnet-4-5",
-    },
-    {
-      value: "anthropic/claude-sonnet-4-5-20250929",
-      label: "Claude Sonnet 4.5 (2025-09-29)",
-      description: "anthropic - anthropic/claude-sonnet-4-5-20250929",
-    },
-    {
-      value: "anthropic/claude-sonnet-4-6",
-      label: "Claude Sonnet 4.6",
-      description: "anthropic - anthropic/claude-sonnet-4-6",
-    },
-    {
-      value: "openai/gpt-5.2",
-      label: "GPT-5.2",
-      description: "openai - openai/gpt-5.2",
-    },
-    {
-      value: "openai/gpt-5.3-codex",
-      label: "GPT-5.3 Codex",
-      description: "openai - openai/gpt-5.3-codex",
-    },
-    {
-      value: "openai/gpt-5.3-codex-spark",
-      label: "GPT-5.3 Codex Spark",
-      description: "openai - openai/gpt-5.3-codex-spark",
-    },
-    {
-      value: "openai/gpt-5.4",
-      label: "GPT-5.4",
-      description: "openai - openai/gpt-5.4",
-    },
-    {
-      value: "openai/gpt-5.4-fast",
-      label: "GPT-5.4 Fast",
-      description: "openai - openai/gpt-5.4-fast",
-    },
-    {
-      value: "openai/gpt-5.4-mini",
-      label: "GPT-5.4 Mini",
-      description: "openai - openai/gpt-5.4-mini",
-    },
-    {
-      value: "openai/gpt-5.4-mini-fast",
-      label: "GPT-5.4 Mini Fast",
-      description: "openai - openai/gpt-5.4-mini-fast",
-    },
-    {
-      value: "openai/gpt-5.5",
-      label: "GPT-5.5",
-      description: "openai - openai/gpt-5.5",
-    },
-    {
-      value: "openai/gpt-5.5-fast",
-      label: "GPT-5.5 Fast",
-      description: "openai - openai/gpt-5.5-fast",
-    },
-    {
-      value: "openai/gpt-5.5-pro",
-      label: "GPT-5.5 Pro",
-      description: "openai - openai/gpt-5.5-pro",
-    },
-  ],
-
-  DEFAULT: "anthropic/claude-sonnet-4-5",
-};
-
-/**
- * Ordered provider registry. Display order in documentation.
- */
-export const PROVIDERS = [
-  { id: "claude", name: "Anthropic", models: CLAUDE_MODELS },
-  { id: "codex", name: "OpenAI", models: CODEX_MODELS },
-  { id: "gemini", name: "Google", models: GEMINI_MODELS },
-  { id: "cursor", name: "Cursor", models: CURSOR_MODELS },
-  { id: "opencode", name: "OpenCode", models: OPENCODE_MODELS },
-];

redirect-package/README.md

@@ -75,7 +75,7 @@
 - **Session Management** - Resume conversations, manage multiple sessions, and track history
 - **Plugin System** - Extend CloudCLI with custom plugins — add new tabs, backend services, and integrations. [Build your own →](https://github.com/cloudcli-ai/cloudcli-plugin-starter)
 - **TaskMaster AI Integration** *(Optional)* - Advanced project management with AI-powered task planning, PRD parsing, and workflow automation
-- **Model Compatibility** - Works with Claude, GPT, and Gemini model families (see [`public/modelConstants.js`](https://github.com/siteboon/claudecodeui/blob/main/public/modelConstants.js) for the full list of supported models)
+- **Model Compatibility** - Works with Claude, GPT, and Gemini model families (the full list of supported models is available at runtime via `GET /api/providers/:provider/models`)
 
 
 ## Quick Start

server/claude-sdk.js

@@ -28,10 +28,14 @@ import {
 } from './services/notification-orchestrator.js';
 import { sessionsService } from './modules/providers/services/sessions.service.js';
 import { providerAuthService } from './modules/providers/services/provider-auth.service.js';
-import { createNormalizedMessage } from './shared/utils.js';
+import { createCompleteMessage, createNormalizedMessage } from './shared/utils.js';
 
 const activeSessions = new Map();
 const pendingToolApprovals = new Map();
+// Sessions cancelled via abort-session. The abort handler already sent the
+// terminal `complete` (aborted: true) to the client, so the run loop must not
+// emit a second one when its generator winds down.
+const abortedSessionIds = new Set();
 
 const TOOL_APPROVAL_TIMEOUT_MS = parseInt(process.env.CLAUDE_TOOL_APPROVAL_TIMEOUT_MS, 10) || 55000;
 
@@ -204,7 +208,7 @@ function mapCliOptionsToSDK(options = {}) {
   sdkOptions.disallowedTools = settings.disallowedTools || [];
 
   // Map model (default to sonnet)
-  // Valid models: sonnet, opus, haiku, opusplan, sonnet[1m]
+  // Valid models: sonnet, opus, haiku, opusplan, sonnet[1m], fable
   sdkOptions.model = options.model || CLAUDE_FALLBACK_MODELS.DEFAULT;
   // Model logged at query start below
 
@@ -304,7 +308,11 @@ function extractTokenBudget(sdkMessage) {
 
   const messageUsage = sdkMessage.message?.usage || sdkMessage.usage;
   if (messageUsage && typeof messageUsage === 'object') {
-    const inputTokens = readNumber(messageUsage.input_tokens ?? messageUsage.inputTokens);
+    const directInputTokens = readNumber(messageUsage.input_tokens ?? messageUsage.inputTokens);
+    const cacheCreationTokens = readNumber(messageUsage.cache_creation_input_tokens ?? messageUsage.cacheCreationInputTokens ?? messageUsage.cacheCreationTokens);
+    const cacheReadTokens = readNumber(messageUsage.cache_read_input_tokens ?? messageUsage.cacheReadInputTokens ?? messageUsage.cacheReadTokens);
+    const cacheTokens = cacheCreationTokens + cacheReadTokens;
+    const inputTokens = directInputTokens + cacheTokens;
     const outputTokens = readNumber(messageUsage.output_tokens ?? messageUsage.outputTokens);
     const totalUsed = inputTokens + outputTokens;
     const contextWindow = parseInt(process.env.CONTEXT_WINDOW, 10) || 160000;
@@ -314,6 +322,9 @@ function extractTokenBudget(sdkMessage) {
       total: contextWindow,
       inputTokens,
       outputTokens,
+      cacheReadTokens,
+      cacheCreationTokens,
+      cacheTokens,
       breakdown: {
         input: inputTokens,
         output: outputTokens,
@@ -724,14 +735,18 @@ async function queryClaudeSDK(command, options = {}, ws) {
     // Clean up temporary image files
     await cleanupTempFiles(tempImagePaths, tempDir);
 
-    // Send completion event
-    ws.send(createNormalizedMessage({ kind: 'complete', exitCode: 0, isNewSession: !sessionId && !!command, sessionId: capturedSessionId, provider: 'claude' }));
+    // Send the terminal completion event — skipped for aborted runs, whose
+    // terminal `complete` (aborted: true) was already sent by abort-session.
+    const wasAborted = capturedSessionId ? abortedSessionIds.delete(capturedSessionId) : false;
+    if (!wasAborted) {
+      ws.send(createCompleteMessage({ provider: 'claude', sessionId: capturedSessionId || sessionId || null, exitCode: 0 }));
+    }
     notifyRunStopped({
       userId: ws?.userId || null,
       provider: 'claude',
       sessionId: capturedSessionId || sessionId || null,
       sessionName: sessionSummary,
-      stopReason: 'completed'
+      stopReason: wasAborted ? 'aborted' : 'completed'
     });
     // Complete
 
@@ -746,14 +761,22 @@ async function queryClaudeSDK(command, options = {}, ws) {
     // Clean up temporary image files on error
     await cleanupTempFiles(tempImagePaths, tempDir);
 
+    const wasAborted = capturedSessionId ? abortedSessionIds.delete(capturedSessionId) : false;
+    if (wasAborted) {
+      // The abort already produced the terminal complete; a generator throw
+      // caused by interrupt() is expected noise, not a user-facing error.
+      return;
+    }
+
     // Check if Claude CLI is installed for a clearer error message
     const installed = await providerAuthService.isProviderInstalled('claude');
     const errorContent = !installed
       ? 'Claude Code is not installed. Please install it first: https://docs.anthropic.com/en/docs/claude-code'
       : error.message;
 
-    // Send error to WebSocket
+    // Send error to WebSocket, then the terminal complete
     ws.send(createNormalizedMessage({ kind: 'error', content: errorContent, sessionId: capturedSessionId || sessionId || null, provider: 'claude' }));
+    ws.send(createCompleteMessage({ provider: 'claude', sessionId: capturedSessionId || sessionId || null, exitCode: 1 }));
     notifyRunFailed({
       userId: ws?.userId || null,
       provider: 'claude',
@@ -780,6 +803,10 @@ async function abortClaudeSDKSession(sessionId) {
   try {
     console.log(`Aborting SDK session: ${sessionId}`);
 
+    // Mark before interrupting so the run loop knows not to emit its own
+    // terminal complete (the abort handler sends the aborted one).
+    abortedSessionIds.add(sessionId);
+
     // Call interrupt() on the query instance
     await session.instance.interrupt();
 
@@ -795,6 +822,8 @@ async function abortClaudeSDKSession(sessionId) {
     return true;
   } catch (error) {
     console.error(`Error aborting session ${sessionId}:`, error);
+    // The run keeps going; let it emit its own terminal complete.
+    abortedSessionIds.delete(sessionId);
     return false;
   }
 }

server/cli.js

@@ -455,7 +455,7 @@ async function sandboxCommand(args) {
             await new Promise(resolve => setTimeout(resolve, 5000));
 
             console.log(`${c.info('▶')} Launching CloudCLI web server...`);
-            sbx(['exec', opts.name, 'bash', '-c', 'cloudcli start --port 3001 &']);
+            sbx(['exec', opts.name, 'bash', '-c', 'nohup cloudcli start --port 3001 > /tmp/cloudcli-ui.log 2>&1 & disown']);
 
             console.log(`${c.info('▶')} Forwarding port ${opts.port} → 3001...`);
             try {
@@ -554,7 +554,7 @@ async function sandboxCommand(args) {
 
             // Step 3: Start CloudCLI inside the sandbox
             console.log(`${c.info('▶')} Launching CloudCLI web server...`);
-            sbx(['exec', opts.name, 'bash', '-c', 'cloudcli start --port 3001 &']);
+            sbx(['exec', opts.name, 'bash', '-c', 'nohup cloudcli start --port 3001 > /tmp/cloudcli-ui.log 2>&1 & disown']);
 
             // Step 4: Forward port
             console.log(`${c.info('▶')} Forwarding port ${opts.port} → 3001...`);

server/cursor-cli.js

@@ -4,7 +4,7 @@ import { notifyRunFailed, notifyRunStopped } from './services/notification-orche
 import { sessionsService } from './modules/providers/services/sessions.service.js';
 import { providerAuthService } from './modules/providers/services/provider-auth.service.js';
 import { providerModelsService } from './modules/providers/services/provider-models.service.js';
-import { createNormalizedMessage } from './shared/utils.js';
+import { createCompleteMessage, createNormalizedMessage } from './shared/utils.js';
 
 // Use cross-spawn on Windows for better command execution
 const spawnFunction = process.platform === 'win32' ? crossSpawn : spawn;
@@ -34,6 +34,10 @@ async function spawnCursor(command, options = {}, ws) {
     let sessionCreatedSent = false; // Track if we've already sent session-created event
     let hasRetriedWithTrust = false;
     let settled = false;
+    // The unified lifecycle contract requires exactly one terminal `complete`
+    // per run. Cursor surfaces completion twice (the `result` JSON line and
+    // the process close), so the first emission wins.
+    let completeSent = false;
 
     // Use tools settings passed from frontend, or defaults
     const settings = toolsSettings || {
@@ -197,15 +201,15 @@ async function spawnCursor(command, options = {}, ws) {
               break;
 
             case 'result': {
-              // Session complete — send stream end + lifecycle complete with result payload
-              const resultText = typeof response.result === 'string' ? response.result : '';
-              ws.send(createNormalizedMessage({
-                kind: 'complete',
-                exitCode: response.subtype === 'success' ? 0 : 1,
-                resultText,
-                isError: response.subtype !== 'success',
-                sessionId: capturedSessionId || sessionId, provider: 'cursor',
-              }));
+              // Session complete — terminal lifecycle event for this run
+              if (!completeSent) {
+                completeSent = true;
+                ws.send(createCompleteMessage({
+                  provider: 'cursor',
+                  sessionId: capturedSessionId || sessionId || null,
+                  exitCode: response.subtype === 'success' ? 0 : 1,
+                }));
+              }
               break;
             }
 
@@ -271,7 +275,12 @@ async function spawnCursor(command, options = {}, ws) {
           return;
         }
 
-        ws.send(createNormalizedMessage({ kind: 'complete', exitCode: code, isNewSession: !sessionId && !!command, sessionId: finalSessionId, provider: 'cursor' }));
+        // Terminal complete — unless the `result` line already sent it, or the
+        // run was aborted (abort-session sent the aborted complete).
+        if (!completeSent && !cursorProcess.aborted) {
+          completeSent = true;
+          ws.send(createCompleteMessage({ provider: 'cursor', sessionId: finalSessionId, exitCode: code }));
+        }
 
         if (code === 0) {
           notifyTerminalState({ code });
@@ -297,6 +306,10 @@ async function spawnCursor(command, options = {}, ws) {
           : error.message;
 
         ws.send(createNormalizedMessage({ kind: 'error', content: errorContent, sessionId: capturedSessionId || sessionId || null, provider: 'cursor' }));
+        if (!completeSent && !cursorProcess.aborted) {
+          completeSent = true;
+          ws.send(createCompleteMessage({ provider: 'cursor', sessionId: capturedSessionId || sessionId || null, exitCode: 1 }));
+        }
         notifyTerminalState({ error });
 
         settleOnce(() => reject(error));
@@ -314,6 +327,9 @@ function abortCursorSession(sessionId) {
   const process = activeCursorProcesses.get(sessionId);
   if (process) {
     console.log(`Aborting Cursor session: ${sessionId}`);
+    // The abort handler sends the terminal complete (aborted: true); flag the
+    // process so its close handler does not emit a second one.
+    process.aborted = true;
     process.kill('SIGTERM');
     activeCursorProcesses.delete(sessionId);
     return true;

server/gemini-cli.js

@@ -10,7 +10,7 @@ import GeminiResponseHandler from './gemini-response-handler.js';
 import { notifyRunFailed, notifyRunStopped } from './services/notification-orchestrator.js';
 import { providerAuthService } from './modules/providers/services/provider-auth.service.js';
 import { providerModelsService } from './modules/providers/services/provider-models.service.js';
-import { createNormalizedMessage } from './shared/utils.js';
+import { createCompleteMessage, createNormalizedMessage } from './shared/utils.js';
 
 // Use cross-spawn on Windows for correct .cmd resolution (same pattern as cursor-cli.js)
 const spawnFunction = process.platform === 'win32' ? crossSpawn : spawn;
@@ -129,6 +129,9 @@ async function spawnGemini(command, options = {}, ws) {
     let capturedSessionId = sessionId; // Track session ID throughout the process
     let sessionCreatedSent = false; // Track if we've already sent session-created event
     let assistantBlocks = []; // Accumulate the full response blocks including tools
+    // Unified lifecycle contract: exactly one terminal `complete` per run
+    // (close and error handlers can both fire for spawn failures).
+    let completeSent = false;
 
     // Use tools settings passed from frontend, or defaults
     const settings = toolsSettings || {
@@ -486,7 +489,12 @@ async function spawnGemini(command, options = {}, ws) {
                 sessionManager.addMessage(finalSessionId, 'assistant', assistantBlocks);
             }
 
-            ws.send(createNormalizedMessage({ kind: 'complete', exitCode: code, isNewSession: !sessionId && !!command, sessionId: finalSessionId, provider: 'gemini' }));
+            // Terminal complete — skipped for aborted runs (abort-session
+            // already sent the aborted complete on this run's behalf).
+            if (!completeSent && !geminiProcess.aborted) {
+                completeSent = true;
+                ws.send(createCompleteMessage({ provider: 'gemini', sessionId: finalSessionId, exitCode: code }));
+            }
 
             // Clean up temporary image files if any
             if (geminiProcess.tempImagePaths && geminiProcess.tempImagePaths.length > 0) {
@@ -566,6 +574,10 @@ async function spawnGemini(command, options = {}, ws) {
 
             const errorSessionId = typeof ws.getSessionId === 'function' ? ws.getSessionId() : finalSessionId;
             ws.send(createNormalizedMessage({ kind: 'error', content: errorContent, sessionId: errorSessionId, provider: 'gemini' }));
+            if (!completeSent && !geminiProcess.aborted) {
+                completeSent = true;
+                ws.send(createCompleteMessage({ provider: 'gemini', sessionId: errorSessionId, exitCode: 1 }));
+            }
             notifyTerminalState({ error });
 
             reject(error);
@@ -590,6 +602,9 @@ function abortGeminiSession(sessionId) {
 
     if (geminiProc) {
         try {
+            // The abort handler sends the terminal complete (aborted: true);
+            // flag the process so its close handler does not emit a second one.
+            geminiProc.aborted = true;
             geminiProc.kill('SIGTERM');
             setTimeout(() => {
                 if (activeGeminiProcesses.has(processKey)) {

server/index.js

@@ -22,35 +22,24 @@ import { findAppRoot, getModuleDir } from './utils/runtime-paths.js';
 import {
     queryClaudeSDK,
     abortClaudeSDKSession,
-    isClaudeSDKSessionActive,
-    getActiveClaudeSDKSessions,
     resolveToolApproval,
     getPendingApprovalsForSession,
-    reconnectSessionWriter,
 } from './claude-sdk.js';
 import {
     spawnCursor,
     abortCursorSession,
-    isCursorSessionActive,
-    getActiveCursorSessions,
 } from './cursor-cli.js';
 import {
     queryCodex,
     abortCodexSession,
-    isCodexSessionActive,
-    getActiveCodexSessions,
 } from './openai-codex.js';
 import {
     spawnGemini,
     abortGeminiSession,
-    isGeminiSessionActive,
-    getActiveGeminiSessions,
 } from './gemini-cli.js';
 import {
     spawnOpenCode,
     abortOpenCodeSession,
-    isOpenCodeSessionActive,
-    getActiveOpenCodeSessions,
 } from './opencode-cli.js';
 import sessionManager from './sessionManager.js';
 import {
@@ -84,9 +73,17 @@ const __dirname = getModuleDir(import.meta.url);
 // Resolving the app root once keeps every repo-level lookup below aligned across both layouts.
 const APP_ROOT = findAppRoot(__dirname);
 const installMode = fs.existsSync(path.join(APP_ROOT, '.git')) ? 'git' : 'npm';
+const MAX_FILE_UPLOAD_SIZE_MB = 200;
+const MAX_FILE_UPLOAD_SIZE_BYTES = MAX_FILE_UPLOAD_SIZE_MB * 1024 * 1024;
+const MAX_FILE_UPLOAD_COUNT = 20;
 
 console.log('SERVER_PORT from env:', process.env.SERVER_PORT);
 
+function readUsageNumber(value) {
+    const parsed = Number(value);
+    return Number.isFinite(parsed) ? parsed : 0;
+}
+
 const app = express();
 const server = http.createServer(app);
 
@@ -97,32 +94,35 @@ const wss = createWebSocketServer(server, {
         authenticateWebSocket,
     },
     chat: {
-        queryClaudeSDK,
-        spawnCursor,
-        queryCodex,
-        spawnGemini,
-        spawnOpenCode,
-        abortClaudeSDKSession,
-        abortCursorSession,
-        abortCodexSession,
-        abortGeminiSession,
-        abortOpenCodeSession,
+        spawnFns: {
+            claude: queryClaudeSDK,
+            cursor: spawnCursor,
+            codex: queryCodex,
+            gemini: spawnGemini,
+            opencode: spawnOpenCode,
+        },
+        abortFns: {
+            claude: abortClaudeSDKSession,
+            cursor: abortCursorSession,
+            codex: abortCodexSession,
+            gemini: abortGeminiSession,
+            opencode: abortOpenCodeSession,
+        },
         resolveToolApproval,
-        isClaudeSDKSessionActive,
-        isCursorSessionActive,
-        isCodexSessionActive,
-        isGeminiSessionActive,
-        isOpenCodeSessionActive,
-        reconnectSessionWriter,
         getPendingApprovalsForSession,
-        getActiveClaudeSDKSessions,
-        getActiveCursorSessions,
-        getActiveCodexSessions,
-        getActiveGeminiSessions,
-        getActiveOpenCodeSessions,
     },
     shell: {
-        getSessionById: (sessionId) => sessionManager.getSession(sessionId),
+        resolveProviderSessionId: (sessionId, provider) => {
+            const dbSession = sessionsDb.getSessionById(sessionId);
+            const legacyGeminiSession =
+                provider === 'gemini' ? sessionManager.getSession(sessionId) : null;
+
+            if (dbSession) {
+                return dbSession.provider_session_id ?? legacyGeminiSession?.cliSessionId ?? null;
+            }
+
+            return legacyGeminiSession?.cliSessionId;
+        },
         stripAnsiSequences,
         normalizeDetectedUrl,
         extractUrlsFromText,
@@ -892,27 +892,27 @@ const uploadFilesHandler = async (req, res) => {
             }
         }),
         limits: {
-            fileSize: 50 * 1024 * 1024, // 50MB limit
-            files: 20 // Max 20 files at once
+            fileSize: MAX_FILE_UPLOAD_SIZE_BYTES,
+            files: MAX_FILE_UPLOAD_COUNT
         }
     });
 
     // Use multer middleware
-    uploadMiddleware.array('files', 20)(req, res, async (err) => {
+    uploadMiddleware.array('files', MAX_FILE_UPLOAD_COUNT)(req, res, async (err) => {
         if (err) {
             console.error('Multer error:', err);
             if (err.code === 'LIMIT_FILE_SIZE') {
-                return res.status(400).json({ error: 'File too large. Maximum size is 50MB.' });
+                return res.status(400).json({ error: `File too large. Maximum size is ${MAX_FILE_UPLOAD_SIZE_MB}MB.` });
             }
             if (err.code === 'LIMIT_FILE_COUNT') {
-                return res.status(400).json({ error: 'Too many files. Maximum is 20 files.' });
+                return res.status(400).json({ error: `Too many files. Maximum is ${MAX_FILE_UPLOAD_COUNT} files.` });
             }
             return res.status(500).json({ error: err.message });
         }
 
         try {
             const { projectId } = req.params;
-            const { targetPath, relativePaths } = req.body;
+            const { targetPath, relativePaths, requestedFileCount: requestedFileCountRaw } = req.body;
 
             // Parse relative paths if provided (for folder uploads)
             let filePaths = [];
@@ -936,6 +936,11 @@ const uploadFilesHandler = async (req, res) => {
                 return res.status(400).json({ error: 'No files provided' });
             }
 
+            const parsedRequestedFileCount = Number.parseInt(requestedFileCountRaw, 10);
+            const requestedFileCount = Number.isFinite(parsedRequestedFileCount) && parsedRequestedFileCount > 0
+                ? parsedRequestedFileCount
+                : req.files.length;
+
             // Resolve the project directory through the DB using the new projectId.
             const projectRoot = await projectsDb.getProjectPathById(projectId);
             if (!projectRoot) {
@@ -1014,8 +1019,10 @@ const uploadFilesHandler = async (req, res) => {
             res.json({
                 success: true,
                 files: uploadedFiles,
+                uploadedCount: uploadedFiles.length,
+                requestedFileCount,
                 targetPath: resolvedTargetDir,
-                message: `Uploaded ${uploadedFiles.length} file(s) successfully`
+                message: `Uploaded ${uploadedFiles.length} ${uploadedFiles.length === 1 ? 'file' : 'files'} successfully`
             });
         } catch (error) {
             console.error('Error uploading files:', error);
@@ -1137,6 +1144,12 @@ app.get('/api/projects/:projectId/sessions/:sessionId/token-usage', authenticate
             return res.status(400).json({ error: 'Invalid sessionId' });
         }
 
+        // Provider artifacts on disk (JSONL file names, OpenCode sqlite rows)
+        // are keyed by the provider-native session id, while the caller sends
+        // the app-facing id. Resolve the mapping once for all branches below.
+        const sessionRow = sessionsDb.getSessionById(safeSessionId);
+        const providerNativeSessionId = sessionRow?.provider_session_id || safeSessionId;
+
         // Handle Cursor sessions - they use SQLite and don't have token usage info
         if (provider === 'cursor') {
             return res.json({
@@ -1237,7 +1250,7 @@ app.get('/api/projects/:projectId/sessions/:sessionId/token-usage', authenticate
                         tokens_cache_write AS cacheWriteTokens
                     FROM session
                     WHERE id = ?
-                `).get(safeSessionId);
+                `).get(providerNativeSessionId);
 
                 if (!row) {
                     return res.status(404).json({ error: 'OpenCode session not found', sessionId: safeSessionId });
@@ -1278,7 +1291,7 @@ app.get('/api/projects/:projectId/sessions/:sessionId/token-usage', authenticate
                         if (entry.isDirectory()) {
                             const found = await findSessionFile(fullPath);
                             if (found) return found;
-                        } else if (entry.name.includes(safeSessionId) && entry.name.endsWith('.jsonl')) {
+                        } else if (entry.name.includes(providerNativeSessionId) && entry.name.endsWith('.jsonl')) {
                             return fullPath;
                         }
                     }
@@ -1362,12 +1375,19 @@ app.get('/api/projects/:projectId/sessions/:sessionId/token-usage', authenticate
         const encodedPath = projectPath.replace(/[^a-zA-Z0-9-]/g, '-');
         const projectDir = path.join(homeDir, '.claude', 'projects', encodedPath);
 
-        const jsonlPath = path.join(projectDir, `${safeSessionId}.jsonl`);
+        // Prefer the indexed transcript path (already produced by the trusted
+        // session synchronizer); fall back to the conventional location
+        // derived from the provider-native session id.
+        let jsonlPath = sessionRow?.jsonl_path;
+        if (!jsonlPath) {
+            jsonlPath = path.join(projectDir, `${providerNativeSessionId}.jsonl`);
 
-        // Constrain to projectDir
-        const rel = path.relative(path.resolve(projectDir), path.resolve(jsonlPath));
-        if (rel.startsWith('..') || path.isAbsolute(rel)) {
-            return res.status(400).json({ error: 'Invalid path' });
+            // Constrain the constructed path to projectDir (the id is
+            // caller-influenced in this fallback branch).
+            const rel = path.relative(path.resolve(projectDir), path.resolve(jsonlPath));
+            if (rel.startsWith('..') || path.isAbsolute(rel)) {
+                return res.status(400).json({ error: 'Invalid path' });
+            }
         }
 
         // Read and parse the JSONL file
@@ -1386,6 +1406,8 @@ app.get('/api/projects/:projectId/sessions/:sessionId/token-usage', authenticate
         const contextWindow = Number.isFinite(parsedContextWindow) ? parsedContextWindow : 160000;
         let inputTokens = 0;
         let outputTokens = 0;
+        let cacheReadTokens = 0;
+        let cacheCreationTokens = 0;
 
         // Find the latest assistant message with usage data (scan from end)
         for (let i = lines.length - 1; i >= 0; i--) {
@@ -1397,8 +1419,11 @@ app.get('/api/projects/:projectId/sessions/:sessionId/token-usage', authenticate
                     const usage = entry.message.usage;
 
                     // Use token counts from latest assistant message only
-                    inputTokens = usage.input_tokens || 0;
-                    outputTokens = usage.output_tokens || 0;
+                    const directInputTokens = readUsageNumber(usage.input_tokens ?? usage.inputTokens);
+                    cacheReadTokens = readUsageNumber(usage.cache_read_input_tokens ?? usage.cacheReadInputTokens ?? usage.cacheReadTokens);
+                    cacheCreationTokens = readUsageNumber(usage.cache_creation_input_tokens ?? usage.cacheCreationInputTokens ?? usage.cacheCreationTokens);
+                    inputTokens = directInputTokens + cacheReadTokens + cacheCreationTokens;
+                    outputTokens = readUsageNumber(usage.output_tokens ?? usage.outputTokens);
 
                     break; // Stop after finding the latest assistant message
                 }
@@ -1409,12 +1434,16 @@ app.get('/api/projects/:projectId/sessions/:sessionId/token-usage', authenticate
         }
 
         const totalUsed = inputTokens + outputTokens;
+        const cacheTokens = cacheReadTokens + cacheCreationTokens;
 
         res.json({
             used: totalUsed,
             total: contextWindow,
             inputTokens,
             outputTokens,
+            cacheReadTokens,
+            cacheCreationTokens,
+            cacheTokens,
             breakdown: {
                 input: inputTokens,
                 output: outputTokens
@@ -1483,74 +1512,133 @@ function permToRwx(perm) {
     return r + w + x;
 }
 
+// Directories that are almost never interesting for a project tree but can
+// contain tens of thousands of files. Skipping them before recursion keeps
+// traversal time bounded on large monorepos and high-latency filesystems
+// (NFS / SMB).
+const IGNORED_DIRS = new Set([
+    // JS / TS toolchains
+    'node_modules', 'dist', 'build', '.next', '.nuxt', '.cache', '.parcel-cache',
+    // VCS
+    '.git', '.svn', '.hg',
+    // Python
+    '__pycache__', '.pytest_cache', '.mypy_cache', '.tox', 'venv', '.venv',
+    // Rust / Go / Java / Ruby
+    'target', 'vendor',
+    // Build output / IDE
+    '.gradle', '.idea', 'coverage', '.nyc_output'
+]);
+
+const DEFAULT_FS_CONCURRENCY = 64;
+const parsedFsConcurrency = Number.parseInt(process.env.FS_CONCURRENCY || '', 10);
+const FS_CONCURRENCY = Number.isFinite(parsedFsConcurrency) && parsedFsConcurrency > 0
+    ? parsedFsConcurrency
+    : DEFAULT_FS_CONCURRENCY;
+let activeFsOperations = 0;
+const pendingFsOperations = [];
+
+async function acquire() {
+    if (activeFsOperations < FS_CONCURRENCY) {
+        activeFsOperations += 1;
+        return;
+    }
+
+    await new Promise((resolve) => {
+        pendingFsOperations.push(resolve);
+    });
+}
+
+function release() {
+    const next = pendingFsOperations.shift();
+    if (next) {
+        next();
+        return;
+    }
+
+    activeFsOperations = Math.max(0, activeFsOperations - 1);
+}
+
 async function getFileTree(dirPath, maxDepth = 3, currentDepth = 0, showHidden = true) {
     // Using fsPromises from import
-    const items = [];
-
+    let entries;
     try {
-        const entries = await fsPromises.readdir(dirPath, { withFileTypes: true });
-
-        for (const entry of entries) {
-            // Debug: log all entries including hidden files
-
-
-            // Skip heavy build directories and VCS directories
-            if (entry.name === 'node_modules' ||
-                entry.name === 'dist' ||
-                entry.name === 'build' ||
-                entry.name === '.git' ||
-                entry.name === '.svn' ||
-                entry.name === '.hg') continue;
-
-            const itemPath = path.join(dirPath, entry.name);
-            const item = {
-                name: entry.name,
-                path: itemPath,
-                type: entry.isDirectory() ? 'directory' : 'file'
-            };
-
-            // Get file stats for additional metadata
-            try {
-                const stats = await fsPromises.stat(itemPath);
-                item.size = stats.size;
-                item.modified = stats.mtime.toISOString();
-
-                // Convert permissions to rwx format
-                const mode = stats.mode;
-                const ownerPerm = (mode >> 6) & 7;
-                const groupPerm = (mode >> 3) & 7;
-                const otherPerm = mode & 7;
-                item.permissions = ((mode >> 6) & 7).toString() + ((mode >> 3) & 7).toString() + (mode & 7).toString();
-                item.permissionsRwx = permToRwx(ownerPerm) + permToRwx(groupPerm) + permToRwx(otherPerm);
-            } catch (statError) {
-                // If stat fails, provide default values
-                item.size = 0;
-                item.modified = null;
-                item.permissions = '000';
-                item.permissionsRwx = '---------';
-            }
-
-            if (entry.isDirectory() && currentDepth < maxDepth) {
-                // Recursively get subdirectories but limit depth
-                try {
-                    // Check if we can access the directory before trying to read it
-                    await fsPromises.access(item.path, fs.constants.R_OK);
-                    item.children = await getFileTree(item.path, maxDepth, currentDepth + 1, showHidden);
-                } catch (e) {
-                    // Silently skip directories we can't access (permission denied, etc.)
-                    item.children = [];
-                }
-            }
-
-            items.push(item);
+        await acquire();
+        try {
+            entries = await fsPromises.readdir(dirPath, { withFileTypes: true });
+        } finally {
+            release();
         }
     } catch (error) {
         // Only log non-permission errors to avoid spam
         if (error.code !== 'EACCES' && error.code !== 'EPERM') {
             console.error('Error reading directory:', error);
         }
+        return [];
     }
 
+    const filteredEntries = entries.filter((entry) => !(entry.isDirectory() && IGNORED_DIRS.has(entry.name)));
+
+    // Process every entry in parallel. On high-latency filesystems (NFS/SMB)
+    // serial stat() was the real bottleneck — issuing them concurrently lets
+    // the kernel pipeline the round-trips and the recursive calls overlap too.
+    const items = await Promise.all(filteredEntries.map(async (entry) => {
+        const itemPath = path.join(dirPath, entry.name);
+        const item = {
+            name: entry.name,
+            path: itemPath,
+            type: entry.isDirectory() ? 'directory' : 'file'
+        };
+
+        // Get file stats for additional metadata
+        try {
+            await acquire();
+            try {
+              const stats = await fsPromises.lstat(itemPath);
+              item.size = stats.size;
+              item.modified = stats.mtime.toISOString();
+
+              // Mark symlinks so UI can distinguish them
+              if (stats.isSymbolicLink()) {
+                item.isSymlink = true;
+              }
+
+              // Convert permissions to rwx format
+              const mode = stats.mode;
+              const ownerPerm = (mode >> 6) & 7;
+              const groupPerm = (mode >> 3) & 7;
+              const otherPerm = mode & 7;
+              item.permissions =
+                ((mode >> 6) & 7).toString() +
+                ((mode >> 3) & 7).toString() +
+                (mode & 7).toString();
+              item.permissionsRwx =
+                permToRwx(ownerPerm) +
+                permToRwx(groupPerm) +
+                permToRwx(otherPerm);
+            } finally {
+                release();
+            }
+        } catch (statError) {
+            // If stat fails, provide default values
+            item.size = 0;
+            item.modified = null;
+            item.permissions = '000';
+            item.permissionsRwx = '---------';
+        }
+
+        if (entry.isDirectory() && currentDepth < maxDepth) {
+            // Recurse. Let readdir's own EACCES bubble up through the catch in
+            // the recursive call rather than doing a separate access() probe
+            // (which doubled the round-trip count on SMB without adding info).
+            // The recursive call starts with a bounded readdir; holding a permit
+            // for the whole subtree can deadlock when sibling directories are
+            // waiting on their own children.
+            item.children = await getFileTree(itemPath, maxDepth, currentDepth + 1, showHidden);
+        }
+
+        return item;
+    }));
+
     return items.sort((a, b) => {
         if (a.type !== b.type) {
             return a.type === 'directory' ? -1 : 1;

server/modules/database/migrations.ts

@@ -382,6 +382,25 @@ const rebuildSessionsTableWithProjectSchema = (db: Database): void => {
   }
 };
 
+/**
+ * Adds the `provider_session_id` mapping column used by the session gateway.
+ *
+ * Rows that existed before this migration were always keyed directly by the
+ * provider-native session id, so backfilling `provider_session_id` with
+ * `session_id` keeps every legacy row resolvable through the new mapping.
+ */
+const addProviderSessionIdMapping = (db: Database): void => {
+  const sessionsTableInfo = getTableInfo(db, 'sessions');
+  const columnNames = sessionsTableInfo.map((column) => column.name);
+
+  addColumnToTableIfNotExists(db, 'sessions', columnNames, 'provider_session_id', 'TEXT');
+  db.exec(`
+    UPDATE sessions
+    SET provider_session_id = session_id
+    WHERE provider_session_id IS NULL
+  `);
+};
+
 const ensureProjectsForSessionPaths = (db: Database): void => {
   if (!tableExists(db, 'sessions')) {
     return;
@@ -428,9 +447,11 @@ export const runMigrations = (db: Database) => {
     migrateLegacyWorkspaceTableIntoProjects(db);
     rebuildSessionsTableWithProjectSchema(db);
     migrateLegacySessionNames(db);
+    addProviderSessionIdMapping(db);
     ensureProjectsForSessionPaths(db);
 
     db.exec('CREATE INDEX IF NOT EXISTS idx_session_ids_lookup ON sessions(session_id)');
+    db.exec('CREATE INDEX IF NOT EXISTS idx_sessions_provider_session_id ON sessions(provider_session_id)');
     db.exec('CREATE INDEX IF NOT EXISTS idx_sessions_project_path ON sessions(project_path)');
     db.exec('CREATE INDEX IF NOT EXISTS idx_sessions_is_archived ON sessions(isArchived)');
     db.exec('CREATE INDEX IF NOT EXISTS idx_projects_is_starred ON projects(isStarred)');

server/modules/database/repositories/notification-preferences.ts

@@ -10,6 +10,7 @@ type NotificationPreferences = {
   channels: {
     inApp: boolean;
     webPush: boolean;
+    sound: boolean;
   };
   events: {
     actionRequired: boolean;
@@ -22,6 +23,7 @@ const DEFAULT_NOTIFICATION_PREFERENCES: NotificationPreferences = {
   channels: {
     inApp: false,
     webPush: false,
+    sound: true,
   },
   events: {
     actionRequired: true,
@@ -37,6 +39,7 @@ function normalizeNotificationPreferences(value: unknown): NotificationPreferenc
     channels: {
       inApp: source.channels?.inApp === true,
       webPush: source.channels?.webPush === true,
+      sound: source.channels?.sound !== false,
     },
     events: {
       actionRequired: source.events?.actionRequired !== false,

server/modules/database/repositories/sessions.db.ts

@@ -5,6 +5,7 @@ import { normalizeProjectPath } from '@/shared/utils.js';
 type SessionRow = {
   session_id: string;
   provider: string;
+  provider_session_id: string | null;
   project_path: string | null;
   jsonl_path: string | null;
   custom_name: string | null;
@@ -13,10 +14,8 @@ type SessionRow = {
   updated_at: string;
 };
 
-type SessionMetadataLookupRow = Pick<
-  SessionRow,
-  'session_id' | 'provider' | 'project_path' | 'jsonl_path' | 'custom_name' | 'isArchived' | 'created_at' | 'updated_at'
->;
+const SESSION_ROW_COLUMNS =
+  'session_id, provider, provider_session_id, project_path, jsonl_path, custom_name, isArchived, created_at, updated_at';
 
 function normalizeTimestamp(value?: string): string | null {
   if (!value) return null;
@@ -35,8 +34,16 @@ function normalizeProjectPathForProvider(provider: string, projectPath: string):
 }
 
 export const sessionsDb = {
+  /**
+   * Upserts one session row discovered on disk by a provider synchronizer.
+   *
+   * The given id is the provider-native session id. Rows are keyed by
+   * `provider_session_id` so a session that was first created by the app
+   * (with an app-allocated `session_id`) is updated in place once its
+   * transcript shows up on disk, instead of producing a duplicate row.
+   */
   createSession(
-    sessionId: string,
+    providerSessionId: string,
     provider: string,
     projectPath: string,
     customName?: string,
@@ -53,19 +60,54 @@ export const sessionsDb = {
     // since it's a foreign key in the sessions table.
     projectsDb.createProjectPath(normalizedProjectPath);
 
+    const existing = db
+      .prepare(
+        `SELECT session_id FROM sessions
+         WHERE provider_session_id = ? AND provider = ?
+         LIMIT 1`
+      )
+      .get(providerSessionId, provider) as { session_id: string } | undefined;
+
+    if (existing) {
+      db.prepare(
+        `UPDATE sessions SET
+           provider = ?,
+           updated_at = COALESCE(?, CURRENT_TIMESTAMP),
+           project_path = ?,
+           jsonl_path = ?,
+           isArchived = 0,
+           custom_name = COALESCE(?, custom_name)
+         WHERE session_id = ?`
+      ).run(
+        provider,
+        updatedAtValue,
+        normalizedProjectPath,
+        jsonlPath ?? null,
+        customName ?? null,
+        existing.session_id
+      );
+
+      return existing.session_id;
+    }
+
+    // Sessions created outside the app (directly via the provider CLI) are
+    // keyed by the provider-native id for both columns. The ON CONFLICT path
+    // covers legacy rows that predate the provider_session_id mapping.
     db.prepare(
-      `INSERT INTO sessions (session_id, provider, custom_name, project_path, jsonl_path, isArchived, created_at, updated_at)
-       VALUES (?, ?, ?, ?, ?, 0, COALESCE(?, CURRENT_TIMESTAMP), COALESCE(?, CURRENT_TIMESTAMP))
+      `INSERT INTO sessions (session_id, provider, provider_session_id, custom_name, project_path, jsonl_path, isArchived, created_at, updated_at)
+       VALUES (?, ?, ?, ?, ?, ?, 0, COALESCE(?, CURRENT_TIMESTAMP), COALESCE(?, CURRENT_TIMESTAMP))
        ON CONFLICT(session_id) DO UPDATE SET
          provider = excluded.provider,
+         provider_session_id = excluded.provider_session_id,
          updated_at = excluded.updated_at,
          project_path = excluded.project_path,
          jsonl_path = excluded.jsonl_path,
          isArchived = 0,
          custom_name = COALESCE(excluded.custom_name, sessions.custom_name)`
     ).run(
-      sessionId,
+      providerSessionId,
       provider,
+      providerSessionId,
       customName ?? null,
       normalizedProjectPath,
       jsonlPath ?? null,
@@ -73,9 +115,77 @@ export const sessionsDb = {
       updatedAtValue
     );
 
+    return providerSessionId;
+  },
+
+  /**
+   * Inserts one app-allocated session row before any provider run happens.
+   *
+   * The session gateway uses this when the frontend starts a brand-new chat:
+   * `session_id` is the stable app-facing id, while `provider_session_id`
+   * stays NULL until the provider runtime announces its own id and
+   * `assignProviderSessionId` records the mapping.
+   */
+  createAppSession(sessionId: string, provider: string, projectPath: string): string {
+    const db = getConnection();
+    const normalizedProjectPath = normalizeProjectPathForProvider(provider, projectPath);
+
+    projectsDb.createProjectPath(normalizedProjectPath);
+
+    db.prepare(
+      `INSERT INTO sessions (session_id, provider, provider_session_id, custom_name, project_path, jsonl_path, isArchived, created_at, updated_at)
+       VALUES (?, ?, NULL, NULL, ?, NULL, 0, CURRENT_TIMESTAMP, CURRENT_TIMESTAMP)`
+    ).run(sessionId, provider, normalizedProjectPath);
+
     return sessionId;
   },
 
+  /**
+   * Records the provider-native session id for one app-allocated session.
+   *
+   * If the filesystem watcher indexed the provider transcript before this
+   * mapping was recorded (a duplicate row keyed by the provider id exists),
+   * the duplicate is merged into the app row: its transcript path and name
+   * are adopted and the duplicate row is removed. Runs in a transaction so
+   * the sidebar can never observe both rows at once.
+   */
+  assignProviderSessionId(sessionId: string, providerSessionId: string): void {
+    const db = getConnection();
+
+    const merge = db.transaction(() => {
+      const duplicate = db
+        .prepare(
+          `SELECT ${SESSION_ROW_COLUMNS} FROM sessions
+           WHERE (session_id = ? OR provider_session_id = ?)
+             AND session_id <> ?
+           LIMIT 1`
+        )
+        .get(providerSessionId, providerSessionId, sessionId) as SessionRow | undefined;
+
+      if (duplicate) {
+        db.prepare('DELETE FROM sessions WHERE session_id = ?').run(duplicate.session_id);
+        db.prepare(
+          `UPDATE sessions SET
+             provider_session_id = ?,
+             jsonl_path = COALESCE(jsonl_path, ?),
+             custom_name = COALESCE(custom_name, ?),
+             updated_at = CURRENT_TIMESTAMP
+           WHERE session_id = ?`
+        ).run(providerSessionId, duplicate.jsonl_path, duplicate.custom_name, sessionId);
+        return;
+      }
+
+      db.prepare(
+        `UPDATE sessions SET
+           provider_session_id = ?,
+           updated_at = CURRENT_TIMESTAMP
+         WHERE session_id = ?`
+      ).run(providerSessionId, sessionId);
+    });
+
+    merge();
+  },
+
   updateSessionCustomName(sessionId: string, customName: string): void {
     const db = getConnection();
     db.prepare(
@@ -85,17 +195,39 @@ export const sessionsDb = {
     ).run(customName, sessionId);
   },
 
-  getSessionById(sessionId: string): SessionMetadataLookupRow | null {
+  getSessionById(sessionId: string): SessionRow | null {
     const db = getConnection();
     const row = db
       .prepare(
-        `SELECT session_id, provider, project_path, jsonl_path, custom_name, isArchived, created_at, updated_at
+        `SELECT ${SESSION_ROW_COLUMNS}
          FROM sessions
          WHERE session_id = ?
          ORDER BY updated_at DESC
          LIMIT 1`
       )
-      .get(sessionId) as SessionMetadataLookupRow | undefined;
+      .get(sessionId) as SessionRow | undefined;
+
+    return row ?? null;
+  },
+
+  /**
+   * Resolves one session row through the provider-native id.
+   *
+   * The filesystem watcher only knows provider ids (they come from transcript
+   * file names), so it uses this lookup to translate disk artifacts back to
+   * the app-facing session row before broadcasting sidebar updates.
+   */
+  getSessionByProviderSessionId(providerSessionId: string): SessionRow | null {
+    const db = getConnection();
+    const row = db
+      .prepare(
+        `SELECT ${SESSION_ROW_COLUMNS}
+         FROM sessions
+         WHERE provider_session_id = ?
+         ORDER BY updated_at DESC
+         LIMIT 1`
+      )
+      .get(providerSessionId) as SessionRow | undefined;
 
     return row ?? null;
   },
@@ -104,7 +236,7 @@ export const sessionsDb = {
     const db = getConnection();
     return db
       .prepare(
-        `SELECT session_id, provider, project_path, jsonl_path, custom_name, isArchived, created_at, updated_at
+        `SELECT ${SESSION_ROW_COLUMNS}
          FROM sessions
          WHERE isArchived = 0`
       )
@@ -119,7 +251,7 @@ export const sessionsDb = {
     const db = getConnection();
     return db
       .prepare(
-        `SELECT session_id, provider, project_path, jsonl_path, custom_name, isArchived, created_at, updated_at
+        `SELECT ${SESSION_ROW_COLUMNS}
          FROM sessions
          WHERE isArchived = 1
          ORDER BY datetime(COALESCE(updated_at, created_at)) DESC, session_id DESC`
@@ -132,7 +264,7 @@ export const sessionsDb = {
     const normalizedProjectPath = normalizeProjectPath(projectPath);
     return db
       .prepare(
-        `SELECT session_id, provider, project_path, jsonl_path, custom_name, isArchived, created_at, updated_at
+        `SELECT ${SESSION_ROW_COLUMNS}
          FROM sessions
          WHERE project_path = ?
            AND isArchived = 0`
@@ -149,7 +281,7 @@ export const sessionsDb = {
     const normalizedProjectPath = normalizeProjectPath(projectPath);
     return db
       .prepare(
-        `SELECT session_id, provider, project_path, jsonl_path, custom_name, isArchived, created_at, updated_at
+        `SELECT ${SESSION_ROW_COLUMNS}
          FROM sessions
          WHERE project_path = ?`
       )
@@ -161,7 +293,7 @@ export const sessionsDb = {
     const normalizedProjectPath = normalizeProjectPath(projectPath);
     return db
       .prepare(
-        `SELECT session_id, provider, project_path, jsonl_path, custom_name, isArchived, created_at, updated_at
+        `SELECT ${SESSION_ROW_COLUMNS}
          FROM sessions
          WHERE project_path = ?
            AND isArchived = 0

server/modules/database/schema.ts

@@ -83,6 +83,12 @@ export const SESSIONS_TABLE_SCHEMA_SQL = `
 CREATE TABLE IF NOT EXISTS sessions (
     session_id TEXT NOT NULL,
     provider TEXT NOT NULL DEFAULT 'claude',
+    -- The session id used by the provider CLI/SDK on disk (JSONL file name,
+    -- store.db folder, sqlite row id, ...). \`session_id\` is the stable
+    -- app-facing id that the frontend uses for the whole session lifetime;
+    -- \`provider_session_id\` is filled in once the provider announces its own
+    -- id mid-run, or equals \`session_id\` for sessions discovered on disk.
+    provider_session_id TEXT,
     custom_name TEXT,
     project_path TEXT,
     jsonl_path TEXT,

server/modules/database/tests/sessions-provider-mapping.test.ts

@@ -0,0 +1,108 @@
+import assert from 'node:assert/strict';
+import { mkdtemp, rm } from 'node:fs/promises';
+import { tmpdir } from 'node:os';
+import path from 'node:path';
+import test from 'node:test';
+
+import { closeConnection } from '@/modules/database/connection.js';
+import { initializeDatabase } from '@/modules/database/init-db.js';
+import { sessionsDb } from '@/modules/database/repositories/sessions.db.js';
+
+async function withIsolatedDatabase(runTest: () => void | Promise<void>): Promise<void> {
+  const previousDatabasePath = process.env.DATABASE_PATH;
+  const tempDirectory = await mkdtemp(path.join(tmpdir(), 'sessions-mapping-'));
+  const databasePath = path.join(tempDirectory, 'auth.db');
+
+  closeConnection();
+  process.env.DATABASE_PATH = databasePath;
+  await initializeDatabase();
+
+  try {
+    await runTest();
+  } finally {
+    closeConnection();
+    if (previousDatabasePath === undefined) {
+      delete process.env.DATABASE_PATH;
+    } else {
+      process.env.DATABASE_PATH = previousDatabasePath;
+    }
+    await rm(tempDirectory, { recursive: true, force: true });
+  }
+}
+
+test('disk-discovered sessions are keyed by the provider id for both columns', async () => {
+  await withIsolatedDatabase(() => {
+    sessionsDb.createSession('provider-abc', 'claude', '/workspace/demo', 'From Disk');
+
+    const row = sessionsDb.getSessionById('provider-abc');
+    assert.equal(row?.session_id, 'provider-abc');
+    assert.equal(row?.provider_session_id, 'provider-abc');
+
+    const byProviderId = sessionsDb.getSessionByProviderSessionId('provider-abc');
+    assert.equal(byProviderId?.session_id, 'provider-abc');
+  });
+});
+
+test('app sessions get the provider id assigned without creating a duplicate row', async () => {
+  await withIsolatedDatabase(() => {
+    sessionsDb.createAppSession('app-id-1', 'claude', '/workspace/demo');
+    sessionsDb.assignProviderSessionId('app-id-1', 'provider-xyz');
+
+    // A later synchronizer pass that discovers the transcript on disk must
+    // update the app row in place instead of inserting a provider-keyed row.
+    const returnedId = sessionsDb.createSession(
+      'provider-xyz',
+      'claude',
+      '/workspace/demo',
+      'Synced Name',
+      undefined,
+      undefined,
+      '/fake/path/provider-xyz.jsonl',
+    );
+
+    assert.equal(returnedId, 'app-id-1');
+    assert.equal(sessionsDb.getAllSessions().length, 1);
+
+    const row = sessionsDb.getSessionById('app-id-1');
+    assert.equal(row?.provider_session_id, 'provider-xyz');
+    assert.equal(row?.jsonl_path, '/fake/path/provider-xyz.jsonl');
+  });
+});
+
+test('assignProviderSessionId merges a watcher-created duplicate into the app row', async () => {
+  await withIsolatedDatabase(() => {
+    sessionsDb.createAppSession('app-id-2', 'codex', '/workspace/demo');
+
+    // Simulate the race: the filesystem watcher indexed the provider
+    // transcript before the runtime announced its session id to the gateway.
+    sessionsDb.createSession(
+      'provider-race',
+      'codex',
+      '/workspace/demo',
+      'Watcher Name',
+      undefined,
+      undefined,
+      '/fake/provider-race.jsonl',
+    );
+    assert.equal(sessionsDb.getAllSessions().length, 2);
+
+    sessionsDb.assignProviderSessionId('app-id-2', 'provider-race');
+
+    const rows = sessionsDb.getAllSessions();
+    assert.equal(rows.length, 1);
+    assert.equal(rows[0]?.session_id, 'app-id-2');
+    assert.equal(rows[0]?.provider_session_id, 'provider-race');
+    // Transcript path and name from the duplicate are adopted.
+    assert.equal(rows[0]?.jsonl_path, '/fake/provider-race.jsonl');
+    assert.equal(rows[0]?.custom_name, 'Watcher Name');
+  });
+});
+
+test('legacy provider-keyed rows stay resolvable through both lookups', async () => {
+  await withIsolatedDatabase(() => {
+    sessionsDb.createSession('legacy-1', 'gemini', '/workspace/demo');
+
+    assert.equal(sessionsDb.getSessionById('legacy-1')?.provider, 'gemini');
+    assert.equal(sessionsDb.getSessionByProviderSessionId('legacy-1')?.session_id, 'legacy-1');
+  });
+});

server/modules/projects/projects.routes.ts

@@ -67,8 +67,17 @@ function resolveRouteErrorMessage(error: unknown): string {
 
 router.get(
   '/',
-  asyncHandler(async (_req, res) => {
-    const projects = await getProjectsWithSessions();
+  asyncHandler(async (req, res) => {
+    const skipSynchronization =
+      readQueryStringValue(req.query.skipSynchronization).trim() === '1' ||
+      readQueryStringValue(req.query.skipSync).trim() === '1';
+    const sessionsLimit = readOptionalNumericQueryValue(req.query.sessionsLimit) ?? undefined;
+    const sessionsOffset = readOptionalNumericQueryValue(req.query.sessionsOffset) ?? undefined;
+    const projects = await getProjectsWithSessions({
+      skipSynchronization,
+      sessionsLimit,
+      sessionsOffset,
+    });
     res.json(projects);
   }),
 );

server/modules/projects/services/projects-with-sessions-fetch.service.ts

@@ -189,10 +189,11 @@ function readProjectSessionsPageByPath(
   };
 }
 
-// Broadcast progress to all connected WebSocket clients
+// Broadcast progress to all connected WebSocket clients.
+// Uses the unified `kind` envelope like every other websocket frame.
 function broadcastProgress(progress: ProgressUpdate) {
   const message = JSON.stringify({
-    type: 'loading_progress',
+    kind: 'loading_progress',
     ...progress,
   });
 

server/modules/providers/README.md

@@ -83,7 +83,7 @@ The existing provider folders are `claude`, `codex`, `cursor`, `gemini`, and
 - Update `server/modules/providers/provider.routes.ts`.
 - Update `server/routes/agent.js` if the provider is launchable from the agent runtime.
 - Update `server/index.js` if the provider needs runtime boot or shutdown wiring.
-- Update `public/modelConstants.js` if the provider appears in README or public API docs.
+- Update the `PROVIDER_ORDER` list in `public/api-docs.html` if the provider should appear in the public API docs.
 - Update `src/components/chat/hooks/useChatProviderState.ts` and
   `src/components/chat/view/subcomponents/ProviderSelectionEmptyState.tsx` if
   the provider should be selectable in chat.

server/modules/providers/list/claude/claude-models.provider.ts

@@ -18,18 +18,28 @@ export const CLAUDE_FALLBACK_MODELS: ProviderModelsDefinition = {
     {
       value: 'default',
       label: 'Default (recommended)',
-      description: 'Use the default model (currently Opus 4.7 (1M context)) · $5/$25 per Mtok',
+      description: 'Use the default model (currently Opus 4.8 (1M context)) · $5/$25 per Mtok',
     },
     {
-      value: 'sonnet',
-      label: 'Sonnet',
-      description: 'Sonnet 4.6 · Best for everyday tasks · $3/$15 per Mtok',
+      value: 'fable',
+      label: 'Fable',
+      description: 'Fable 5 · Most capable for your hardest and longest-running tasks · Uses your limits ~2× faster than Opus',
+    },
+    {
+      value: "sonnet",
+      label: "Sonnet",
+      description: "Sonnet 4.6 · Best for everyday tasks · $3/$15 per Mtok",
     },
     {
       value: 'sonnet[1m]',
       label: 'Sonnet (1M context)',
       description: 'Sonnet 4.6 for long sessions · $3/$15 per Mtok',
     },
+    {
+      value: 'opus[1m]',
+      label: 'Opus 4.8 (1M context)',
+      description: 'Opus 4.8 with 1M context · Most capable for complex work · $5/$25 per Mtok',
+    },
     {
       value: 'haiku',
       label: 'Haiku',

server/modules/providers/list/claude/claude-session-synchronizer.provider.ts

@@ -111,7 +111,10 @@ export class ClaudeSessionSynchronizer implements IProviderSessionSynchronizer {
       return null;
     }
 
-    const existingSession = sessionsDb.getSessionById(parsed.sessionId);
+    // App-created sessions are keyed by an app id, so disk-discovered provider
+    // ids must be resolved through the provider-id mapping first.
+    const existingSession = sessionsDb.getSessionByProviderSessionId(parsed.sessionId)
+      ?? sessionsDb.getSessionById(parsed.sessionId);
     const existingSessionName = existingSession?.custom_name;
     if (existingSessionName && existingSessionName !== 'Untitled Claude Session') {
       return {

server/modules/providers/list/claude/claude-sessions.provider.ts

@@ -5,7 +5,7 @@ import readline from 'node:readline';
 
 import type { IProviderSessions } from '@/shared/interfaces.js';
 import type { AnyRecord, FetchHistoryOptions, FetchHistoryResult, NormalizedMessage } from '@/shared/types.js';
-import { createNormalizedMessage, generateMessageId, readObjectRecord } from '@/shared/utils.js';
+import { createNormalizedMessage, generateMessageId, readObjectRecord, sliceTailPage } from '@/shared/utils.js';
 import { sessionsDb } from '@/modules/database/index.js';
 
 const PROVIDER = 'claude';
@@ -103,10 +103,13 @@ async function parseAgentTools(filePath: string): Promise<AnyRecord[]> {
 
 async function getSessionMessages(
   sessionId: string,
+  providerSessionId: string,
   limit: number | null,
   offset: number,
 ): Promise<ClaudeHistoryMessagesResult> {
   try {
+    // The DB row is keyed by the app-facing session id, while the JSONL rows
+    // on disk carry the provider-native id — both ids are needed here.
     const jsonLPath = sessionsDb.getSessionById(sessionId)?.jsonl_path;
 
     if (!jsonLPath) {
@@ -133,7 +136,7 @@ async function getSessionMessages(
 
       try {
         const entry = JSON.parse(line) as AnyRecord;
-        if (entry.sessionId === sessionId) {
+        if (entry.sessionId === providerSessionId) {
           messages.push(entry);
         }
       } catch {
@@ -553,12 +556,13 @@ export class ClaudeSessionsProvider implements IProviderSessions {
     options: FetchHistoryOptions = {},
   ): Promise<FetchHistoryResult> {
     const { limit = null, offset = 0 } = options;
+    const providerSessionId = options.providerSessionId ?? sessionId;
 
     let result: ClaudeHistoryResult;
     try {
       // Load full history first so `total` reflects frontend-normalized messages,
       // not raw JSONL records.
-      result = await getSessionMessages(sessionId, null, 0);
+      result = await getSessionMessages(sessionId, providerSessionId, null, 0);
     } catch (error) {
       const message = error instanceof Error ? error.message : String(error);
       console.warn(`[ClaudeProvider] Failed to load session ${sessionId}:`, message);
@@ -606,7 +610,6 @@ export class ClaudeSessionsProvider implements IProviderSessions {
       }
     }
 
-    const totalNormalized = normalized.length;
     let total = 0;
     for (const msg of normalized) {
       if (msg.kind !== 'tool_result') {
@@ -615,18 +618,10 @@ export class ClaudeSessionsProvider implements IProviderSessions {
     }
     const normalizedOffset = Math.max(0, offset);
     const normalizedLimit = limit === null ? null : Math.max(0, limit);
-    const messages = normalizedLimit === null
-      ? normalized
-      : normalized.slice(
-          Math.max(0, totalNormalized - normalizedOffset - normalizedLimit),
-          Math.max(0, totalNormalized - normalizedOffset),
-        );
-    const hasMore = normalizedLimit === null
-      ? false
-      : Math.max(0, totalNormalized - normalizedOffset - normalizedLimit) > 0;
+    const { page, hasMore } = sliceTailPage(normalized, normalizedLimit, normalizedOffset);
 
     return {
-      messages,
+      messages: page,
       total,
       hasMore,
       offset: normalizedOffset,

server/modules/providers/list/codex/codex-session-synchronizer.provider.ts

@@ -43,11 +43,12 @@ export class CodexSessionSynchronizer implements IProviderSessionSynchronizer {
         continue;
       }
 
-      const existingSession = sessionsDb.getSessionById(parsed.sessionId);
+      const existingSession = sessionsDb.getSessionByProviderSessionId(parsed.sessionId)
+        ?? sessionsDb.getSessionById(parsed.sessionId);
       if (existingSession) {
         // If session name is untitled and we now have a name, update it
         if (existingSession.custom_name === 'Untitled Codex Session' && parsed.sessionName && parsed.sessionName !== 'Untitled Codex Session') {
-          sessionsDb.updateSessionCustomName(parsed.sessionId, parsed.sessionName);
+          sessionsDb.updateSessionCustomName(existingSession.session_id, parsed.sessionName);
         }
       }
 
@@ -120,7 +121,10 @@ export class CodexSessionSynchronizer implements IProviderSessionSynchronizer {
       return null;
     }
 
-    const existingSession = sessionsDb.getSessionById(parsed.sessionId);
+    // App-created sessions are keyed by an app id, so disk-discovered provider
+    // ids must be resolved through the provider-id mapping first.
+    const existingSession = sessionsDb.getSessionByProviderSessionId(parsed.sessionId)
+      ?? sessionsDb.getSessionById(parsed.sessionId);
     const existingSessionName = existingSession?.custom_name;
     if (existingSessionName && existingSessionName !== 'Untitled Codex Session') {
       return {

server/modules/providers/list/codex/codex-sessions.provider.ts

@@ -4,7 +4,7 @@ import readline from 'node:readline';
 import { sessionsDb } from '@/modules/database/index.js';
 import type { IProviderSessions } from '@/shared/interfaces.js';
 import type { AnyRecord, FetchHistoryOptions, FetchHistoryResult, NormalizedMessage } from '@/shared/types.js';
-import { createNormalizedMessage, generateMessageId, readObjectRecord } from '@/shared/utils.js';
+import { createNormalizedMessage, generateMessageId, readObjectRecord, sliceTailPage } from '@/shared/utils.js';
 
 const PROVIDER = 'codex';
 
@@ -552,7 +552,6 @@ export class CodexSessionsProvider implements IProviderSessions {
       }
     }
 
-    const totalNormalized = normalized.length;
     let total = 0;
     for (const msg of normalized) {
       if (msg.kind !== 'tool_result') {
@@ -561,18 +560,10 @@ export class CodexSessionsProvider implements IProviderSessions {
     }
     const normalizedOffset = Math.max(0, offset);
     const normalizedLimit = limit === null ? null : Math.max(0, limit);
-    const messages = normalizedLimit === null
-      ? normalized
-      : normalized.slice(
-          Math.max(0, totalNormalized - normalizedOffset - normalizedLimit),
-          Math.max(0, totalNormalized - normalizedOffset),
-        );
-    const hasMore = normalizedLimit === null
-      ? false
-      : Math.max(0, totalNormalized - normalizedOffset - normalizedLimit) > 0;
+    const { page, hasMore } = sliceTailPage(normalized, normalizedLimit, normalizedOffset);
 
     return {
-      messages,
+      messages: page,
       total,
       hasMore,
       offset: normalizedOffset,

server/modules/providers/list/cursor/cursor-sessions.provider.ts

@@ -9,6 +9,7 @@ import {
   generateMessageId,
   readObjectRecord,
   sanitizeLeafDirectoryName,
+  sliceTailPage,
 } from '@/shared/utils.js';
 
 const PROVIDER = 'cursor';
@@ -363,42 +364,32 @@ export class CursorSessionsProvider implements IProviderSessions {
 
   /**
    * Fetches and paginates Cursor session history from its project-scoped store.db.
+   *
+   * Pagination follows the shared tail contract (`sliceTailPage`): offset 0 is
+   * the most recent page, matching every other provider.
    */
   async fetchHistory(
     sessionId: string,
     options: FetchHistoryOptions = {},
   ): Promise<FetchHistoryResult> {
     const { projectPath = '', limit = null, offset = 0 } = options;
+    // The store.db folder on disk is named after the provider-native id, not
+    // the app-facing session id this method is addressed with.
+    const providerSessionId = options.providerSessionId ?? sessionId;
 
     try {
-      const blobs = await this.loadCursorBlobs(sessionId, projectPath);
+      const blobs = await this.loadCursorBlobs(providerSessionId, projectPath);
       const allNormalized = this.normalizeCursorBlobs(blobs, sessionId);
       const renderableMessages = allNormalized.filter((msg) => msg.kind !== 'tool_result');
       const total = renderableMessages.length;
-
-      if (limit !== null) {
-        const start = offset;
-        const page = limit === 0
-          ? []
-          : renderableMessages.slice(start, start + limit);
-        const hasMore = limit === 0
-          ? start < total
-          : start + limit < total;
-        return {
-          messages: page,
-          total,
-          hasMore,
-          offset,
-          limit,
-        };
-      }
+      const { page, hasMore } = sliceTailPage(renderableMessages, limit, offset);
 
       return {
-        messages: renderableMessages,
+        messages: page,
         total,
-        hasMore: false,
-        offset: 0,
-        limit: null,
+        hasMore,
+        offset,
+        limit,
       };
     } catch (error) {
       const message = error instanceof Error ? error.message : String(error);

server/modules/providers/list/gemini/gemini-models.provider.ts

@@ -12,17 +12,14 @@ import {
 
 export const GEMINI_FALLBACK_MODELS: ProviderModelsDefinition = {
   OPTIONS: [
-    { value: 'gemini-3.1-pro-preview', label: 'Gemini 3.1 Pro Preview' },
-    { value: 'gemini-3-pro-preview', label: 'Gemini 3 Pro Preview' },
     { value: 'gemini-3-flash-preview', label: 'Gemini 3 Flash Preview' },
+    { value: 'gemini-3.1-flash-lite-preview', label: 'Gemini 3.1 Flash Lite Preview' },
     { value: 'gemini-2.5-flash', label: 'Gemini 2.5 Flash' },
-    { value: 'gemini-2.5-pro', label: 'Gemini 2.5 Pro' },
-    { value: 'gemini-2.0-flash-lite', label: 'Gemini 2.0 Flash Lite' },
-    { value: 'gemini-2.0-flash', label: 'Gemini 2.0 Flash' },
-    { value: 'gemini-2.0-pro-exp', label: 'Gemini 2.0 Pro Experimental' },
-    { value: 'gemini-2.0-flash-thinking-exp', label: 'Gemini 2.0 Flash Thinking' },
+    { value: 'gemini-2.5-flash-lite', label: 'Gemini 2.5 Flash Lite' },
+    { value: 'gemma-4-31b-it', label: 'Gemma 4 31B IT' },
+    { value: 'gemma-4-26b-a4b-it', label: 'Gemma 4 26B A4B IT' },
   ],
-  DEFAULT: 'gemini-3.1-pro-preview',
+  DEFAULT: 'gemini-3-flash-preview',
 };
 
 export class GeminiProviderModels implements IProviderModels {

server/modules/providers/list/gemini/gemini-sessions.provider.ts

@@ -5,7 +5,7 @@ import readline from 'node:readline';
 import { sessionsDb } from '@/modules/database/index.js';
 import type { IProviderSessions } from '@/shared/interfaces.js';
 import type { AnyRecord, FetchHistoryOptions, FetchHistoryResult, NormalizedMessage } from '@/shared/types.js';
-import { createNormalizedMessage, generateMessageId, readObjectRecord } from '@/shared/utils.js';
+import { createNormalizedMessage, generateMessageId, readObjectRecord, sliceTailPage } from '@/shared/utils.js';
 
 const PROVIDER = 'gemini';
 
@@ -518,9 +518,9 @@ export class GeminiSessionsProvider implements IProviderSessions {
 
     const start = Math.max(0, offset);
     const pageLimit = limit === null ? null : Math.max(0, limit);
-    const messages = pageLimit === null
-      ? normalized.slice(start)
-      : normalized.slice(start, start + pageLimit);
+    // Tail pagination via the shared contract: offset 0 returns the most
+    // recent page, matching every other provider.
+    const { page, hasMore } = sliceTailPage(normalized, pageLimit, start);
     let total = 0;
     for (const msg of normalized) {
       if (msg.kind !== 'tool_result') {
@@ -529,9 +529,9 @@ export class GeminiSessionsProvider implements IProviderSessions {
     }
 
     return {
-      messages,
+      messages: page,
       total,
-      hasMore: pageLimit === null ? false : start + pageLimit < normalized.length,
+      hasMore,
       offset: start,
       limit: pageLimit,
       tokenUsage: result.tokenUsage,

server/modules/providers/list/opencode/opencode-session-synchronizer.provider.ts

@@ -112,7 +112,10 @@ export class OpenCodeSessionSynchronizer implements IProviderSessionSynchronizer
     }
 
     const fallbackTitle = 'Untitled OpenCode Session';
-    const existingSession = sessionsDb.getSessionById(sessionId);
+    // App-created sessions are keyed by an app id, so disk-discovered provider
+    // ids must be resolved through the provider-id mapping first.
+    const existingSession = sessionsDb.getSessionByProviderSessionId(sessionId)
+      ?? sessionsDb.getSessionById(sessionId);
     const existingName = existingSession?.custom_name;
     const nextName = existingName && existingName !== fallbackTitle
       ? existingName

server/modules/providers/list/opencode/opencode-sessions.provider.ts

@@ -12,6 +12,7 @@ import {
   readObjectRecord,
   readJsonRecord,
   readOptionalString,
+  sliceTailPage,
 } from '@/shared/utils.js';
 
 const PROVIDER = 'opencode';
@@ -325,6 +326,9 @@ export class OpenCodeSessionsProvider implements IProviderSessions {
     options: FetchHistoryOptions = {},
   ): Promise<FetchHistoryResult> {
     const { limit = null, offset = 0 } = options;
+    // OpenCode's shared sqlite database keys messages by the provider-native
+    // session id, not the app-facing id this method is addressed with.
+    const providerSessionId = options.providerSessionId ?? sessionId;
     const db = openOpenCodeDatabase();
     if (!db) {
       return { messages: [], total: 0, hasMore: false, offset: 0, limit: null };
@@ -349,27 +353,20 @@ export class OpenCodeSessionsProvider implements IProviderSessions {
           m.id,
           COALESCE(p.time_created, 0),
           p.id
-      `).all(sessionId) as OpenCodeHistoryRow[];
+      `).all(providerSessionId) as OpenCodeHistoryRow[];
 
       const normalized = this.normalizeHistoryRows(rows, sessionId);
-      const tokenUsage = aggregateOpenCodeSessionTokenUsage(db, sessionId);
+      const tokenUsage = aggregateOpenCodeSessionTokenUsage(db, providerSessionId);
 
       const normalizedOffset = Math.max(0, offset);
       const normalizedLimit = limit === null ? null : Math.max(0, limit);
       const total = normalized.length;
-      const messages = normalizedLimit === null
-        ? normalized
-        : normalized.slice(
-            Math.max(0, total - normalizedOffset - normalizedLimit),
-            Math.max(0, total - normalizedOffset),
-          );
+      const { page, hasMore } = sliceTailPage(normalized, normalizedLimit, normalizedOffset);
 
       return {
-        messages,
+        messages: page,
         total,
-        hasMore: normalizedLimit === null
-          ? false
-          : Math.max(0, total - normalizedOffset - normalizedLimit) > 0,
+        hasMore,
         offset: normalizedOffset,
         limit: normalizedLimit,
         tokenUsage,

server/modules/providers/provider.routes.ts

@@ -1,6 +1,7 @@
 import express, { type Request, type Response } from 'express';
 
 import { providerAuthService } from '@/modules/providers/services/provider-auth.service.js';
+import { providerCapabilitiesService } from '@/modules/providers/services/provider-capabilities.service.js';
 import { providerMcpService } from '@/modules/providers/services/mcp.service.js';
 import { providerModelsService } from '@/modules/providers/services/provider-models.service.js';
 import { providerSkillsService } from '@/modules/providers/services/skills.service.js';
@@ -382,7 +383,51 @@ router.post(
   }),
 );
 
+router.get(
+  '/capabilities',
+  asyncHandler(async (_req: Request, res: Response) => {
+    res.json(createApiSuccessResponse({
+      providers: providerCapabilitiesService.listAllProviderCapabilities(),
+    }));
+  }),
+);
+
+router.get(
+  '/:provider/capabilities',
+  asyncHandler(async (req: Request, res: Response) => {
+    const provider = parseProvider(req.params.provider);
+    res.json(createApiSuccessResponse(
+      providerCapabilitiesService.getProviderCapabilities(provider),
+    ));
+  }),
+);
+
 // ----------------- Session routes -----------------
+/**
+ * Session gateway entry point: allocates the stable app-facing session id for
+ * a brand-new chat. The frontend must call this before the first `chat.send`
+ * so the session id in the URL, the store, and the websocket all agree from
+ * the very first message — there is no client-visible session-id handoff.
+ */
+router.post(
+  '/sessions',
+  asyncHandler(async (req: Request, res: Response) => {
+    const body = (req.body ?? {}) as Record<string, unknown>;
+    const provider = parseProvider(body.provider);
+    const projectPath = typeof body.projectPath === 'string' ? body.projectPath : '';
+    const result = sessionsService.createAppSession(provider, projectPath);
+    res.status(201).json(createApiSuccessResponse(result));
+  }),
+);
+
+router.get(
+  '/sessions/running',
+  asyncHandler(async (_req: Request, res: Response) => {
+    const sessions = sessionsService.listRunningSessions();
+    res.json(createApiSuccessResponse({ sessions }));
+  }),
+);
+
 router.get(
   '/sessions/archived',
   asyncHandler(async (_req: Request, res: Response) => {
@@ -459,7 +504,7 @@ router.get(
       limit,
       offset,
     });
-    res.json(result);
+    res.json(createApiSuccessResponse(result));
   }),
 );
 

server/modules/providers/services/provider-capabilities.service.ts

@@ -0,0 +1,91 @@
+import type { LLMProvider } from '@/shared/types.js';
+
+/**
+ * Static, backend-owned description of what one provider integration supports.
+ *
+ * The frontend renders its composer UI (permission mode picker, image upload,
+ * abort button, ...) purely from this shape, which is what keeps the frontend
+ * free of per-provider conditionals. New provider features should be exposed
+ * here instead of branching on the provider id in React components.
+ */
+type ProviderCapabilities = {
+  provider: LLMProvider;
+  /** Permission modes the provider runtime understands, in cycle order. */
+  permissionModes: string[];
+  defaultPermissionMode: string;
+  /** Whether image attachments can be included in a chat.send. */
+  supportsImages: boolean;
+  /** Whether an in-flight run can be cancelled via chat.abort. */
+  supportsAbort: boolean;
+  /** Whether interactive tool permission prompts can reach the UI. */
+  supportsPermissionRequests: boolean;
+  /** Whether the token-usage endpoint has data for this provider. */
+  supportsTokenUsage: boolean;
+};
+
+/**
+ * The capability matrix mirrors what each runtime actually implements today:
+ * - permission modes match the option sets accepted by each CLI/SDK.
+ * - only the Claude SDK integration surfaces interactive permission requests.
+ * - Cursor has no token usage endpoint support (its store.db has no usage rows).
+ */
+const PROVIDER_CAPABILITIES: Record<LLMProvider, ProviderCapabilities> = {
+  claude: {
+    provider: 'claude',
+    permissionModes: ['default', 'auto', 'acceptEdits', 'bypassPermissions', 'plan'],
+    defaultPermissionMode: 'default',
+    supportsImages: true,
+    supportsAbort: true,
+    supportsPermissionRequests: true,
+    supportsTokenUsage: true,
+  },
+  cursor: {
+    provider: 'cursor',
+    permissionModes: ['default', 'acceptEdits', 'bypassPermissions', 'plan'],
+    defaultPermissionMode: 'default',
+    supportsImages: false,
+    supportsAbort: true,
+    supportsPermissionRequests: false,
+    supportsTokenUsage: false,
+  },
+  codex: {
+    provider: 'codex',
+    permissionModes: ['default', 'acceptEdits', 'bypassPermissions'],
+    defaultPermissionMode: 'default',
+    supportsImages: false,
+    supportsAbort: true,
+    supportsPermissionRequests: false,
+    supportsTokenUsage: true,
+  },
+  gemini: {
+    provider: 'gemini',
+    permissionModes: ['default', 'acceptEdits', 'bypassPermissions', 'plan'],
+    defaultPermissionMode: 'default',
+    supportsImages: false,
+    supportsAbort: true,
+    supportsPermissionRequests: false,
+    supportsTokenUsage: true,
+  },
+  opencode: {
+    provider: 'opencode',
+    permissionModes: ['default'],
+    defaultPermissionMode: 'default',
+    supportsImages: false,
+    supportsAbort: true,
+    supportsPermissionRequests: false,
+    supportsTokenUsage: true,
+  },
+};
+
+/**
+ * Application service exposing the provider capability matrix.
+ */
+export const providerCapabilitiesService = {
+  getProviderCapabilities(provider: LLMProvider): ProviderCapabilities {
+    return PROVIDER_CAPABILITIES[provider];
+  },
+
+  listAllProviderCapabilities(): ProviderCapabilities[] {
+    return Object.values(PROVIDER_CAPABILITIES);
+  },
+};

server/modules/providers/services/provider-models.service.ts

@@ -17,6 +17,7 @@ import { readProviderSessionActiveModelChange } from '@/shared/utils.js';
 
 export const PROVIDER_MODELS_CACHE_TTL_MS = 3 * 24 * 60 * 60 * 1000;
 const PROVIDER_MODELS_CACHE_VERSION = 1;
+const UNCACHED_PROVIDERS = new Set<LLMProvider>(['claude', 'gemini']);
 
 type ProviderModelsServiceDependencies = {
   resolveProvider?: (provider: LLMProvider) => Pick<IProvider, 'models'>;
@@ -232,10 +233,42 @@ export const createProviderModelsService = (dependencies: ProviderModelsServiceD
     return request;
   };
 
+  const loadDirectModels = (
+    provider: LLMProvider,
+  ): Promise<ProviderModelsResult> => {
+    const request = resolveProvider(provider).models.getSupportedModels()
+      .then((models) => {
+        const currentTime = now();
+        return {
+          models,
+          cache: {
+            updatedAt: new Date(currentTime).toISOString(),
+            expiresAt: new Date(currentTime).toISOString(),
+            source: 'fresh' as const,
+          },
+        };
+      })
+      .finally(() => {
+        pendingRequests.delete(provider);
+      });
+
+    pendingRequests.set(provider, request);
+    return request;
+  };
+
   const getProviderModels = async (
     provider: LLMProvider,
     options: ProviderModelsOptions = {},
   ): Promise<ProviderModelsResult> => {
+    if (UNCACHED_PROVIDERS.has(provider)) {
+      const pendingRequest = pendingRequests.get(provider);
+      if (pendingRequest) {
+        return pendingRequest;
+      }
+
+      return loadDirectModels(provider);
+    }
+
     if (options.bypassCache) {
       const pendingRequest = pendingRequests.get(provider);
       if (pendingRequest) {

server/modules/providers/services/sessions-watcher.service.ts

@@ -4,10 +4,11 @@ import { promises as fsPromises } from 'node:fs';
 
 import chokidar, { type FSWatcher } from 'chokidar';
 
+import { projectsDb, sessionsDb } from '@/modules/database/index.js';
 import { sessionSynchronizerService } from '@/modules/providers/services/session-synchronizer.service.js';
 import { WS_OPEN_STATE, connectedClients } from '@/modules/websocket/index.js';
 import type { LLMProvider } from '@/shared/types.js';
-import { getProjectsWithSessions } from '@/modules/projects/index.js';
+import { generateDisplayName } from '@/modules/projects/index.js';
 
 type WatcherEventType = 'add' | 'change';
 
@@ -58,6 +59,11 @@ const watchers: FSWatcher[] = [];
 type PendingWatcherUpdate = {
   providers: Set<LLMProvider>;
   changeTypes: Set<WatcherEventType>;
+  /**
+   * Provider-native session ids reported by the synchronizers. They are
+   * translated back to app-facing session rows at flush time, because the
+   * transcript file names on disk only ever contain provider ids.
+   */
   updatedSessionIds: Set<string>;
 };
 
@@ -131,6 +137,50 @@ function queuePendingWatcherUpdate(
   schedulePendingWatcherFlush();
 }
 
+/**
+ * Builds one `session_upserted` delta event for a provider-native session id.
+ *
+ * The event carries everything a sidebar needs to upsert the session in place
+ * (session summary plus owning-project metadata), so clients never need a full
+ * project-list refetch when a transcript file changes on disk. Returns `null`
+ * when the id cannot be resolved to an indexed session row.
+ */
+async function buildSessionUpsertedEvent(updatedProviderSessionId: string): Promise<string | null> {
+  const row = sessionsDb.getSessionByProviderSessionId(updatedProviderSessionId)
+    ?? sessionsDb.getSessionById(updatedProviderSessionId);
+  if (!row || row.isArchived) {
+    return null;
+  }
+
+  const projectPath = row.project_path;
+  const project = projectPath ? projectsDb.getProjectPath(projectPath) : null;
+  const displayName = project?.custom_project_name?.trim()
+    ? project.custom_project_name
+    : await generateDisplayName(path.basename(projectPath ?? '') || (projectPath ?? ''), projectPath);
+
+  return JSON.stringify({
+    kind: 'session_upserted',
+    sessionId: row.session_id,
+    provider: row.provider,
+    session: {
+      id: row.session_id,
+      summary: row.custom_name || '',
+      messageCount: 0,
+      lastActivity: row.updated_at ?? row.created_at ?? new Date().toISOString(),
+    },
+    project: project
+      ? {
+        projectId: project.project_id,
+        path: project.project_path,
+        fullPath: project.project_path,
+        displayName,
+        isStarred: Boolean(project.isStarred),
+      }
+      : null,
+    timestamp: new Date().toISOString(),
+  });
+}
+
 async function flushPendingWatcherUpdate(): Promise<void> {
   clearPendingWatcherFlushTimer();
 
@@ -149,33 +199,29 @@ async function flushPendingWatcherUpdate(): Promise<void> {
   watcherRefreshInFlight = true;
 
   try {
-    const updatedProjects = await getProjectsWithSessions({ skipSynchronization: true });
-    const changeTypes = Array.from(queuedUpdate.changeTypes);
-    const watchProviders = Array.from(queuedUpdate.providers);
-    const updatedSessionIds = Array.from(queuedUpdate.updatedSessionIds);
-
-    // Backward-compatible fields stay populated with the first queued values.
-    const updateMessage = JSON.stringify({
-      type: 'projects_updated',
-      projects: updatedProjects,
-      timestamp: new Date().toISOString(),
-      changeType: changeTypes[0] ?? 'change',
-      updatedSessionId: updatedSessionIds[0] ?? undefined,
-      watchProvider: watchProviders[0] ?? undefined,
-      changeTypes,
-      updatedSessionIds,
-      watchProviders,
-      batched: true,
-    });
-
-    connectedClients.forEach(client => {
-      if (client.readyState === WS_OPEN_STATE) {
-        client.send(updateMessage);
+    // Per-session deltas instead of full project snapshots: an upsert of one
+    // session can never clobber unrelated client state, so the frontend needs
+    // no "suppress updates while a run is active" protection logic.
+    const events: string[] = [];
+    for (const updatedSessionId of queuedUpdate.updatedSessionIds) {
+      const event = await buildSessionUpsertedEvent(updatedSessionId);
+      if (event) {
+        events.push(event);
       }
-    });
+    }
+
+    if (events.length > 0) {
+      connectedClients.forEach(client => {
+        if (client.readyState === WS_OPEN_STATE) {
+          for (const event of events) {
+            client.send(event);
+          }
+        }
+      });
+    }
   } catch (error) {
     const message = error instanceof Error ? error.message : String(error);
-    console.error('Session watcher refresh failed while broadcasting projects_updated', { error: message });
+    console.error('Session watcher refresh failed while broadcasting session_upserted', { error: message });
   } finally {
     watcherRefreshInFlight = false;
 

server/modules/providers/services/sessions.service.ts

@@ -1,7 +1,9 @@
+import { randomUUID } from 'node:crypto';
 import fsp from 'node:fs/promises';
 import path from 'node:path';
 
 import { projectsDb, sessionsDb } from '@/modules/database/index.js';
+import { chatRunRegistry } from '@/modules/websocket/index.js';
 import { providerRegistry } from '@/modules/providers/provider.registry.js';
 import type {
   FetchHistoryOptions,
@@ -11,6 +13,12 @@ import type {
 } from '@/shared/types.js';
 import { AppError } from '@/shared/utils.js';
 
+type CreateAppSessionResult = {
+  sessionId: string;
+  provider: LLMProvider;
+  projectPath: string;
+};
+
 type ArchivedSessionListItem = {
   sessionId: string;
   provider: LLMProvider;
@@ -77,6 +85,21 @@ export const sessionsService = {
     return providerRegistry.listProviders().map((provider) => provider.id);
   },
 
+  /**
+   * Returns app-facing ids for provider runs that are currently processing.
+   *
+   * This is intentionally status-only: callers that only need sidebar activity
+   * indicators should not attach to chat streams or request replayed messages.
+   */
+  listRunningSessions(): Array<{
+    sessionId: string;
+    provider: LLMProvider;
+    startedAt: number;
+    lastSeq: number;
+  }> {
+    return chatRunRegistry.listRunningRuns();
+  },
+
   /**
    * Normalizes one provider-native event into frontend session message events.
    */
@@ -89,12 +112,43 @@ export const sessionsService = {
   },
 
   /**
-   * Fetches persisted history by session id.
+   * Allocates a stable app-facing session id before any provider run happens.
+   *
+   * This is the entry point of the session gateway: the frontend calls this
+   * (via `POST /api/providers/sessions`) when the user starts a brand-new
+   * chat, navigates to the returned id immediately, and the id never changes
+   * for the lifetime of the conversation. The provider-native id is mapped to
+   * this row later, when the provider runtime announces it mid-run.
+   */
+  createAppSession(provider: LLMProvider, projectPath: string): CreateAppSessionResult {
+    const normalizedProjectPath = projectPath.trim();
+    if (!normalizedProjectPath) {
+      throw new AppError('projectPath is required.', {
+        code: 'PROJECT_PATH_REQUIRED',
+        statusCode: 400,
+      });
+    }
+
+    const sessionId = randomUUID();
+    sessionsDb.createAppSession(sessionId, provider, normalizedProjectPath);
+
+    return {
+      sessionId,
+      provider,
+      projectPath: normalizedProjectPath,
+    };
+  },
+
+  /**
+   * Fetches persisted history by app session id.
    *
    * Provider and provider-specific lookup hints are resolved from the indexed
-   * session metadata in the database.
+   * session metadata in the database. The provider adapter receives the
+   * provider-native session id (the one written into transcripts on disk),
+   * and every returned message is remapped back to the app session id so
+   * provider ids never reach the frontend.
    */
-  fetchHistory(
+  async fetchHistory(
     sessionId: string,
     options: Pick<FetchHistoryOptions, 'limit' | 'offset'> = {},
   ): Promise<FetchHistoryResult> {
@@ -106,12 +160,33 @@ export const sessionsService = {
       });
     }
 
+    // App-created sessions that never produced a provider transcript yet
+    // (e.g. first message still streaming) simply have no history.
+    if (!session.provider_session_id) {
+      return {
+        messages: [],
+        total: 0,
+        hasMore: false,
+        offset: options.offset ?? 0,
+        limit: options.limit ?? null,
+      };
+    }
+
     const provider = session.provider as LLMProvider;
-    return providerRegistry.resolveProvider(provider).sessions.fetchHistory(sessionId, {
+    const result = await providerRegistry.resolveProvider(provider).sessions.fetchHistory(sessionId, {
       limit: options.limit ?? null,
       offset: options.offset ?? 0,
       projectPath: session.project_path ?? '',
+      providerSessionId: session.provider_session_id,
     });
+
+    return {
+      ...result,
+      messages: result.messages.map((message) => ({
+        ...message,
+        sessionId,
+      })),
+    };
   },
 
   /**

server/modules/providers/tests/provider-models.service.test.ts

@@ -130,6 +130,37 @@ test('provider models are cached for the three-day ttl', async () => {
   }
 });
 
+test('claude provider models are always loaded directly from the provider', async () => {
+  const tempRoot = await mkdtemp(path.join(os.tmpdir(), 'provider-model-cache-claude-direct-'));
+  let loadCount = 0;
+
+  try {
+    const service = createProviderModelsService({
+      cachePath: path.join(tempRoot, 'models-cache.json'),
+      resolveProvider: (provider) => ({
+        models: {
+          getSupportedModels: async () => {
+            loadCount += 1;
+            return createModels(`${provider}-${loadCount}`);
+          },
+          getCurrentActiveModel: async () => createCurrentActiveModel(`${provider}-active`),
+          changeActiveModel: async (input) => createSessionActiveModelChange(provider, input),
+        },
+      }),
+    });
+
+    const first = await service.getProviderModels('claude');
+    const second = await service.getProviderModels('claude');
+
+    assert.equal(loadCount, 2);
+    assert.equal(first.models.DEFAULT, 'claude-1');
+    assert.equal(second.models.DEFAULT, 'claude-2');
+    assert.equal(second.cache.source, 'fresh');
+  } finally {
+    await rm(tempRoot, { recursive: true, force: true });
+  }
+});
+
 test('provider model cache is persisted across service instances', async () => {
   const tempRoot = await mkdtemp(path.join(os.tmpdir(), 'provider-model-cache-file-'));
   const cachePath = path.join(tempRoot, 'models-cache.json');

server/modules/websocket/README.md

@@ -33,10 +33,12 @@ Benefits:
 |---|---|
 | `services/websocket-server.service.ts` | Creates `WebSocketServer`, binds `verifyClient`, routes connection by pathname |
 | `services/websocket-auth.service.ts` | Authenticates upgrade requests and attaches `request.user` |
-| `services/chat-websocket.service.ts` | Handles `/ws` chat protocol and provider command/session control messages |
+| `services/chat-websocket.service.ts` | Handles the `/ws` chat protocol (`chat.send` / `chat.abort` / `chat.subscribe` / `chat.permission-response`) |
+| `services/chat-run-registry.service.ts` | Tracks live provider runs per app session id: seq numbering, event replay buffer, provider-id mapping, completion state |
+| `services/chat-session-writer.service.ts` | Gateway writer handed to provider runtimes: remaps provider session ids to app ids, swallows `session_created`, assigns `seq` |
 | `services/shell-websocket.service.ts` | Handles `/shell` PTY lifecycle, reconnect buffering, auth URL detection |
 | `services/plugin-websocket-proxy.service.ts` | Bridges client socket to plugin socket |
-| `services/websocket-writer.service.ts` | Adapts raw WebSocket to writer interface (`send`, `setSessionId`, `getSessionId`) |
+| `services/websocket-writer.service.ts` | Adapts raw WebSocket to writer interface (`send`, `setSessionId`, `getSessionId`) for non-chat writer consumers |
 | `services/websocket-state.service.ts` | Holds shared chat client set and open-state constant |
 
 ## High-Level Architecture
@@ -52,12 +54,12 @@ flowchart LR
   D -->|other| H[close()]
 
   E --> I[connectedClients Set]
-  E --> J[WebSocketWriter]
+  E --> J[chatRunRegistry + ChatSessionWriter]
   F --> K[ptySessionsMap]
   G --> L[Upstream Plugin ws://127.0.0.1:port/ws]
 
-  I --> M[projects.service broadcastProgress]
-  I --> N[sessions-watcher.service projects_updated]
+  I --> M[projects.service loading_progress]
+  I --> N[sessions-watcher.service session_upserted]
 ```
 
 ## Connection Handshake + Routing
@@ -105,37 +107,41 @@ sequenceDiagram
 When a chat socket connects:
 
 1. Add socket to `connectedClients`.
-2. Build `WebSocketWriter` (captures `userId` from authenticated request).
-3. Parse each incoming message with `parseIncomingJsonObject`.
-4. Dispatch by `data.type`.
-5. On close, remove socket from `connectedClients`.
+2. Parse each incoming message with `parseIncomingJsonObject`.
+3. Dispatch by `data.type` (four message types, none provider-specific).
+4. On close, remove socket from `connectedClients`.
+
+### Session identity model
+
+The frontend only ever knows the **app session id** (allocated by
+`POST /api/providers/sessions` or discovered via the session index). The
+provider-native id (JSONL file name, CLI resume id) stays inside the backend:
+
+1. `chat.send` resolves the app id to `{ provider, provider_session_id, project_path }` from the sessions DB.
+2. The provider runtime receives the provider-native id for resume.
+3. The `ChatSessionWriter` remaps every outbound event back to the app id, and turns `session_created` announcements into a DB mapping update instead of forwarding them.
 
 ### Chat Message Dispatch
 
 ```mermaid
 flowchart TD
   A[Incoming WS message] --> B[parseIncomingJsonObject]
-  B -->|invalid| C[send {type:error}]
+  B -->|invalid| C[send kind:protocol_error]
   B -->|ok| D{data.type}
 
-  D -->|claude-command| E[queryClaudeSDK]
-  D -->|cursor-command| F[spawnCursor]
-  D -->|codex-command| G[queryCodex]
-  D -->|gemini-command| H[spawnGemini]
-  D -->|cursor-resume| I[spawnCursor resume]
-  D -->|abort-session| J[abort by provider]
-  D -->|claude-permission-response| K[resolveToolApproval]
-  D -->|cursor-abort| L[abortCursorSession]
-  D -->|check-session-status| M[is*SessionActive + optional reconnectSessionWriter]
-  D -->|get-pending-permissions| N[getPendingApprovalsForSession]
-  D -->|get-active-sessions| O[getActive*Sessions]
+  D -->|chat.send| E[resolve session row -> startRun -> spawnFns provider]
+  D -->|chat.abort| F[abortFns provider + synthetic complete]
+  D -->|chat.subscribe| G[chat_subscribed ack + attach socket + replay events seq > lastSeq]
+  D -->|chat.permission-response| H[resolveToolApproval]
+  D -->|other| I[send kind:protocol_error]
 ```
 
 ### Chat Notes
 
-1. `abort-session` returns a normalized `complete` message with `aborted: true`.
-2. `check-session-status` returns `{ type: "session-status", isProcessing }`.
-3. Claude status checks can reconnect output stream to the new socket via `reconnectSessionWriter`.
+1. **Unified envelope**: every server-to-client frame carries a `kind` — either a provider `NormalizedMessage` kind or a gateway kind (`chat_subscribed`, `session_upserted`, `loading_progress`, `protocol_error`). There is no second `type`-based protocol.
+2. **Unified terminal lifecycle**: every provider run ends with exactly one `complete` message built by `createCompleteMessage()` (`server/shared/utils.ts`): `{ kind: "complete", sessionId, actualSessionId, exitCode, success, aborted }`. The chat handler emits a synthetic `complete` for runs that crash or get aborted, and the run registry drops duplicate completes.
+3. **Per-run event log**: every live event gets a monotonically increasing `seq`. `chat.subscribe { sessions: [{ sessionId, lastSeq }] }` re-attaches the live stream to the requesting socket (any provider, not just Claude) and replays events with `seq > lastSeq`. If the buffer no longer covers `lastSeq`, the client refreshes over REST.
+4. `chat_subscribed` includes `isProcessing` (replaces `check-session-status`) and `pendingPermissions` (replaces `get-pending-permissions`).
 
 ## `/shell` Terminal Flow
 
@@ -223,9 +229,9 @@ Only chat sockets (`/ws`) are tracked in `connectedClients`.
 That shared set is consumed by:
 
 1. `modules/projects/services/projects-with-sessions-fetch.service.ts`
-Broadcasts `loading_progress` while project snapshots are being built.
+Broadcasts `kind: loading_progress` while project snapshots are being built.
 2. `modules/providers/services/sessions-watcher.service.ts`
-Broadcasts `projects_updated` when provider session artifacts change.
+Broadcasts per-session `kind: session_upserted` deltas when provider session artifacts change (no full project snapshots).
 
 This design centralizes cross-module realtime fanout without requiring route-local references to WebSocket internals.
 
@@ -252,7 +258,7 @@ Current explicit close codes in this module:
 
 Other errors:
 
-1. Chat handler catches and emits `{ type: "error", error }`.
+1. Chat handler catches and emits `{ kind: "protocol_error", code, error }`.
 2. Shell handler catches and writes terminal-visible error output.
 3. Unknown websocket paths are closed immediately.
 

server/modules/websocket/index.ts

@@ -1,2 +1,3 @@
 export { WS_OPEN_STATE, connectedClients } from './services/websocket-state.service.js';
 export { createWebSocketServer } from './services/websocket-server.service.js';
+export { chatRunRegistry } from './services/chat-run-registry.service.js';

server/modules/websocket/services/chat-run-registry.service.ts

@@ -0,0 +1,273 @@
+import { sessionsDb } from '@/modules/database/index.js';
+import { ChatSessionWriter } from '@/modules/websocket/services/chat-session-writer.service.js';
+import type {
+  LLMProvider,
+  NormalizedMessage,
+  RealtimeClientConnection,
+} from '@/shared/types.js';
+
+type ChatRunStatus = 'running' | 'completed';
+
+/**
+ * One live (or recently finished) provider run for a single app session.
+ *
+ * State notes — why each mutable field is essential:
+ * - `providerSessionId`: the provider-native id captured mid-run. The abort
+ *   handler needs it to address the provider runtime, and the DB mapping is
+ *   written from it so history/resume work after the run.
+ * - `status`: drives `chat_subscribed.isProcessing`, prevents double sends
+ *   into the same session, and guards the synthetic-complete fallback in the
+ *   chat handler (only emitted when a runtime died without completing).
+ * - `lastSeq` / `events`: the per-run event log. Every live event gets a
+ *   monotonically increasing `seq` and is buffered so a reconnecting client
+ *   can replay exactly the events it missed via `chat.subscribe`.
+ */
+type ChatRun = {
+  appSessionId: string;
+  provider: LLMProvider;
+  providerSessionId: string | null;
+  status: ChatRunStatus;
+  lastSeq: number;
+  events: NormalizedMessage[];
+  writer: ChatSessionWriter;
+  startedAt: number;
+  completedAt: number | null;
+};
+
+/**
+ * How long a completed run stays available for replay. Covers the window
+ * between a run finishing and the client refreshing history over REST (for
+ * example when the browser tab was asleep while the run completed).
+ */
+const COMPLETED_RUN_RETENTION_MS = 5 * 60 * 1000;
+
+/**
+ * Upper bound on buffered events per run so a very long tool-heavy run cannot
+ * grow memory unbounded. When exceeded, the oldest events are dropped —
+ * a reconnecting client whose `lastSeq` predates the buffer falls back to a
+ * REST history refresh, which is always the authoritative source.
+ */
+const MAX_BUFFERED_EVENTS_PER_RUN = 5000;
+
+/**
+ * Active and recently-completed runs keyed by app session id.
+ *
+ * This map is the single in-memory source of truth for "is something running
+ * for this session" — the chat websocket handler, abort path, and subscribe
+ * path all consult it instead of asking each provider runtime individually.
+ */
+const runs = new Map<string, ChatRun>();
+
+function evictRunLater(appSessionId: string): void {
+  const timer = setTimeout(() => {
+    const run = runs.get(appSessionId);
+    if (run && run.status === 'completed') {
+      runs.delete(appSessionId);
+    }
+  }, COMPLETED_RUN_RETENTION_MS);
+
+  // Never keep the process alive just to evict a buffered run.
+  timer.unref?.();
+}
+
+/**
+ * Decorates one outbound live event for a run and records it in the event log.
+ *
+ * Responsibilities:
+ * 1. Remap `sessionId` (and `actualSessionId` on `complete`) to the stable
+ *    app session id — provider-native ids never leave the backend.
+ * 2. Assign the next `seq` so clients can detect/replay gaps.
+ * 3. Buffer the event for `chat.subscribe` replay.
+ * 4. Flip the run to `completed` when the terminal `complete` event passes by.
+ */
+function decorateAndRecordEvent(run: ChatRun, message: NormalizedMessage): NormalizedMessage | null {
+  // Exactly-one-complete contract: when a run is aborted the chat handler
+  // emits the terminal `complete` immediately, but the killed runtime may
+  // still emit its own `complete` from its exit handler moments later.
+  // Whichever arrives first wins; the duplicate is dropped here.
+  if (message.kind === 'complete' && run.status === 'completed') {
+    return null;
+  }
+
+  run.lastSeq += 1;
+
+  const outbound: NormalizedMessage = {
+    ...message,
+    sessionId: run.appSessionId,
+    seq: run.lastSeq,
+  };
+
+  if (message.kind === 'complete') {
+    // The provider may report its own id here; the frontend only ever knows
+    // the app id, so the "actual" id is by definition the app id as well.
+    outbound.actualSessionId = run.appSessionId;
+    run.status = 'completed';
+    run.completedAt = Date.now();
+    evictRunLater(run.appSessionId);
+  }
+
+  run.events.push(outbound);
+  if (run.events.length > MAX_BUFFERED_EVENTS_PER_RUN) {
+    run.events.splice(0, run.events.length - MAX_BUFFERED_EVENTS_PER_RUN);
+  }
+
+  return outbound;
+}
+
+/**
+ * Records the provider-native session id for a run and persists the
+ * app-id-to-provider-id mapping so history fetches and future resumes can
+ * address the provider transcript.
+ *
+ * Called from the gateway writer when the runtime either calls
+ * `setSessionId(...)` or emits its `session_created` event — whichever
+ * happens first wins; later calls with the same id are no-ops.
+ */
+function recordProviderSessionId(run: ChatRun, providerSessionId: string): void {
+  if (!providerSessionId || run.providerSessionId === providerSessionId) {
+    return;
+  }
+
+  run.providerSessionId = providerSessionId;
+
+  try {
+    sessionsDb.assignProviderSessionId(run.appSessionId, providerSessionId);
+  } catch (error) {
+    const message = error instanceof Error ? error.message : String(error);
+    console.error('[ChatRunRegistry] Failed to persist provider session id mapping', {
+      appSessionId: run.appSessionId,
+      providerSessionId,
+      error: message,
+    });
+  }
+}
+
+/**
+ * Registry of live provider runs keyed by the stable app session id.
+ *
+ * The registry is what makes the websocket protocol provider-independent:
+ * every run gets a `ChatSessionWriter` that remaps provider-native session
+ * ids to the app id, assigns `seq` numbers, and buffers events for replay —
+ * regardless of which provider runtime produced them.
+ */
+export const chatRunRegistry = {
+  /**
+   * Starts tracking a run and returns it, or `null` when a run is already in
+   * progress for the session (callers must reject the duplicate send).
+   */
+  startRun(input: {
+    appSessionId: string;
+    provider: LLMProvider;
+    providerSessionId: string | null;
+    connection: RealtimeClientConnection;
+    userId: string | number | null;
+  }): ChatRun | null {
+    const existing = runs.get(input.appSessionId);
+    if (existing && existing.status === 'running') {
+      return null;
+    }
+
+    const run: ChatRun = {
+      appSessionId: input.appSessionId,
+      provider: input.provider,
+      providerSessionId: input.providerSessionId,
+      status: 'running',
+      lastSeq: 0,
+      events: [],
+      writer: null as unknown as ChatSessionWriter,
+      startedAt: Date.now(),
+      completedAt: null,
+    };
+
+    run.writer = new ChatSessionWriter({
+      connection: input.connection,
+      userId: input.userId,
+      provider: input.provider,
+      providerSessionId: input.providerSessionId,
+      onProviderSessionId: (providerSessionId) => {
+        recordProviderSessionId(run, providerSessionId);
+      },
+      decorateOutboundEvent: (message) => decorateAndRecordEvent(run, message),
+    });
+
+    runs.set(input.appSessionId, run);
+    return run;
+  },
+
+  getRun(appSessionId: string): ChatRun | undefined {
+    return runs.get(appSessionId);
+  },
+
+  isProcessing(appSessionId: string): boolean {
+    return runs.get(appSessionId)?.status === 'running';
+  },
+
+  listRunningRuns(): Array<{
+    sessionId: string;
+    provider: LLMProvider;
+    startedAt: number;
+    lastSeq: number;
+  }> {
+    return Array.from(runs.values())
+      .filter((run) => run.status === 'running')
+      .map((run) => ({
+        sessionId: run.appSessionId,
+        provider: run.provider,
+        startedAt: run.startedAt,
+        lastSeq: run.lastSeq,
+      }));
+  },
+
+  /**
+   * Re-attaches a run's outbound stream to a (new) websocket connection.
+   *
+   * This is the generic replacement for the Claude-only writer reconnect:
+   * after a page refresh the new socket subscribes and immediately starts
+   * receiving the still-running stream, for every provider.
+   */
+  attachConnection(appSessionId: string, connection: RealtimeClientConnection): boolean {
+    const run = runs.get(appSessionId);
+    if (!run) {
+      return false;
+    }
+
+    run.writer.updateWebSocket(connection);
+    return true;
+  },
+
+  /**
+   * Returns buffered events with `seq` greater than `afterSeq` for replay.
+   *
+   * An empty array with `run.lastSeq > afterSeq` not covered by the buffer
+   * means the buffer was truncated; the client should refresh over REST.
+   */
+  replayEvents(appSessionId: string, afterSeq: number): NormalizedMessage[] {
+    const run = runs.get(appSessionId);
+    if (!run) {
+      return [];
+    }
+
+    return run.events.filter((event) => typeof event.seq === 'number' && event.seq > afterSeq);
+  },
+
+  /**
+   * Emits a synthetic terminal `complete` if (and only if) the run is still
+   * marked running. Used when a provider runtime throws or resolves without
+   * having produced its own terminal event, and by the abort path.
+   */
+  completeRun(appSessionId: string, opts: { exitCode: number; aborted?: boolean }): void {
+    const run = runs.get(appSessionId);
+    if (!run || run.status !== 'running') {
+      return;
+    }
+
+    run.writer.sendComplete(opts);
+  },
+
+  /**
+   * Test-only escape hatch: clears every tracked run.
+   */
+  clearAll(): void {
+    runs.clear();
+  },
+};

server/modules/websocket/services/chat-session-writer.service.ts

@@ -0,0 +1,145 @@
+import { WS_OPEN_STATE } from '@/modules/websocket/services/websocket-state.service.js';
+import type {
+  LLMProvider,
+  NormalizedMessage,
+  RealtimeClientConnection,
+} from '@/shared/types.js';
+import { createCompleteMessage, readObjectRecord } from '@/shared/utils.js';
+
+type ChatSessionWriterOptions = {
+  connection: RealtimeClientConnection;
+  userId: string | number | null;
+  provider: LLMProvider;
+  /** Provider-native id when resuming an existing session, otherwise null. */
+  providerSessionId: string | null;
+  /**
+   * Invoked the moment the provider runtime reveals its native session id
+   * (either via `setSessionId` or a `session_created` event). The registry
+   * persists the app-id-to-provider-id mapping from this callback.
+   */
+  onProviderSessionId: (providerSessionId: string) => void;
+  /**
+   * Remaps/sequences/buffers one outbound live event. Implemented by the chat
+   * run registry; the writer never forwards a provider event untouched.
+   * Returns `null` when the event must be dropped (duplicate terminal
+   * `complete` after an abort already completed the run).
+   */
+  decorateOutboundEvent: (message: NormalizedMessage) => NormalizedMessage | null;
+};
+
+/**
+ * Gateway writer handed to provider runtimes instead of a raw websocket writer.
+ *
+ * It exposes the exact same surface as `WebSocketWriter` (`send`,
+ * `setSessionId`, `getSessionId`, `updateWebSocket`, `userId`,
+ * `isWebSocketWriter`) so the provider runtimes (`claude-sdk.js`,
+ * `cursor-cli.js`, ...) need zero changes — but everything that flows through
+ * it is translated from the provider's world into the app's protocol:
+ *
+ * - `session_created` events are swallowed and turned into a provider-id
+ *   mapping; the frontend never learns provider-native ids.
+ * - every other event gets `sessionId` remapped to the app session id and a
+ *   per-run `seq` assigned before being forwarded.
+ * - `setSessionId(...)` calls (used by runtimes to label captured ids) are
+ *   intercepted and recorded as the provider-id mapping as well.
+ */
+export class ChatSessionWriter {
+  ws: RealtimeClientConnection;
+  userId: string | number | null;
+  /**
+   * Some runtimes feature-detect their writer with this flag; keep it so the
+   * gateway writer is a drop-in replacement for `WebSocketWriter`.
+   */
+  isWebSocketWriter = true;
+
+  private readonly options: ChatSessionWriterOptions;
+  /**
+   * The provider-native session id as the runtime knows it. Kept locally
+   * (besides the registry) because runtimes read it back via `getSessionId()`
+   * to label their own outgoing events — those labels are remapped on send
+   * anyway, but the runtime-visible value must stay provider-native.
+   */
+  private providerSessionId: string | null;
+
+  constructor(options: ChatSessionWriterOptions) {
+    this.options = options;
+    this.ws = options.connection;
+    this.userId = options.userId;
+    this.providerSessionId = options.providerSessionId;
+  }
+
+  send(data: unknown): void {
+    const record = readObjectRecord(data);
+    if (!record || typeof record.kind !== 'string') {
+      // Provider runtimes only emit kind-based normalized messages. Anything
+      // else indicates a programming error; drop it rather than leaking an
+      // un-remapped payload to the client.
+      console.error('[ChatSessionWriter] Dropping non-normalized outbound payload', data);
+      return;
+    }
+
+    const message = record as NormalizedMessage;
+
+    if (message.kind === 'session_created') {
+      const announcedId =
+        typeof message.newSessionId === 'string' && message.newSessionId
+          ? message.newSessionId
+          : message.sessionId;
+      if (announcedId) {
+        this.captureProviderSessionId(announcedId);
+      }
+      // Swallowed on purpose: the frontend already has the stable app session
+      // id, so there is no client-side handoff to perform anymore.
+      return;
+    }
+
+    const outbound = this.options.decorateOutboundEvent(message);
+    if (outbound) {
+      this.forward(outbound);
+    }
+  }
+
+  /**
+   * Emits the synthetic terminal `complete` for runs that ended without one
+   * (runtime crash before completing, or user abort).
+   */
+  sendComplete(opts: { exitCode: number; aborted?: boolean }): void {
+    const message = createCompleteMessage({
+      provider: this.options.provider,
+      sessionId: this.providerSessionId,
+      exitCode: opts.exitCode,
+      aborted: opts.aborted,
+    });
+    const outbound = this.options.decorateOutboundEvent(message);
+    if (outbound) {
+      this.forward(outbound);
+    }
+  }
+
+  updateWebSocket(newConnection: RealtimeClientConnection): void {
+    this.ws = newConnection;
+  }
+
+  setSessionId(sessionId: string): void {
+    this.captureProviderSessionId(sessionId);
+  }
+
+  getSessionId(): string | null {
+    return this.providerSessionId;
+  }
+
+  private captureProviderSessionId(providerSessionId: string): void {
+    if (!providerSessionId || this.providerSessionId === providerSessionId) {
+      return;
+    }
+
+    this.providerSessionId = providerSessionId;
+    this.options.onProviderSessionId(providerSessionId);
+  }
+
+  private forward(message: NormalizedMessage): void {
+    if (this.ws.readyState === WS_OPEN_STATE) {
+      this.ws.send(JSON.stringify(message));
+    }
+  }
+}

server/modules/websocket/services/chat-websocket.service.ts

@@ -1,40 +1,35 @@
 import type { WebSocket } from 'ws';
 
-import { connectedClients } from '@/modules/websocket/services/websocket-state.service.js';
-import { WebSocketWriter } from '@/modules/websocket/services/websocket-writer.service.js';
+import { sessionsDb } from '@/modules/database/index.js';
+import { chatRunRegistry } from '@/modules/websocket/services/chat-run-registry.service.js';
+import { connectedClients, WS_OPEN_STATE } from '@/modules/websocket/services/websocket-state.service.js';
 import type {
   AnyRecord,
   AuthenticatedWebSocketRequest,
   LLMProvider,
 } from '@/shared/types.js';
-import { createNormalizedMessage, parseIncomingJsonObject } from '@/shared/utils.js';
+import { parseIncomingJsonObject } from '@/shared/utils.js';
 
-type ChatIncomingMessage = AnyRecord & {
-  type?: string;
-  command?: string;
-  options?: AnyRecord;
-  provider?: string;
-  sessionId?: string;
-  requestId?: string;
-  allow?: unknown;
-  updatedInput?: unknown;
-  message?: unknown;
-  rememberEntry?: unknown;
-};
-
-const DEFAULT_PROVIDER: LLMProvider = 'claude';
+/**
+ * One provider runtime entry point. All five runtimes share this signature,
+ * which lets the chat handler dispatch through a provider-keyed map instead
+ * of provider-specific branches.
+ */
+type ProviderSpawnFn = (
+  command: string,
+  options: AnyRecord,
+  writer: unknown
+) => Promise<unknown>;
 
 type ChatWebSocketDependencies = {
-  queryClaudeSDK: (command: string, options: unknown, writer: WebSocketWriter) => Promise<unknown>;
-  spawnCursor: (command: string, options: unknown, writer: WebSocketWriter) => Promise<unknown>;
-  queryCodex: (command: string, options: unknown, writer: WebSocketWriter) => Promise<unknown>;
-  spawnGemini: (command: string, options: unknown, writer: WebSocketWriter) => Promise<unknown>;
-  spawnOpenCode: (command: string, options: unknown, writer: WebSocketWriter) => Promise<unknown>;
-  abortClaudeSDKSession: (sessionId: string) => Promise<boolean>;
-  abortCursorSession: (sessionId: string) => boolean;
-  abortCodexSession: (sessionId: string) => boolean;
-  abortGeminiSession: (sessionId: string) => boolean;
-  abortOpenCodeSession: (sessionId: string) => boolean;
+  /** Provider runtimes keyed by provider id. */
+  spawnFns: Record<LLMProvider, ProviderSpawnFn>;
+  /**
+   * Abort functions keyed by provider id. They are addressed with the
+   * provider-native session id (that is how runtimes key their process maps).
+   * The Claude abort is async; the rest are sync — both shapes are accepted.
+   */
+  abortFns: Record<LLMProvider, (providerSessionId: string) => boolean | Promise<boolean>>;
   resolveToolApproval: (
     requestId: string,
     payload: {
@@ -44,31 +39,10 @@ type ChatWebSocketDependencies = {
       rememberEntry?: unknown;
     }
   ) => void;
-  isClaudeSDKSessionActive: (sessionId: string) => boolean;
-  isCursorSessionActive: (sessionId: string) => boolean;
-  isCodexSessionActive: (sessionId: string) => boolean;
-  isGeminiSessionActive: (sessionId: string) => boolean;
-  isOpenCodeSessionActive: (sessionId: string) => boolean;
-  reconnectSessionWriter: (sessionId: string, ws: WebSocket) => boolean;
-  getPendingApprovalsForSession: (sessionId: string) => unknown[];
-  getActiveClaudeSDKSessions: () => unknown;
-  getActiveCursorSessions: () => unknown;
-  getActiveCodexSessions: () => unknown;
-  getActiveGeminiSessions: () => unknown;
-  getActiveOpenCodeSessions: () => unknown;
+  /** Claude-only today: pending tool approvals included in `chat_subscribed`. */
+  getPendingApprovalsForSession: (providerSessionId: string) => unknown[];
 };
 
-/**
- * Normalizes potentially invalid provider names coming from websocket payloads.
- */
-function readProvider(value: unknown): LLMProvider {
-  if (value === 'claude' || value === 'cursor' || value === 'codex' || value === 'gemini' || value === 'opencode') {
-    return value;
-  }
-
-  return DEFAULT_PROVIDER;
-}
-
 /**
  * Extracts the authenticated request user id in the formats currently produced
  * by platform and OSS auth code paths.
@@ -92,8 +66,258 @@ function readRequestUserId(
   return null;
 }
 
+function sendJson(ws: WebSocket, payload: unknown): void {
+  if (ws.readyState === WS_OPEN_STATE) {
+    ws.send(JSON.stringify(payload));
+  }
+}
+
+/**
+ * Reports a protocol-level failure to the requesting client.
+ *
+ * Protocol errors deliberately use their own `kind` (instead of the provider
+ * `error` message kind) so the frontend can distinguish "your request was
+ * invalid" from "the model run produced an error" without inspecting text.
+ */
+function sendProtocolError(
+  ws: WebSocket,
+  code: string,
+  error: string,
+  sessionId?: string
+): void {
+  sendJson(ws, {
+    kind: 'protocol_error',
+    code,
+    error,
+    sessionId: sessionId ?? null,
+    timestamp: new Date().toISOString(),
+  });
+}
+
+function readRequiredSessionId(data: AnyRecord): string | null {
+  const sessionId = typeof data.sessionId === 'string' ? data.sessionId.trim() : '';
+  return sessionId.length > 0 ? sessionId : null;
+}
+
+/**
+ * Handles `chat.send`: resolves the session row (provider, project path, and
+ * provider-native id all come from the database — never from the client),
+ * registers the run, and dispatches to the provider runtime.
+ */
+async function handleChatSend(
+  ws: WebSocket,
+  userId: string | number | null,
+  data: AnyRecord,
+  dependencies: ChatWebSocketDependencies
+): Promise<void> {
+  const sessionId = readRequiredSessionId(data);
+  if (!sessionId) {
+    sendProtocolError(ws, 'SESSION_ID_REQUIRED', 'chat.send requires a sessionId.');
+    return;
+  }
+
+  const session = sessionsDb.getSessionById(sessionId);
+  if (!session) {
+    sendProtocolError(
+      ws,
+      'SESSION_NOT_FOUND',
+      `Session "${sessionId}" was not found. Create it via POST /api/providers/sessions first.`,
+      sessionId
+    );
+    return;
+  }
+
+  const provider = session.provider as LLMProvider;
+  const spawnFn = dependencies.spawnFns[provider];
+  if (!spawnFn) {
+    sendProtocolError(ws, 'UNSUPPORTED_PROVIDER', `Provider "${provider}" is not available.`, sessionId);
+    return;
+  }
+
+  const run = chatRunRegistry.startRun({
+    appSessionId: sessionId,
+    provider,
+    providerSessionId: session.provider_session_id,
+    connection: ws,
+    userId,
+  });
+
+  if (!run) {
+    sendProtocolError(
+      ws,
+      'RUN_IN_PROGRESS',
+      `Session "${sessionId}" already has a run in progress.`,
+      sessionId
+    );
+    return;
+  }
+
+  const clientOptions = (data.options ?? {}) as AnyRecord;
+  const command = typeof data.content === 'string' ? data.content : '';
+
+  // The provider runtimes receive the provider-native session id (that is the
+  // id their CLI/SDK understands for resume). Brand-new sessions have no
+  // provider id yet, so the runtime starts fresh and announces one, which the
+  // gateway writer captures and maps back to the app session id.
+  const runtimeOptions: AnyRecord = {
+    ...clientOptions,
+    sessionId: session.provider_session_id ?? undefined,
+    resume: Boolean(session.provider_session_id),
+    cwd: clientOptions.cwd ?? session.project_path ?? undefined,
+    projectPath: session.project_path ?? clientOptions.projectPath,
+  };
+
+  try {
+    await spawnFn(command, runtimeOptions, run.writer);
+  } catch (error) {
+    const message = error instanceof Error ? error.message : String(error);
+    console.error(`[Chat] Provider runtime "${provider}" failed`, { sessionId, error: message });
+  } finally {
+    // Safety net: a runtime that crashed (or resolved) without emitting its
+    // terminal `complete` would otherwise leave the session stuck in
+    // "processing" forever on every connected client.
+    chatRunRegistry.completeRun(sessionId, { exitCode: 1 });
+  }
+}
+
+/**
+ * Handles `chat.abort`: cancels the run for one app session and emits the
+ * terminal `complete` on its behalf (runtimes skip their own complete for
+ * aborted runs, and the registry drops any duplicate).
+ */
+async function handleChatAbort(
+  ws: WebSocket,
+  data: AnyRecord,
+  dependencies: ChatWebSocketDependencies
+): Promise<void> {
+  const sessionId = readRequiredSessionId(data);
+  if (!sessionId) {
+    sendProtocolError(ws, 'SESSION_ID_REQUIRED', 'chat.abort requires a sessionId.');
+    return;
+  }
+
+  const run = chatRunRegistry.getRun(sessionId);
+  if (!run || run.status !== 'running') {
+    sendProtocolError(ws, 'NO_ACTIVE_RUN', `Session "${sessionId}" has no active run.`, sessionId);
+    return;
+  }
+
+  const abortFn = dependencies.abortFns[run.provider];
+  let success = false;
+  if (abortFn && run.providerSessionId) {
+    success = Boolean(await abortFn(run.providerSessionId));
+  }
+
+  chatRunRegistry.completeRun(sessionId, {
+    exitCode: success ? 0 : 1,
+    aborted: true,
+  });
+}
+
+/**
+ * Handles `chat.subscribe`: for each requested session, reports whether a run
+ * is processing, re-attaches the live stream to this socket, replays missed
+ * events (seq > lastSeq), and includes pending permission requests.
+ *
+ * This single message replaces the old `check-session-status`,
+ * `get-pending-permissions`, and Claude-only writer reconnect flows.
+ */
+function handleChatSubscribe(
+  ws: WebSocket,
+  data: AnyRecord,
+  dependencies: ChatWebSocketDependencies
+): void {
+  const targets = Array.isArray(data.sessions) ? data.sessions : [];
+
+  for (const target of targets) {
+    if (!target || typeof target !== 'object') {
+      continue;
+    }
+
+    const sessionId = typeof (target as AnyRecord).sessionId === 'string'
+      ? ((target as AnyRecord).sessionId as string).trim()
+      : '';
+    if (!sessionId) {
+      continue;
+    }
+
+    const lastSeqRaw = (target as AnyRecord).lastSeq;
+    const lastSeq = typeof lastSeqRaw === 'number' && Number.isFinite(lastSeqRaw)
+      ? Math.max(0, Math.floor(lastSeqRaw))
+      : 0;
+
+    const run = chatRunRegistry.getRun(sessionId);
+    const isProcessing = chatRunRegistry.isProcessing(sessionId);
+
+    // Future live events for this run should land on the socket that asked —
+    // this is what makes mid-stream page refreshes work for all providers.
+    if (isProcessing) {
+      chatRunRegistry.attachConnection(sessionId, ws);
+    }
+
+    // Pending approvals are tracked under the provider-native id inside the
+    // Claude runtime; remap their sessionId so the client only sees app ids.
+    const pendingPermissions = (run?.providerSessionId
+      ? dependencies.getPendingApprovalsForSession(run.providerSessionId)
+      : []
+    ).map((approval) =>
+      approval && typeof approval === 'object'
+        ? { ...(approval as AnyRecord), sessionId }
+        : approval,
+    );
+
+    sendJson(ws, {
+      kind: 'chat_subscribed',
+      sessionId,
+      isProcessing,
+      lastSeq: run?.lastSeq ?? 0,
+      pendingPermissions,
+      timestamp: new Date().toISOString(),
+    });
+
+    // Replay only for RUNNING runs, strictly after the ack. Completed runs
+    // are fully persisted to the provider transcript and served over REST —
+    // replaying them (e.g. after a page reload where the client's lastSeq is
+    // 0) would duplicate messages the history fetch already returned.
+    if (isProcessing) {
+      for (const event of chatRunRegistry.replayEvents(sessionId, lastSeq)) {
+        sendJson(ws, event);
+      }
+    }
+  }
+}
+
+/**
+ * Handles `chat.permission-response`: forwards a tool-approval decision to the
+ * pending approval resolver (Claude is the only provider with interactive
+ * approvals today, but the message is intentionally provider-neutral).
+ */
+function handlePermissionResponse(data: AnyRecord, dependencies: ChatWebSocketDependencies): void {
+  if (typeof data.requestId !== 'string' || data.requestId.length === 0) {
+    return;
+  }
+
+  dependencies.resolveToolApproval(data.requestId, {
+    allow: Boolean(data.allow),
+    updatedInput: data.updatedInput,
+    message: typeof data.message === 'string' ? data.message : undefined,
+    rememberEntry: data.rememberEntry,
+  });
+}
+
 /**
  * Handles authenticated chat websocket messages used by the main chat panel.
+ *
+ * Inbound protocol (client to server):
+ * - `chat.send`                { sessionId, content, options? }
+ * - `chat.abort`               { sessionId }
+ * - `chat.subscribe`           { sessions: [{ sessionId, lastSeq? }] }
+ * - `chat.permission-response` { requestId, allow, updatedInput?, message?, rememberEntry? }
+ *
+ * Outbound protocol (server to client): every frame is `kind`-based — either
+ * a provider `NormalizedMessage` (with `seq`) or a gateway event
+ * (`chat_subscribed`, `session_upserted`, `loading_progress`,
+ * `protocol_error`).
  */
 export function handleChatConnection(
   ws: WebSocket,
@@ -103,7 +327,7 @@ export function handleChatConnection(
   console.log('[INFO] Chat WebSocket connected');
   connectedClients.add(ws);
 
-  const writer = new WebSocketWriter(ws, readRequestUserId(request));
+  const userId = readRequestUserId(request);
 
   ws.on('message', async (rawMessage) => {
     try {
@@ -112,169 +336,30 @@ export function handleChatConnection(
         throw new Error('Invalid websocket payload');
       }
 
-      const data = parsed as ChatIncomingMessage;
-      const messageType = data.type;
-      if (!messageType) {
-        throw new Error('Message type is required');
-      }
+      const data = parsed as AnyRecord;
+      const messageType = typeof data.type === 'string' ? data.type : '';
 
-      if (messageType === 'claude-command') {
-        await dependencies.queryClaudeSDK(data.command ?? '', data.options, writer);
-        return;
-      }
-
-      if (messageType === 'cursor-command') {
-        await dependencies.spawnCursor(data.command ?? '', data.options, writer);
-        return;
-      }
-
-      if (messageType === 'codex-command') {
-        await dependencies.queryCodex(data.command ?? '', data.options, writer);
-        return;
-      }
-
-      if (messageType === 'gemini-command') {
-        await dependencies.spawnGemini(data.command ?? '', data.options, writer);
-        return;
-      }
-
-      if (messageType === 'opencode-command') {
-        await dependencies.spawnOpenCode(data.command ?? '', data.options, writer);
-        return;
-      }
-
-      if (messageType === 'cursor-resume') {
-        await dependencies.spawnCursor(
-          '',
-          {
-            sessionId: data.sessionId,
-            resume: true,
-            cwd: data.options?.cwd,
-          },
-          writer
-        );
-        return;
-      }
-
-      if (messageType === 'abort-session') {
-        const provider = readProvider(data.provider);
-        const sessionId = typeof data.sessionId === 'string' ? data.sessionId : '';
-        let success = false;
-
-        if (provider === 'cursor') {
-          success = dependencies.abortCursorSession(sessionId);
-        } else if (provider === 'codex') {
-          success = dependencies.abortCodexSession(sessionId);
-        } else if (provider === 'gemini') {
-          success = dependencies.abortGeminiSession(sessionId);
-        } else if (provider === 'opencode') {
-          success = dependencies.abortOpenCodeSession(sessionId);
-        } else {
-          success = await dependencies.abortClaudeSDKSession(sessionId);
-        }
-
-        writer.send(
-          createNormalizedMessage({
-            kind: 'complete',
-            exitCode: success ? 0 : 1,
-            aborted: true,
-            success,
-            sessionId,
-            provider,
-          })
-        );
-        return;
-      }
-
-      if (messageType === 'claude-permission-response') {
-        if (typeof data.requestId === 'string' && data.requestId.length > 0) {
-          dependencies.resolveToolApproval(data.requestId, {
-            allow: Boolean(data.allow),
-            updatedInput: data.updatedInput,
-            message: typeof data.message === 'string' ? data.message : undefined,
-            rememberEntry: data.rememberEntry,
-          });
-        }
-        return;
-      }
-
-      if (messageType === 'cursor-abort') {
-        const sessionId = typeof data.sessionId === 'string' ? data.sessionId : '';
-        const success = dependencies.abortCursorSession(sessionId);
-        writer.send(
-          createNormalizedMessage({
-            kind: 'complete',
-            exitCode: success ? 0 : 1,
-            aborted: true,
-            success,
-            sessionId,
-            provider: 'cursor',
-          })
-        );
-        return;
-      }
-
-      if (messageType === 'check-session-status') {
-        const provider = readProvider(data.provider);
-        const sessionId = typeof data.sessionId === 'string' ? data.sessionId : '';
-        let isActive = false;
-
-        if (provider === 'cursor') {
-          isActive = dependencies.isCursorSessionActive(sessionId);
-        } else if (provider === 'codex') {
-          isActive = dependencies.isCodexSessionActive(sessionId);
-        } else if (provider === 'gemini') {
-          isActive = dependencies.isGeminiSessionActive(sessionId);
-        } else if (provider === 'opencode') {
-          isActive = dependencies.isOpenCodeSessionActive(sessionId);
-        } else {
-          isActive = dependencies.isClaudeSDKSessionActive(sessionId);
-          if (isActive) {
-            dependencies.reconnectSessionWriter(sessionId, ws);
-          }
-        }
-
-        writer.send({
-          type: 'session-status',
-          sessionId,
-          provider,
-          isProcessing: isActive,
-        });
-        return;
-      }
-
-      if (messageType === 'get-pending-permissions') {
-        const sessionId = typeof data.sessionId === 'string' ? data.sessionId : '';
-        if (sessionId && dependencies.isClaudeSDKSessionActive(sessionId)) {
-          const pending = dependencies.getPendingApprovalsForSession(sessionId);
-          writer.send({
-            type: 'pending-permissions-response',
-            sessionId,
-            data: pending,
-          });
-        }
-        return;
-      }
-
-      if (messageType === 'get-active-sessions') {
-        writer.send({
-          type: 'active-sessions',
-          sessions: {
-            claude: dependencies.getActiveClaudeSDKSessions(),
-            cursor: dependencies.getActiveCursorSessions(),
-            codex: dependencies.getActiveCodexSessions(),
-            gemini: dependencies.getActiveGeminiSessions(),
-            opencode: dependencies.getActiveOpenCodeSessions(),
-          },
-        });
+      switch (messageType) {
+        case 'chat.send':
+          await handleChatSend(ws, userId, data, dependencies);
+          return;
+        case 'chat.abort':
+          await handleChatAbort(ws, data, dependencies);
+          return;
+        case 'chat.subscribe':
+          handleChatSubscribe(ws, data, dependencies);
+          return;
+        case 'chat.permission-response':
+          handlePermissionResponse(data, dependencies);
+          return;
+        default:
+          sendProtocolError(ws, 'UNKNOWN_MESSAGE_TYPE', `Unknown message type "${messageType}".`);
+          return;
       }
     } catch (error) {
       const message = error instanceof Error ? error.message : String(error);
       console.error('[ERROR] Chat WebSocket error:', message);
-      writer.send({
-        type: 'error',
-        error: message,
-      });
+      sendProtocolError(ws, 'INTERNAL_ERROR', message);
     }
   });
 

server/modules/websocket/services/shell-websocket.service.ts

@@ -18,6 +18,7 @@ type ShellIncomingMessage = {
   provider?: string;
   initialCommand?: string;
   isPlainShell?: boolean;
+  forceRestart?: boolean;
 };
 
 type PtySessionEntry = {
@@ -34,7 +35,10 @@ const PTY_SESSION_TIMEOUT = 30 * 60 * 1000;
 const SHELL_URL_PARSE_BUFFER_LIMIT = 32768;
 
 type ShellWebSocketDependencies = {
-  getSessionById: (sessionId: string) => { cliSessionId?: string } | null | undefined;
+  resolveProviderSessionId: (
+    sessionId: string,
+    provider: string,
+  ) => string | null | undefined;
   stripAnsiSequences: (content: string) => string;
   normalizeDetectedUrl: (url: string) => string | null;
   extractUrlsFromText: (content: string) => string[];
@@ -75,6 +79,36 @@ function parseShellMessage(rawMessage: RawData): ShellIncomingMessage | null {
   return payload as ShellIncomingMessage;
 }
 
+const SAFE_SESSION_ID_PATTERN = /^[a-zA-Z0-9_.\-:]+$/;
+
+function resolveResumeSessionId(
+  message: ShellIncomingMessage,
+  dependencies: ShellWebSocketDependencies
+): string {
+  const hasSession = readBoolean(message.hasSession);
+  const sessionId = readString(message.sessionId);
+  const provider = readString(message.provider, 'claude');
+
+  if (!hasSession || !sessionId) {
+    return '';
+  }
+
+  let resumeSessionId: string | null | undefined;
+  try {
+    resumeSessionId = dependencies.resolveProviderSessionId(sessionId, provider);
+  } catch (error) {
+    console.error('Failed to resolve provider session ID:', error);
+    resumeSessionId = undefined;
+  }
+
+  const resolvedSessionId = resumeSessionId === undefined ? sessionId : resumeSessionId;
+  if (!resolvedSessionId || !SAFE_SESSION_ID_PATTERN.test(resolvedSessionId)) {
+    return '';
+  }
+
+  return resolvedSessionId;
+}
+
 /**
  * Resolves provider command line for plain shell and agent-backed shell modes.
  */
@@ -83,10 +117,9 @@ function buildShellCommand(
   dependencies: ShellWebSocketDependencies
 ): string {
   const hasSession = readBoolean(message.hasSession);
-  const sessionId = readString(message.sessionId);
   const initialCommand = readString(message.initialCommand);
   const provider = readString(message.provider, 'claude');
-  const safeSessionIdPattern = /^[a-zA-Z0-9_.\-:]+$/;
+  const resumeSessionId = resolveResumeSessionId(message, dependencies);
   const isPlainShell =
     readBoolean(message.isPlainShell) ||
     (!!initialCommand && !hasSession) ||
@@ -97,58 +130,43 @@ function buildShellCommand(
   }
 
   if (provider === 'cursor') {
-    if (hasSession && sessionId) {
-      return `cursor-agent --resume="${sessionId}"`;
+    if (resumeSessionId) {
+      return `cursor-agent --resume="${resumeSessionId}"`;
     }
     return 'cursor-agent';
   }
 
   if (provider === 'codex') {
-    if (hasSession && sessionId) {
+    if (resumeSessionId) {
       if (os.platform() === 'win32') {
-        return `codex resume "${sessionId}"; if ($LASTEXITCODE -ne 0) { codex }`;
+        return `codex resume "${resumeSessionId}"; if ($LASTEXITCODE -ne 0) { codex }`;
       }
-      return `codex resume "${sessionId}" || codex`;
+      return `codex resume "${resumeSessionId}" || codex`;
     }
     return 'codex';
   }
 
   if (provider === 'gemini') {
     const command = initialCommand || 'gemini';
-    let resumeId = sessionId;
-    if (hasSession && sessionId) {
-      try {
-        const existingSession = dependencies.getSessionById(sessionId);
-        if (existingSession && existingSession.cliSessionId) {
-          resumeId = existingSession.cliSessionId;
-          if (!safeSessionIdPattern.test(resumeId)) {
-            resumeId = '';
-          }
-        }
-      } catch (error) {
-        console.error('Failed to get Gemini CLI session ID:', error);
-      }
-    }
-
-    if (hasSession && resumeId) {
-      return `${command} --resume "${resumeId}"`;
+    if (resumeSessionId) {
+      return `${command} --resume "${resumeSessionId}"`;
     }
     return command;
   }
 
   if (provider === 'opencode') {
-    if (hasSession && sessionId) {
-      return `opencode --session "${sessionId}"`;
+    if (resumeSessionId) {
+      return `opencode --session "${resumeSessionId}"`;
     }
     return initialCommand || 'opencode';
   }
 
   const command = initialCommand || 'claude';
-  if (hasSession && sessionId) {
+  if (resumeSessionId) {
     if (os.platform() === 'win32') {
-      return `claude --resume "${sessionId}"; if ($LASTEXITCODE -ne 0) { claude }`;
+      return `claude --resume "${resumeSessionId}"; if ($LASTEXITCODE -ne 0) { claude }`;
     }
-    return `claude --resume "${sessionId}" || claude`;
+    return `claude --resume "${resumeSessionId}" || claude`;
   }
   return command;
 }
@@ -180,6 +198,7 @@ export function handleShellConnection(
         const hasSession = readBoolean(data.hasSession);
         const provider = readString(data.provider, 'claude');
         const initialCommand = readString(data.initialCommand);
+        const forceRestart = readBoolean(data.forceRestart);
         const isPlainShell =
           readBoolean(data.isPlainShell) ||
           (!!initialCommand && !hasSession) ||
@@ -200,7 +219,7 @@ export function handleShellConnection(
             : '';
         ptySessionKey = `${projectPath}_${sessionId ?? 'default'}${commandSuffix}`;
 
-        if (isLoginCommand) {
+        if (isLoginCommand || forceRestart) {
           const oldSession = ptySessionsMap.get(ptySessionKey);
           if (oldSession) {
             if (oldSession.timeoutId) {
@@ -211,7 +230,8 @@ export function handleShellConnection(
           }
         }
 
-        const existingSession = isLoginCommand ? null : ptySessionsMap.get(ptySessionKey);
+        const existingSession =
+          isLoginCommand || forceRestart ? null : ptySessionsMap.get(ptySessionKey);
         if (existingSession) {
           shellProcess = existingSession.pty;
           if (existingSession.timeoutId) {
@@ -258,6 +278,7 @@ export function handleShellConnection(
         }
 
         const shellCommand = buildShellCommand(data, dependencies);
+        const resumeSessionId = resolveResumeSessionId(data, dependencies);
         const shell = os.platform() === 'win32' ? 'powershell.exe' : 'bash';
         const shellArgs =
           os.platform() === 'win32' ? ['-Command', shellCommand] : ['-c', shellCommand];
@@ -368,6 +389,10 @@ export function handleShellConnection(
           }
 
           const session = ptySessionsMap.get(ptySessionKey);
+          if (session && session.pty !== shellProcess) {
+            return;
+          }
+
           if (session && session.ws && session.ws.readyState === WebSocket.OPEN) {
             session.ws.send(
               JSON.stringify({
@@ -399,8 +424,8 @@ export function handleShellConnection(
                   : provider === 'opencode'
                     ? 'OpenCode'
                   : 'Claude';
-          welcomeMsg = hasSession
-            ? `\x1b[36mResuming ${providerName} session ${sessionId} in: ${projectPath}\x1b[0m\r\n`
+          welcomeMsg = hasSession && resumeSessionId
+            ? `\x1b[36mResuming ${providerName} session ${resumeSessionId} in: ${projectPath}\x1b[0m\r\n`
             : `\x1b[36mStarting new ${providerName} session in: ${projectPath}\x1b[0m\r\n`;
         }
 
@@ -451,6 +476,10 @@ export function handleShellConnection(
 
     session.ws = null;
     session.timeoutId = setTimeout(() => {
+      if (ptySessionsMap.get(ptySessionKey as string) !== session) {
+        return;
+      }
+
       session.pty.kill();
       ptySessionsMap.delete(ptySessionKey as string);
     }, PTY_SESSION_TIMEOUT);

server/modules/websocket/services/websocket-auth.service.ts

@@ -20,7 +20,13 @@ export function verifyWebSocketClient(
   dependencies: WebSocketAuthDependencies
 ): boolean {
   const request = info.req as AuthenticatedWebSocketRequest;
-  console.log('WebSocket connection attempt to:', request.url);
+  const upgradeUrl = new URL(request.url ?? '/', 'http://localhost');
+  const loggedUrl = new URL(upgradeUrl);
+  if (loggedUrl.searchParams.has('token')) {
+    loggedUrl.searchParams.set('token', 'REDACTED');
+  }
+
+  console.log('WebSocket connection attempt to:', `${loggedUrl.pathname}${loggedUrl.search}`);
 
   // Platform mode: use the first DB user and skip token checks.
   if (dependencies.isPlatform) {
@@ -36,7 +42,6 @@ export function verifyWebSocketClient(
   }
 
   // OSS mode: read JWT from query string first, then Authorization header.
-  const upgradeUrl = new URL(request.url ?? '/', 'http://localhost');
   const token =
     upgradeUrl.searchParams.get('token') ??
     request.headers.authorization?.split(' ')[1] ??

server/modules/websocket/services/websocket-server.service.ts

@@ -31,6 +31,24 @@ export function createWebSocketServer(
   });
 
   wss.on('connection', (ws, request) => {
+    // Keep WebSocket alive across reverse-proxy idle timeouts (Cloudflare ~100s,
+    // AWS ALB 60s, nginx 60s, etc.). Without app-level pings these connections
+    // are silently torn down even when the UI is active, causing repeated
+    // reconnect cycles. ws library heartbeat is opt-in.
+    const HEARTBEAT_INTERVAL_MS = 30_000;
+    const heartbeat = setInterval(() => {
+      if (ws.readyState === ws.OPEN) {
+        try {
+          ws.ping();
+        } catch {
+          // socket may have been closed concurrently — interval will be cleared below
+        }
+      }
+    }, HEARTBEAT_INTERVAL_MS);
+    const stopHeartbeat = () => clearInterval(heartbeat);
+    ws.on('close', stopHeartbeat);
+    ws.on('error', stopHeartbeat);
+
     const incomingRequest = request as AuthenticatedWebSocketRequest;
     const url = incomingRequest.url ?? '/';
     const pathname = new URL(url, 'http://localhost').pathname;

server/modules/websocket/tests/chat-run-registry.test.ts

@@ -0,0 +1,239 @@
+import assert from 'node:assert/strict';
+import { mkdtemp, rm } from 'node:fs/promises';
+import { tmpdir } from 'node:os';
+import path from 'node:path';
+import test from 'node:test';
+
+import { closeConnection, initializeDatabase, sessionsDb } from '@/modules/database/index.js';
+import { chatRunRegistry } from '@/modules/websocket/services/chat-run-registry.service.js';
+import type { NormalizedMessage } from '@/shared/types.js';
+
+/**
+ * Minimal stand-in for a websocket connection: collects every JSON frame the
+ * gateway writer forwards so assertions can inspect the outbound protocol.
+ */
+class FakeConnection {
+  readyState = 1; // WS_OPEN_STATE
+  frames: NormalizedMessage[] = [];
+
+  send(data: string): void {
+    this.frames.push(JSON.parse(data) as NormalizedMessage);
+  }
+}
+
+async function withIsolatedDatabase(runTest: () => void | Promise<void>): Promise<void> {
+  const previousDatabasePath = process.env.DATABASE_PATH;
+  const tempDirectory = await mkdtemp(path.join(tmpdir(), 'chat-run-registry-'));
+  const databasePath = path.join(tempDirectory, 'auth.db');
+
+  closeConnection();
+  process.env.DATABASE_PATH = databasePath;
+  await initializeDatabase();
+
+  try {
+    await runTest();
+  } finally {
+    chatRunRegistry.clearAll();
+    closeConnection();
+    if (previousDatabasePath === undefined) {
+      delete process.env.DATABASE_PATH;
+    } else {
+      process.env.DATABASE_PATH = previousDatabasePath;
+    }
+    await rm(tempDirectory, { recursive: true, force: true });
+  }
+}
+
+test('live events are remapped to the app session id and sequenced', async () => {
+  await withIsolatedDatabase(() => {
+    sessionsDb.createAppSession('app-run-1', 'claude', '/workspace/demo');
+    const connection = new FakeConnection();
+    const run = chatRunRegistry.startRun({
+      appSessionId: 'app-run-1',
+      provider: 'claude',
+      providerSessionId: null,
+      connection,
+      userId: 'user-1',
+    });
+    assert.ok(run);
+
+    run.writer.send({ kind: 'stream_delta', provider: 'claude', sessionId: 'provider-id-9', content: 'hello' });
+    run.writer.send({ kind: 'text', provider: 'claude', sessionId: 'provider-id-9', content: 'hello world' });
+
+    assert.equal(connection.frames.length, 2);
+    assert.equal(connection.frames[0]?.sessionId, 'app-run-1');
+    assert.equal(connection.frames[0]?.seq, 1);
+    assert.equal(connection.frames[1]?.sessionId, 'app-run-1');
+    assert.equal(connection.frames[1]?.seq, 2);
+  });
+});
+
+test('session_created is swallowed and persisted as the provider-id mapping', async () => {
+  await withIsolatedDatabase(() => {
+    sessionsDb.createAppSession('app-run-2', 'cursor', '/workspace/demo');
+    const connection = new FakeConnection();
+    const run = chatRunRegistry.startRun({
+      appSessionId: 'app-run-2',
+      provider: 'cursor',
+      providerSessionId: null,
+      connection,
+      userId: null,
+    });
+    assert.ok(run);
+
+    run.writer.send({
+      kind: 'session_created',
+      provider: 'cursor',
+      sessionId: 'cursor-native-7',
+      newSessionId: 'cursor-native-7',
+    });
+
+    // Never forwarded to the client...
+    assert.equal(connection.frames.length, 0);
+    // ...but recorded in the registry and persisted in the database.
+    assert.equal(run.providerSessionId, 'cursor-native-7');
+    assert.equal(sessionsDb.getSessionById('app-run-2')?.provider_session_id, 'cursor-native-7');
+  });
+});
+
+test('complete marks the run finished and duplicate completes are dropped', async () => {
+  await withIsolatedDatabase(() => {
+    sessionsDb.createAppSession('app-run-3', 'codex', '/workspace/demo');
+    const connection = new FakeConnection();
+    const run = chatRunRegistry.startRun({
+      appSessionId: 'app-run-3',
+      provider: 'codex',
+      providerSessionId: null,
+      connection,
+      userId: null,
+    });
+    assert.ok(run);
+
+    run.writer.send({ kind: 'complete', provider: 'codex', sessionId: 'native-3', exitCode: 0 });
+    // Late duplicate from a killed runtime's exit handler.
+    run.writer.send({ kind: 'complete', provider: 'codex', sessionId: 'native-3', exitCode: 1 });
+
+    const completes = connection.frames.filter((frame) => frame.kind === 'complete');
+    assert.equal(completes.length, 1);
+    assert.equal(completes[0]?.actualSessionId, 'app-run-3');
+    assert.equal(chatRunRegistry.isProcessing('app-run-3'), false);
+
+    // completeRun is also a no-op once the run already completed.
+    chatRunRegistry.completeRun('app-run-3', { exitCode: 1 });
+    assert.equal(connection.frames.filter((frame) => frame.kind === 'complete').length, 1);
+  });
+});
+
+test('listRunningRuns returns only currently running app sessions', async () => {
+  await withIsolatedDatabase(() => {
+    sessionsDb.createAppSession('app-run-7', 'claude', '/workspace/demo');
+    sessionsDb.createAppSession('app-run-8', 'codex', '/workspace/demo');
+    const connection = new FakeConnection();
+
+    const completedRun = chatRunRegistry.startRun({
+      appSessionId: 'app-run-7',
+      provider: 'claude',
+      providerSessionId: null,
+      connection,
+      userId: null,
+    });
+    assert.ok(completedRun);
+
+    const runningRun = chatRunRegistry.startRun({
+      appSessionId: 'app-run-8',
+      provider: 'codex',
+      providerSessionId: null,
+      connection,
+      userId: null,
+    });
+    assert.ok(runningRun);
+
+    chatRunRegistry.completeRun('app-run-7', { exitCode: 0 });
+
+    const runningSessions = chatRunRegistry.listRunningRuns();
+    assert.deepEqual(runningSessions.map((session) => session.sessionId), ['app-run-8']);
+    assert.equal(runningSessions[0]?.provider, 'codex');
+  });
+});
+
+test('replayEvents returns only events after the requested seq', async () => {
+  await withIsolatedDatabase(() => {
+    sessionsDb.createAppSession('app-run-4', 'claude', '/workspace/demo');
+    const connection = new FakeConnection();
+    const run = chatRunRegistry.startRun({
+      appSessionId: 'app-run-4',
+      provider: 'claude',
+      providerSessionId: null,
+      connection,
+      userId: null,
+    });
+    assert.ok(run);
+
+    run.writer.send({ kind: 'stream_delta', provider: 'claude', sessionId: 'x', content: 'a' });
+    run.writer.send({ kind: 'stream_delta', provider: 'claude', sessionId: 'x', content: 'b' });
+    run.writer.send({ kind: 'stream_delta', provider: 'claude', sessionId: 'x', content: 'c' });
+
+    const replayed = chatRunRegistry.replayEvents('app-run-4', 1);
+    assert.deepEqual(replayed.map((event) => event.content), ['b', 'c']);
+    assert.deepEqual(replayed.map((event) => event.seq), [2, 3]);
+  });
+});
+
+test('attachConnection reroutes the live stream to a new socket', async () => {
+  await withIsolatedDatabase(() => {
+    sessionsDb.createAppSession('app-run-5', 'gemini', '/workspace/demo');
+    const firstConnection = new FakeConnection();
+    const run = chatRunRegistry.startRun({
+      appSessionId: 'app-run-5',
+      provider: 'gemini',
+      providerSessionId: null,
+      connection: firstConnection,
+      userId: null,
+    });
+    assert.ok(run);
+
+    run.writer.send({ kind: 'stream_delta', provider: 'gemini', sessionId: 'g', content: 'before' });
+
+    const secondConnection = new FakeConnection();
+    assert.equal(chatRunRegistry.attachConnection('app-run-5', secondConnection), true);
+    run.writer.send({ kind: 'stream_delta', provider: 'gemini', sessionId: 'g', content: 'after' });
+
+    assert.deepEqual(firstConnection.frames.map((frame) => frame.content), ['before']);
+    assert.deepEqual(secondConnection.frames.map((frame) => frame.content), ['after']);
+  });
+});
+
+test('startRun rejects a second concurrent run for the same session', async () => {
+  await withIsolatedDatabase(() => {
+    sessionsDb.createAppSession('app-run-6', 'opencode', '/workspace/demo');
+    const connection = new FakeConnection();
+    const first = chatRunRegistry.startRun({
+      appSessionId: 'app-run-6',
+      provider: 'opencode',
+      providerSessionId: null,
+      connection,
+      userId: null,
+    });
+    assert.ok(first);
+
+    const second = chatRunRegistry.startRun({
+      appSessionId: 'app-run-6',
+      provider: 'opencode',
+      providerSessionId: null,
+      connection,
+      userId: null,
+    });
+    assert.equal(second, null);
+
+    // After the run finishes a new one is allowed again.
+    chatRunRegistry.completeRun('app-run-6', { exitCode: 0 });
+    const third = chatRunRegistry.startRun({
+      appSessionId: 'app-run-6',
+      provider: 'opencode',
+      providerSessionId: null,
+      connection,
+      userId: null,
+    });
+    assert.ok(third);
+  });
+});

server/openai-codex.js

@@ -18,7 +18,7 @@ import { notifyRunFailed, notifyRunStopped } from './services/notification-orche
 import { sessionsService } from './modules/providers/services/sessions.service.js';
 import { providerAuthService } from './modules/providers/services/provider-auth.service.js';
 import { providerModelsService } from './modules/providers/services/provider-models.service.js';
-import { createNormalizedMessage } from './shared/utils.js';
+import { createCompleteMessage, createNormalizedMessage } from './shared/utils.js';
 
 // Track active sessions
 const activeCodexSessions = new Map();
@@ -352,21 +352,26 @@ export async function queryCodex(command, options = {}, ws) {
       }
     }
 
-    // Send completion event
-    if (!terminalFailure) {
-      sendMessage(ws, createNormalizedMessage({
-        kind: 'complete',
-        actualSessionId: capturedSessionId || thread.id || sessionId || null,
-        sessionId: capturedSessionId || sessionId || null,
-        provider: 'codex'
-      }));
-      notifyRunStopped({
-        userId: ws?.userId || null,
+    // Send the terminal completion event — skipped for aborted runs, whose
+    // terminal `complete` (aborted: true) was already sent by abort-session.
+    const runSession = capturedSessionId ? activeCodexSessions.get(capturedSessionId) : null;
+    const runAborted = runSession?.status === 'aborted' || abortController.signal.aborted;
+    if (!runAborted) {
+      sendMessage(ws, createCompleteMessage({
         provider: 'codex',
         sessionId: capturedSessionId || sessionId || null,
-        sessionName: sessionSummary,
-        stopReason: 'completed'
-      });
+        actualSessionId: capturedSessionId || thread.id || sessionId || null,
+        exitCode: terminalFailure ? 1 : 0,
+      }));
+      if (!terminalFailure) {
+        notifyRunStopped({
+          userId: ws?.userId || null,
+          provider: 'codex',
+          sessionId: capturedSessionId || sessionId || null,
+          sessionName: sessionSummary,
+          stopReason: 'completed'
+        });
+      }
     }
 
   } catch (error) {
@@ -386,6 +391,11 @@ export async function queryCodex(command, options = {}, ws) {
         : error.message;
 
       sendMessage(ws, createNormalizedMessage({ kind: 'error', content: errorContent, sessionId: capturedSessionId || sessionId || null, provider: 'codex' }));
+      sendMessage(ws, createCompleteMessage({
+        provider: 'codex',
+        sessionId: capturedSessionId || sessionId || null,
+        exitCode: 1,
+      }));
       if (!terminalFailure) {
         notifyRunFailed({
           userId: ws?.userId || null,

server/opencode-cli.js

@@ -8,7 +8,7 @@ import { sessionsService } from './modules/providers/services/sessions.service.j
 import { providerAuthService } from './modules/providers/services/provider-auth.service.js';
 import { providerModelsService } from './modules/providers/services/provider-models.service.js';
 import { notifyRunFailed, notifyRunStopped } from './services/notification-orchestrator.js';
-import { createNormalizedMessage, getOpenCodeDatabasePath } from './shared/utils.js';
+import { createCompleteMessage, createNormalizedMessage, getOpenCodeDatabasePath } from './shared/utils.js';
 
 const spawnFunction = process.platform === 'win32' ? crossSpawn : spawn;
 
@@ -92,6 +92,9 @@ async function spawnOpenCode(command, options = {}, ws) {
     let stdoutLineBuffer = '';
     let terminalNotificationSent = false;
     let opencodeProcess = null;
+    // Unified lifecycle contract: exactly one terminal `complete` per run
+    // (close and error handlers can both fire for spawn failures).
+    let completeSent = false;
 
     const notifyTerminalState = ({ code = null, error = null } = {}) => {
       if (terminalNotificationSent) {
@@ -256,13 +259,12 @@ async function spawnOpenCode(command, options = {}, ws) {
           }));
         }
 
-        ws.send(createNormalizedMessage({
-          kind: 'complete',
-          exitCode: code,
-          isNewSession: !sessionId && !!command,
-          sessionId: finalSessionId,
-          provider: 'opencode',
-        }));
+        // Terminal complete — skipped for aborted runs (abort-session
+        // already sent the aborted complete on this run's behalf).
+        if (!completeSent && !opencodeProcess.aborted) {
+          completeSent = true;
+          ws.send(createCompleteMessage({ provider: 'opencode', sessionId: finalSessionId, exitCode: code }));
+        }
 
         if (code === 0) {
           notifyTerminalState({ code });
@@ -302,6 +304,10 @@ async function spawnOpenCode(command, options = {}, ws) {
           sessionId: finalSessionId,
           provider: 'opencode',
         }));
+        if (!completeSent && !opencodeProcess.aborted) {
+          completeSent = true;
+          ws.send(createCompleteMessage({ provider: 'opencode', sessionId: finalSessionId, exitCode: 1 }));
+        }
         notifyTerminalState({ error });
         reject(error);
       });
@@ -315,6 +321,9 @@ function abortOpenCodeSession(sessionId) {
     return false;
   }
 
+  // The abort handler sends the terminal complete (aborted: true); flag the
+  // process so its close handler does not emit a second one.
+  process.aborted = true;
   process.kill('SIGTERM');
   activeOpenCodeProcesses.delete(sessionId);
   return true;

server/routes/agent.js

@@ -592,12 +592,14 @@ class ResponseCollector {
       }
     }
 
+    const inputTokens = totalInput + totalCacheRead + totalCacheCreation;
+
     return {
-      inputTokens: totalInput,
+      inputTokens,
       outputTokens: totalOutput,
       cacheReadTokens: totalCacheRead,
       cacheCreationTokens: totalCacheCreation,
-      totalTokens: totalInput + totalOutput + totalCacheRead + totalCacheCreation
+      totalTokens: inputTokens + totalOutput
     };
   }
 }
@@ -644,7 +646,7 @@ class ResponseCollector {
  *
  * @param {string} model - (Optional) Model identifier for providers.
  *
- *                        Claude models: 'sonnet' (default), 'opus', 'haiku', 'opusplan', 'sonnet[1m]'
+ *                        Claude models: 'sonnet' (default), 'opus', 'haiku', 'opusplan', 'sonnet[1m]', 'fable'
  *                        Cursor models: 'gpt-5' (default), 'gpt-5.2', 'gpt-5.2-high', 'sonnet-4.5', 'opus-4.5',
  *                                       'gemini-3-pro', 'composer-1', 'auto', 'gpt-5.1', 'gpt-5.1-high',
  *                                       'gpt-5.1-codex', 'gpt-5.1-codex-high', 'gpt-5.1-codex-max',

server/routes/commands.js

@@ -268,16 +268,35 @@ Custom commands can be created in:
           tokenUsage.contextWindow ??
           0,
       ) || 0;
-    const inputTokensRaw =
+    const normalizedInputValue =
+      tokenUsage.inputTokens ??
+      tokenUsage.input ??
+      tokenUsage.cumulativeInputTokens ??
+      tokenUsage.breakdown?.input ??
+      tokenUsage.promptTokens;
+    const directInputTokens =
       Number(
-        tokenUsage.inputTokens ??
-          tokenUsage.input ??
+        normalizedInputValue ??
           tokenUsage.input_tokens ??
-          tokenUsage.cumulativeInputTokens ??
-          tokenUsage.breakdown?.input ??
-          tokenUsage.promptTokens ??
+          0
+      ) || 0;
+    const cacheReadTokens =
+      Number(
+        tokenUsage.cacheReadTokens ??
+          tokenUsage.cache_read_input_tokens ??
+          tokenUsage.cacheReadInputTokens ??
           0,
       ) || 0;
+    const cacheCreationTokens =
+      Number(
+        tokenUsage.cacheCreationTokens ??
+          tokenUsage.cache_creation_input_tokens ??
+          tokenUsage.cacheCreationInputTokens ??
+          0,
+      ) || 0;
+    const inputTokens = normalizedInputValue == null
+      ? directInputTokens + cacheReadTokens + cacheCreationTokens
+      : directInputTokens;
     const outputTokens =
       Number(
         tokenUsage.outputTokens ??
@@ -288,8 +307,9 @@ Custom commands can be created in:
           tokenUsage.completionTokens ??
           0,
       ) || 0;
-    const hasTokenBreakdown = inputTokensRaw > 0 || outputTokens > 0;
-    const used = reportedUsed || inputTokensRaw + outputTokens;
+    const computedUsed = inputTokens + outputTokens;
+    const hasTokenBreakdown = computedUsed > 0;
+    const used = Math.max(reportedUsed, computedUsed);
 
     return {
       type: "builtin",
@@ -302,7 +322,7 @@ Custom commands can be created in:
         ...(hasTokenBreakdown
           ? {
               tokenBreakdown: {
-                input: inputTokensRaw,
+                input: inputTokens,
                 output: outputTokens,
               },
             }

server/services/notification-orchestrator.js

@@ -98,6 +98,44 @@ function normalizeSessionName(sessionName) {
   return normalized.length > 80 ? `${normalized.slice(0, 77)}...` : normalized;
 }
 
+function rowMatchesProvider(row, provider) {
+  return row && (!provider || row.provider === provider);
+}
+
+function resolveSessionRow(sessionId, provider) {
+  if (!sessionId) {
+    return null;
+  }
+
+  const appSessionRow = sessionsDb.getSessionById(sessionId);
+  if (rowMatchesProvider(appSessionRow, provider)) {
+    return appSessionRow;
+  }
+
+  const providerSessionRow = sessionsDb.getSessionByProviderSessionId(sessionId);
+  if (rowMatchesProvider(providerSessionRow, provider)) {
+    return providerSessionRow;
+  }
+
+  return null;
+}
+
+function normalizeNotificationSession(event) {
+  if (!event?.sessionId || !event.provider || event.provider === 'system') {
+    return event;
+  }
+
+  const row = resolveSessionRow(event.sessionId, event.provider);
+  if (!row || row.session_id === event.sessionId) {
+    return event;
+  }
+
+  return {
+    ...event,
+    sessionId: row.session_id
+  };
+}
+
 function resolveSessionName(event) {
   const explicitSessionName = normalizeSessionName(event.meta?.sessionName);
   if (explicitSessionName) {
@@ -112,28 +150,29 @@ function resolveSessionName(event) {
 }
 
 function buildPushBody(event) {
+  const normalizedEvent = normalizeNotificationSession(event);
   const CODE_MAP = {
-    'permission.required': event.meta?.toolName
-      ? `Action Required: Tool "${event.meta.toolName}" needs approval`
+    'permission.required': normalizedEvent.meta?.toolName
+      ? `Action Required: Tool "${normalizedEvent.meta.toolName}" needs approval`
       : 'Action Required: A tool needs your approval',
-    'run.stopped': event.meta?.stopReason || 'Run Stopped: The run has stopped',
-    'run.failed': event.meta?.error ? `Run Failed: ${event.meta.error}` : 'Run Failed: The run encountered an error',
-    'agent.notification': event.meta?.message ? String(event.meta.message) : 'You have a new notification',
+    'run.stopped': normalizedEvent.meta?.stopReason || 'Run Stopped: The run has stopped',
+    'run.failed': normalizedEvent.meta?.error ? `Run Failed: ${normalizedEvent.meta.error}` : 'Run Failed: The run encountered an error',
+    'agent.notification': normalizedEvent.meta?.message ? String(normalizedEvent.meta.message) : 'You have a new notification',
     'push.enabled': 'Push notifications are now enabled!'
   };
-  const providerLabel = PROVIDER_LABELS[event.provider] || 'Assistant';
-  const sessionName = resolveSessionName(event);
-  const message = CODE_MAP[event.code] || 'You have a new notification';
+  const providerLabel = PROVIDER_LABELS[normalizedEvent.provider] || 'Assistant';
+  const sessionName = resolveSessionName(normalizedEvent);
+  const message = CODE_MAP[normalizedEvent.code] || 'You have a new notification';
 
   return {
     title: sessionName || 'CloudCLI',
     body: `${providerLabel}: ${message}`,
     data: {
-      sessionId: event.sessionId || null,
-      code: event.code,
-      provider: event.provider || null,
+      sessionId: normalizedEvent.sessionId || null,
+      code: normalizedEvent.code,
+      provider: normalizedEvent.provider || null,
       sessionName,
-      tag: `${event.provider || 'assistant'}:${event.sessionId || 'none'}:${event.code}`
+      tag: `${normalizedEvent.provider || 'assistant'}:${normalizedEvent.sessionId || 'none'}:${normalizedEvent.code}`
     }
   };
 }
@@ -175,15 +214,16 @@ function notifyUserIfEnabled({ userId, event }) {
     return;
   }
 
+  const normalizedEvent = normalizeNotificationSession(event);
   const preferences = notificationPreferencesDb.getPreferences(userId);
-  if (!shouldSendPush(preferences, event)) {
+  if (!shouldSendPush(preferences, normalizedEvent)) {
     return;
   }
-  if (isDuplicate(event)) {
+  if (isDuplicate(normalizedEvent)) {
     return;
   }
 
-  sendWebPush(userId, event).catch((err) => {
+  sendWebPush(userId, normalizedEvent).catch((err) => {
     console.error('Web push send error:', err);
   });
 }

server/services/notification-orchestrator.test.js

@@ -0,0 +1,80 @@
+import assert from 'node:assert/strict';
+import { mkdtemp, rm } from 'node:fs/promises';
+import { tmpdir } from 'node:os';
+import path from 'node:path';
+import test from 'node:test';
+
+import webPush from 'web-push';
+
+import {
+  closeConnection,
+  initializeDatabase,
+  notificationPreferencesDb,
+  pushSubscriptionsDb,
+  sessionsDb,
+  userDb,
+} from '../modules/database/index.js';
+
+import { notifyRunStopped } from './notification-orchestrator.js';
+
+async function withIsolatedDatabase(runTest) {
+  const previousDatabasePath = process.env.DATABASE_PATH;
+  const tempDirectory = await mkdtemp(path.join(tmpdir(), 'notification-orchestrator-'));
+  const databasePath = path.join(tempDirectory, 'auth.db');
+
+  closeConnection();
+  process.env.DATABASE_PATH = databasePath;
+  await initializeDatabase();
+
+  try {
+    await runTest();
+  } finally {
+    closeConnection();
+    if (previousDatabasePath === undefined) {
+      delete process.env.DATABASE_PATH;
+    } else {
+      process.env.DATABASE_PATH = previousDatabasePath;
+    }
+    await rm(tempDirectory, { recursive: true, force: true });
+  }
+}
+
+test('push payload uses the app session id when notified with a provider session id', async () => {
+  const originalSendNotification = webPush.sendNotification;
+  const sentPayloads = [];
+
+  webPush.sendNotification = async (_subscription, payload) => {
+    sentPayloads.push(JSON.parse(payload));
+    return {};
+  };
+
+  try {
+    await withIsolatedDatabase(async () => {
+      const user = userDb.createUser('notify-user', 'hash');
+      const userId = Number(user.id);
+
+      notificationPreferencesDb.updatePreferences(userId, {
+        channels: { webPush: true },
+        events: { actionRequired: true, stop: true, error: true },
+      });
+      pushSubscriptionsDb.saveSubscription(userId, 'https://example.test/push', 'p256dh', 'auth');
+      sessionsDb.createAppSession('app-session-1', 'claude', '/workspace/demo');
+      sessionsDb.assignProviderSessionId('app-session-1', 'claude-native-1');
+
+      notifyRunStopped({
+        userId,
+        provider: 'claude',
+        sessionId: 'claude-native-1',
+        stopReason: 'completed',
+      });
+
+      await new Promise((resolve) => setImmediate(resolve));
+
+      assert.equal(sentPayloads.length, 1);
+      assert.equal(sentPayloads[0]?.data?.sessionId, 'app-session-1');
+      assert.match(sentPayloads[0]?.data?.tag, /app-session-1/);
+    });
+  } finally {
+    webPush.sendNotification = originalSendNotification;
+  }
+});

server/shared/tests/slice-tail-page.test.ts

@@ -0,0 +1,42 @@
+import assert from 'node:assert/strict';
+import test from 'node:test';
+
+import { sliceTailPage } from '@/shared/utils.js';
+
+const ITEMS = ['a', 'b', 'c', 'd', 'e'];
+
+test('offset 0 returns the most recent page', () => {
+  const { page, hasMore } = sliceTailPage(ITEMS, 2, 0);
+  assert.deepEqual(page, ['d', 'e']);
+  assert.equal(hasMore, true);
+});
+
+test('increasing offsets walk backwards in time', () => {
+  const { page, hasMore } = sliceTailPage(ITEMS, 2, 2);
+  assert.deepEqual(page, ['b', 'c']);
+  assert.equal(hasMore, true);
+});
+
+test('the oldest page reports hasMore false', () => {
+  const { page, hasMore } = sliceTailPage(ITEMS, 2, 4);
+  assert.deepEqual(page, ['a']);
+  assert.equal(hasMore, false);
+});
+
+test('null limit returns everything', () => {
+  const { page, hasMore } = sliceTailPage(ITEMS, null, 0);
+  assert.deepEqual(page, ITEMS);
+  assert.equal(hasMore, false);
+});
+
+test('offsets past the start return an empty page', () => {
+  const { page, hasMore } = sliceTailPage(ITEMS, 3, 10);
+  assert.deepEqual(page, []);
+  assert.equal(hasMore, false);
+});
+
+test('zero limit returns an empty page but keeps hasMore accurate', () => {
+  const { page, hasMore } = sliceTailPage(ITEMS, 0, 0);
+  assert.deepEqual(page, []);
+  assert.equal(hasMore, true);
+});

server/shared/types.ts

@@ -68,7 +68,7 @@ export type AuthenticatedWebSocketRequest = IncomingMessage & {
 export type LLMProvider = 'claude' | 'codex' | 'gemini' | 'cursor' | 'opencode';
 
 /**
- * One selectable model row (matches the documentation `public/modelConstants.js` option shape).
+ * One selectable model row in a provider model catalog.
  */
 export type ProviderModelOption = {
   value: string;
@@ -175,6 +175,30 @@ export type MessageKind =
   | 'interactive_prompt'
   | 'task_notification';
 
+/**
+ * Event kinds added by the chat gateway layer on top of provider message kinds.
+ *
+ * These are app-level realtime events (subscription acks, sidebar deltas,
+ * project loading progress, protocol failures) that are not produced by any
+ * provider adapter. Together with `MessageKind` they form the complete set of
+ * `kind` values a websocket client can receive, so the frontend only ever
+ * needs one kind-based switch.
+ */
+export type GatewayEventKind =
+  | 'chat_subscribed'
+  | 'session_upserted'
+  | 'loading_progress'
+  | 'protocol_error';
+
+/**
+ * Complete set of `kind` values emitted to websocket clients.
+ *
+ * Every server-to-client websocket frame carries a `kind` from this union.
+ * Provider runtimes emit `MessageKind` values; gateway services emit
+ * `GatewayEventKind` values.
+ */
+export type ServerEventKind = MessageKind | GatewayEventKind;
+
 /**
  * Provider-neutral message envelope used in REST responses and realtime channels.
  *
@@ -187,6 +211,13 @@ export type NormalizedMessage = {
   timestamp: string;
   provider: LLMProvider;
   kind: MessageKind;
+  /**
+   * Monotonic per-run sequence number assigned by the chat run registry when a
+   * live event is forwarded to the websocket. History messages loaded over
+   * REST do not carry it. Clients use it with `chat.subscribe` to replay only
+   * the live events they missed across websocket reconnects.
+   */
+  seq?: number;
   role?: 'user' | 'assistant';
   content?: string;
   /**
@@ -237,11 +268,18 @@ export type NormalizedMessage = {
  *
  * Consumers should pass provider-specific lookup hints (`projectPath`) only
  * when the selected provider requires them.
+ *
+ * `providerSessionId` is the provider-native session id from the sessions
+ * index (transcript file name / provider database key). Provider adapters
+ * must use it — never the app-facing session id they were called with — when
+ * matching transcript rows on disk, because app-created sessions use an
+ * app-allocated id that the provider has never seen.
  */
 export type FetchHistoryOptions = {
   projectPath?: string;
   limit?: number | null;
   offset?: number;
+  providerSessionId?: string;
 };
 
 /**

server/shared/utils.ts

@@ -346,6 +346,84 @@ export function createNormalizedMessage(fields: NormalizedMessageInput): Normali
   };
 }
 
+/**
+ * Build the unified terminal `complete` lifecycle message.
+ *
+ * Contract: every provider run ends with exactly one `complete` (the
+ * abort-session handler emits it on behalf of cancelled runs, so aborted runs
+ * must NOT emit their own). The frontend treats `complete` as the only
+ * terminal signal and never needs provider-specific handling:
+ *
+ * - `sessionId`     — the id the client knows this run by ('' if never discovered)
+ * - `actualSessionId` — canonical id after the run; equals `sessionId` unless
+ *                       the provider rewrote it mid-run
+ * - `exitCode`      — 0 on success; a missing/null code (e.g. killed process)
+ *                     is reported as failure
+ * - `success`       — exitCode === 0 and not aborted
+ * - `aborted`       — run was cancelled by the user
+ */
+export function createCompleteMessage(opts: {
+  provider: NormalizedMessage['provider'];
+  sessionId?: string | null;
+  actualSessionId?: string | null;
+  exitCode?: number | null;
+  aborted?: boolean;
+}): NormalizedMessage {
+  const exitCode = typeof opts.exitCode === 'number' ? opts.exitCode : 1;
+  const aborted = Boolean(opts.aborted);
+
+  return createNormalizedMessage({
+    kind: 'complete',
+    provider: opts.provider,
+    sessionId: opts.sessionId || null,
+    actualSessionId: opts.actualSessionId || opts.sessionId || null,
+    exitCode,
+    success: exitCode === 0 && !aborted,
+    aborted,
+  });
+}
+
+// ---------------------------
+//----------------- CONVERSATION HISTORY PAGINATION UTILITIES ------------
+/**
+ * Slices one page from the END of a chronologically ordered message list.
+ *
+ * This is the single pagination contract for conversation history across all
+ * providers: `offset = 0` returns the most recent `limit` items, increasing
+ * offsets walk backwards in time (for "scroll up to load older" UIs), and a
+ * `null` limit returns everything. Items must already be sorted oldest-first;
+ * the returned page preserves that order.
+ *
+ * Every provider history reader must use this helper instead of slicing
+ * manually so `offset`/`limit` query params behave identically regardless of
+ * which provider produced the session.
+ */
+export function sliceTailPage<T>(
+  items: T[],
+  limit: number | null,
+  offset: number,
+): { page: T[]; hasMore: boolean } {
+  const total = items.length;
+  const normalizedOffset = Math.max(0, offset);
+
+  if (limit === null) {
+    // A null limit returns the full list; offset still trims newest entries
+    // so "everything before the page I already have" stays expressible.
+    const end = Math.max(0, total - normalizedOffset);
+    return {
+      page: items.slice(0, end),
+      hasMore: false,
+    };
+  }
+
+  const end = Math.max(0, total - normalizedOffset);
+  const start = Math.max(0, end - Math.max(0, limit));
+  return {
+    page: items.slice(start, end),
+    hasMore: start > 0,
+  };
+}
+
 // ---------------------------
 //----------------- MCP CONFIG PARSING UTILITIES ------------
 /**

server/utils/plugin-loader.js

@@ -1,7 +1,8 @@
 import fs from 'fs';
 import path from 'path';
 import os from 'os';
-import { spawn } from 'child_process';
+
+import { spawn } from 'cross-spawn';
 
 const PLUGINS_DIR = path.join(os.homedir(), '.claude-code-ui', 'plugins');
 const PLUGINS_CONFIG_PATH = path.join(os.homedir(), '.claude-code-ui', 'plugins.json');

server/utils/plugin-process-manager.js

@@ -7,6 +7,41 @@ const runningPlugins = new Map();
 // Map<pluginName, Promise<port>> — in-flight start operations
 const startingPlugins = new Map();
 
+/**
+ * Build the environment handed to a plugin server subprocess.
+ *
+ * Intentionally minimal: only non-secret essentials, never the host's full
+ * environment. On Windows a handful of system variables are required for any
+ * child to bootstrap (Node itself, and any Python or CLI a plugin shells out
+ * to). Without APPDATA a `pip install --user` tool cannot locate its
+ * site-packages and fails to import; SystemRoot, PATHEXT and TEMP are needed to
+ * resolve system DLLs, executable extensions and a temp directory. None of
+ * these carry secrets, so the ones that are set get passed straight through.
+ */
+function buildPluginEnv(name) {
+  const env = {
+    PATH: process.env.PATH,
+    HOME: process.env.HOME,
+    NODE_ENV: process.env.NODE_ENV || 'production',
+    PLUGIN_NAME: name,
+  };
+
+  if (process.platform === 'win32') {
+    const WINDOWS_ESSENTIALS = [
+      'SystemRoot', 'windir', 'SystemDrive',
+      'USERPROFILE', 'APPDATA', 'LOCALAPPDATA',
+      'TEMP', 'TMP', 'PATHEXT',
+    ];
+    for (const key of WINDOWS_ESSENTIALS) {
+      if (process.env[key] !== undefined) {
+        env[key] = process.env[key];
+      }
+    }
+  }
+
+  return env;
+}
+
 /**
  * Start a plugin's server subprocess.
  * The plugin's server entry must print a JSON line with { ready: true, port: <number> }
@@ -26,15 +61,9 @@ export function startPluginServer(name, pluginDir, serverEntry) {
 
     const serverPath = path.join(pluginDir, serverEntry);
 
-    // Restricted env — only essentials, no host secrets
     const pluginProcess = spawn('node', [serverPath], {
       cwd: pluginDir,
-      env: {
-        PATH: process.env.PATH,
-        HOME: process.env.HOME,
-        NODE_ENV: process.env.NODE_ENV || 'production',
-        PLUGIN_NAME: name,
-      },
+      env: buildPluginEnv(name),
       stdio: ['ignore', 'pipe', 'pipe'],
     });
 

src/components/app/AppContent.tsx

@@ -1,4 +1,4 @@
-import { useEffect, useRef } from 'react';
+import { useCallback, useEffect } from 'react';
 import { useNavigate, useParams } from 'react-router-dom';
 import { useTranslation } from 'react-i18next';
 
@@ -10,6 +10,33 @@ import { PaletteOpsProvider, usePaletteOpsRegister } from '../../contexts/Palett
 import { useDeviceSettings } from '../../hooks/useDeviceSettings';
 import { useSessionProtection } from '../../hooks/useSessionProtection';
 import { useProjectsState } from '../../hooks/useProjectsState';
+import { api } from '../../utils/api';
+
+type RunningSessionApiItem = {
+  sessionId?: unknown;
+  startedAt?: unknown;
+  statusText?: unknown;
+  canInterrupt?: unknown;
+};
+
+type RunningSessionsApiPayload = {
+  data?: {
+    sessions?: RunningSessionApiItem[];
+  };
+};
+
+const parseStartedAt = (value: unknown): number | undefined => {
+  if (typeof value === 'number' && Number.isFinite(value) && value > 0) {
+    return value;
+  }
+
+  if (typeof value !== 'string') {
+    return undefined;
+  }
+
+  const parsed = Date.parse(value);
+  return Number.isFinite(parsed) ? parsed : undefined;
+};
 
 export default function AppContent() {
   return (
@@ -24,16 +51,13 @@ function AppContentInner() {
   const { sessionId } = useParams<{ sessionId?: string }>();
   const { t } = useTranslation('common');
   const { isMobile } = useDeviceSettings({ trackPWA: false });
-  const { ws, sendMessage, latestMessage, isConnected } = useWebSocket();
-  const wasConnectedRef = useRef(false);
+  const { ws, sendMessage, subscribe } = useWebSocket();
 
   const {
-    activeSessions,
     processingSessions,
-    markSessionAsActive,
-    markSessionAsInactive,
-    markSessionAsProcessing,
-    markSessionAsNotProcessing,
+    markSessionProcessing,
+    markSessionIdle,
+    syncProcessingSessions,
   } = useSessionProtection();
 
   const {
@@ -50,16 +74,64 @@ function AppContentInner() {
     setShowSettings,
     openSettings,
     refreshProjectsSilently,
+    registerOptimisticSession,
     sidebarSharedProps,
     handleNewSession,
   } = useProjectsState({
     sessionId,
     navigate,
-    latestMessage,
+    subscribe,
     isMobile,
-    activeSessions,
+    activeSessions: processingSessions,
   });
 
+  const refreshRunningSessions = useCallback(async () => {
+    try {
+      const response = await api.runningSessions();
+      if (!response.ok) {
+        return;
+      }
+
+      const payload = (await response.json()) as RunningSessionsApiPayload;
+      const sessions = Array.isArray(payload.data?.sessions) ? payload.data.sessions : [];
+
+      syncProcessingSessions(
+        sessions
+          .map((session) => {
+            if (typeof session.sessionId !== 'string' || !session.sessionId) {
+              return null;
+            }
+
+            return {
+              sessionId: session.sessionId,
+              startedAt: parseStartedAt(session.startedAt),
+              statusText: typeof session.statusText === 'string' ? session.statusText : undefined,
+              canInterrupt: typeof session.canInterrupt === 'boolean' ? session.canInterrupt : undefined,
+            };
+          })
+          .filter((session): session is NonNullable<typeof session> => Boolean(session)),
+      );
+    } catch (error) {
+      console.error('[AppContent] Failed to sync running sessions:', error);
+    }
+  }, [syncProcessingSessions]);
+
+  useEffect(() => {
+    void refreshRunningSessions();
+  }, [refreshRunningSessions]);
+
+  useEffect(() => {
+    if (processingSessions.size === 0) {
+      return;
+    }
+
+    const interval = window.setInterval(() => {
+      void refreshRunningSessions();
+    }, 5000);
+
+    return () => window.clearInterval(interval);
+  }, [processingSessions.size, refreshRunningSessions]);
+
   usePaletteOpsRegister({
     openSettings,
     refreshProjects: refreshProjectsSilently,
@@ -99,23 +171,9 @@ function AppContentInner() {
     };
   }, [navigate, refreshProjectsSilently, setActiveTab, setSidebarOpen]);
 
-  // Permission recovery: query pending permissions on WebSocket reconnect or session change
-  useEffect(() => {
-    const isReconnect = isConnected && !wasConnectedRef.current;
-
-    if (isReconnect) {
-      wasConnectedRef.current = true;
-    } else if (!isConnected) {
-      wasConnectedRef.current = false;
-    }
-
-    if (isConnected && selectedSession?.id) {
-      sendMessage({
-        type: 'get-pending-permissions',
-        sessionId: selectedSession.id
-      });
-    }
-  }, [isConnected, selectedSession?.id, sendMessage]);
+  // Pending tool permissions are recovered through the `chat.subscribe` flow:
+  // the `chat_subscribed` ack carries them on session open and on reconnect,
+  // so no separate permission-recovery message is needed here.
 
   // Adjust the app container to stay above the virtual keyboard on iOS Safari.
   // On Chrome for Android the layout viewport already shrinks when the keyboard opens,
@@ -180,19 +238,19 @@ function AppContentInner() {
           setActiveTab={setActiveTab}
           ws={ws}
           sendMessage={sendMessage}
-          latestMessage={latestMessage}
           isMobile={isMobile}
           onMenuClick={() => setSidebarOpen(true)}
           isLoading={isLoadingProjects}
           onInputFocusChange={setIsInputFocused}
-          onSessionActive={markSessionAsActive}
-          onSessionInactive={markSessionAsInactive}
-          onSessionProcessing={markSessionAsProcessing}
-          onSessionNotProcessing={markSessionAsNotProcessing}
+          onSessionProcessing={markSessionProcessing}
+          onSessionIdle={markSessionIdle}
           processingSessions={processingSessions}
           onNavigateToSession={(targetSessionId: string, options) =>
             navigate(`/session/${targetSessionId}`, { replace: Boolean(options?.replace) })
           }
+          onSessionEstablished={(targetSessionId, context) =>
+            registerOptimisticSession({ sessionId: targetSessionId, ...context })
+          }
           onShowSettings={() => setShowSettings(true)}
           externalMessageUpdate={externalMessageUpdate}
           newSessionTrigger={newSessionTrigger}

src/components/chat/constants/thinkingModes.ts

@@ -1,44 +0,0 @@
-import { Brain, Zap, Sparkles, Atom } from 'lucide-react';
-
-export const thinkingModes = [
-  {
-    id: 'none',
-    name: 'Standard',
-    description: 'Regular Claude response',
-    icon: null,
-    prefix: '',
-    color: 'text-gray-600'
-  },
-  {
-    id: 'think',
-    name: 'Think',
-    description: 'Basic extended thinking',
-    icon: Brain,
-    prefix: 'think',
-    color: 'text-blue-600'
-  },
-  {
-    id: 'think-hard',
-    name: 'Think Hard',
-    description: 'More thorough evaluation',
-    icon: Zap,
-    prefix: 'think hard',
-    color: 'text-purple-600'
-  },
-  {
-    id: 'think-harder',
-    name: 'Think Harder',
-    description: 'Deep analysis with alternatives',
-    icon: Sparkles,
-    prefix: 'think harder',
-    color: 'text-indigo-600'
-  },
-  {
-    id: 'ultrathink',
-    name: 'Ultrathink',
-    description: 'Maximum thinking budget',
-    icon: Atom,
-    prefix: 'ultrathink',
-    color: 'text-red-600'
-  }
-];

src/components/chat/hooks/useChatComposerState.ts

@@ -12,13 +12,14 @@ import type {
 import { useDropzone } from 'react-dropzone';
 
 import { authenticatedFetch } from '../../../utils/api';
-import { thinkingModes } from '../constants/thinkingModes';
+import type { MarkSessionProcessing } from '../../../hooks/useSessionProtection';
 import { grantClaudeToolPermission } from '../utils/chatPermissions';
 import { safeLocalStorage } from '../utils/chatStorage';
 import type {
   ChatMessage,
   PendingPermissionRequest,
   PermissionMode,
+  SessionEstablishedContext,
 } from '../types/types';
 import type { Project, ProjectSession, LLMProvider, ProviderModelsCacheInfo } from '../../../types/app';
 import { escapeRegExp } from '../utils/chatFormatting';
@@ -26,10 +27,6 @@ import { escapeRegExp } from '../utils/chatFormatting';
 import { useFileMentions } from './useFileMentions';
 import { type SlashCommand, useSlashCommands } from './useSlashCommands';
 
-type PendingViewSession = {
-  startedAt: number;
-};
-
 interface UseChatComposerStateArgs {
   selectedProject: Project | null;
   selectedSession: ProjectSession | null;
@@ -47,17 +44,20 @@ interface UseChatComposerStateArgs {
   tokenBudget: Record<string, unknown> | null;
   sendMessage: (message: unknown) => void;
   sendByCtrlEnter?: boolean;
-  onSessionActive?: (sessionId?: string | null) => void;
-  onSessionProcessing?: (sessionId?: string | null) => void;
+  onSessionProcessing?: MarkSessionProcessing;
+  /**
+   * Invoked with the freshly allocated session id when the user sends the
+   * first message of a brand-new conversation. The backend allocates the id
+   * via POST /api/providers/sessions BEFORE the websocket send, so the id is
+   * stable for the conversation's whole lifetime — the consumer navigates to
+   * /session/:id and records it as the current session.
+   */
+  onSessionEstablished?: (sessionId: string, context: SessionEstablishedContext) => void;
   onInputFocusChange?: (focused: boolean) => void;
   onFileOpen?: (filePath: string, diffInfo?: unknown) => void;
   onShowSettings?: () => void;
-  pendingViewSessionRef: { current: PendingViewSession | null };
   scrollToBottom: () => void;
   addMessage: (msg: ChatMessage) => void;
-  setIsLoading: (loading: boolean) => void;
-  setCanAbortSession: (canAbort: boolean) => void;
-  setClaudeStatus: (status: { text: string; tokens: number; can_interrupt: boolean } | null) => void;
   setIsUserScrolledUp: (isScrolledUp: boolean) => void;
   setPendingPermissionRequests: Dispatch<SetStateAction<PendingPermissionRequest[]>>;
 }
@@ -178,17 +178,13 @@ export function useChatComposerState({
   tokenBudget,
   sendMessage,
   sendByCtrlEnter,
-  onSessionActive,
   onSessionProcessing,
+  onSessionEstablished,
   onInputFocusChange,
   onFileOpen,
   onShowSettings,
-  pendingViewSessionRef,
   scrollToBottom,
   addMessage,
-  setIsLoading,
-  setCanAbortSession,
-  setClaudeStatus,
   setIsUserScrolledUp,
   setPendingPermissionRequests,
 }: UseChatComposerStateArgs) {
@@ -204,7 +200,6 @@ export function useChatComposerState({
   const [uploadingImages, setUploadingImages] = useState<Map<string, number>>(new Map());
   const [imageErrors, setImageErrors] = useState<Map<string, string>>(new Map());
   const [isTextareaExpanded, setIsTextareaExpanded] = useState(false);
-  const [thinkingMode, setThinkingMode] = useState('none');
   const [commandModalPayload, setCommandModalPayload] = useState<CommandModalPayload | null>(null);
 
   const textareaRef = useRef<HTMLTextAreaElement>(null);
@@ -313,7 +308,7 @@ export function useChatComposerState({
   }, [addMessage]);
 
   const executeCommand = useCallback(
-    async (command: SlashCommand, rawInput?: string) => {
+    async (command: SlashCommand, rawInput?: string, options?: { preserveInput?: boolean }) => {
       if (!command || !selectedProject) {
         return;
       }
@@ -370,8 +365,10 @@ export function useChatComposerState({
         const result = (await response.json()) as CommandExecutionResult;
         if (result.type === 'builtin') {
           handleBuiltInCommand(result);
-          setInput('');
-          inputValueRef.current = '';
+          if (!options?.preserveInput) {
+            setInput('');
+            inputValueRef.current = '';
+          }
         } else if (result.type === 'custom') {
           await handleCustomCommand(result);
         }
@@ -402,6 +399,19 @@ export function useChatComposerState({
     ],
   );
 
+  const showCostModal = useCallback(() => {
+    executeCommand(
+      {
+        name: '/cost',
+        description: 'Display token usage information',
+        namespace: 'builtin',
+        metadata: { type: 'builtin' },
+      } as SlashCommand,
+      '/cost',
+      { preserveInput: true },
+    );
+  }, [executeCommand]);
+
   const {
     slashCommands,
     slashCommandsCount,
@@ -562,11 +572,7 @@ export function useChatComposerState({
         }
       }
 
-      let messageContent = currentInput;
-      const selectedThinkingMode = thinkingModes.find((mode: { id: string; prefix?: string }) => mode.id === thinkingMode);
-      if (selectedThinkingMode && selectedThinkingMode.prefix) {
-        messageContent = `${selectedThinkingMode.prefix}: ${currentInput}`;
-      }
+      const messageContent = currentInput;
 
       let uploadedImages: unknown[] = [];
       if (attachedImages.length > 0) {
@@ -600,8 +606,54 @@ export function useChatComposerState({
         }
       }
 
-      const effectiveSessionId =
-        currentSessionId || selectedSession?.id || sessionStorage.getItem('cursorSessionId');
+      const resolvedProjectPath = selectedProject.fullPath || selectedProject.path || '';
+      const sessionSummary = getNotificationSessionSummary(selectedSession, currentInput);
+
+      // The conversation always has a stable backend-allocated session id
+      // BEFORE the first websocket send: brand-new chats allocate one here
+      // via the session gateway. There is no client-visible session-id
+      // handoff later — this id stays valid for the conversation's lifetime.
+      let targetSessionId = selectedSession?.id || currentSessionId || null;
+      if (!targetSessionId) {
+        try {
+          const response = await authenticatedFetch('/api/providers/sessions', {
+            method: 'POST',
+            body: JSON.stringify({
+              provider,
+              projectPath: resolvedProjectPath,
+            }),
+          });
+          if (!response.ok) {
+            throw new Error(`Failed to create session (${response.status})`);
+          }
+          const body = await response.json();
+          targetSessionId = body?.data?.sessionId || null;
+        } catch (error) {
+          const message = error instanceof Error ? error.message : 'Unknown error';
+          console.error('Session creation failed:', error);
+          addMessage({
+            type: 'error',
+            content: `Failed to start a new session: ${message}`,
+            timestamp: new Date(),
+          });
+          return;
+        }
+
+        if (!targetSessionId) {
+          addMessage({
+            type: 'error',
+            content: 'Failed to start a new session: no session id returned.',
+            timestamp: new Date(),
+          });
+          return;
+        }
+
+        onSessionEstablished?.(targetSessionId, {
+          provider,
+          project: selectedProject,
+          summary: sessionSummary,
+        });
+      }
 
       const userMessage: ChatMessage = {
         type: 'user',
@@ -611,27 +663,17 @@ export function useChatComposerState({
       };
 
       addMessage(userMessage);
-      setIsLoading(true); // Processing banner starts
-      setCanAbortSession(true);
-      setClaudeStatus({
-        text: 'Processing',
-        tokens: 0,
-        can_interrupt: true,
+      // Mark this request as processing in the per-session activity map (the
+      // single source of truth the indicator derives from). The id is always
+      // concrete at this point — no pending placeholder exists anymore.
+      onSessionProcessing?.(targetSessionId, {
+        statusText: null,
+        canInterrupt: true,
       });
 
       setIsUserScrolledUp(false);
       setTimeout(() => scrollToBottom(), 100);
 
-      if (!effectiveSessionId && !selectedSession?.id) {
-        // This tracks only that a request is in flight before the provider has
-        // emitted its real session id; routing still waits for session_created.
-        pendingViewSessionRef.current = { startedAt: Date.now() };
-      }
-      if (effectiveSessionId) {
-        onSessionActive?.(effectiveSessionId);
-        onSessionProcessing?.(effectiveSessionId);
-      }
-
       const getToolsSettings = () => {
         try {
           const settingsKey =
@@ -660,87 +702,35 @@ export function useChatComposerState({
       };
 
       const toolsSettings = getToolsSettings();
-      const resolvedProjectPath = selectedProject.fullPath || selectedProject.path || '';
-      const sessionSummary = getNotificationSessionSummary(selectedSession, currentInput);
+      const model =
+        provider === 'cursor'
+          ? cursorModel
+          : provider === 'codex'
+            ? codexModel
+            : provider === 'gemini'
+              ? geminiModel
+              : provider === 'opencode'
+                ? opencodeModel
+                : claudeModel;
 
-      if (provider === 'cursor') {
-        sendMessage({
-          type: 'cursor-command',
-          command: messageContent,
-          sessionId: effectiveSessionId,
-          options: {
-            cwd: resolvedProjectPath,
-            projectPath: resolvedProjectPath,
-            sessionId: effectiveSessionId,
-            resume: Boolean(effectiveSessionId),
-            model: cursorModel,
-            skipPermissions: toolsSettings?.skipPermissions || false,
-            sessionSummary,
-            toolsSettings,
-          },
-        });
-      } else if (provider === 'codex') {
-        sendMessage({
-          type: 'codex-command',
-          command: messageContent,
-          sessionId: effectiveSessionId,
-          options: {
-            cwd: resolvedProjectPath,
-            projectPath: resolvedProjectPath,
-            sessionId: effectiveSessionId,
-            resume: Boolean(effectiveSessionId),
-            model: codexModel,
-            sessionSummary,
-            permissionMode: permissionMode === 'plan' ? 'default' : permissionMode,
-          },
-        });
-      } else if (provider === 'gemini') {
-        sendMessage({
-          type: 'gemini-command',
-          command: messageContent,
-          sessionId: effectiveSessionId,
-          options: {
-            cwd: resolvedProjectPath,
-            projectPath: resolvedProjectPath,
-            sessionId: effectiveSessionId,
-            resume: Boolean(effectiveSessionId),
-            model: geminiModel,
-            sessionSummary,
-            permissionMode,
-            toolsSettings,
-          },
-        });
-      } else if (provider === 'opencode') {
-        sendMessage({
-          type: 'opencode-command',
-          command: messageContent,
-          sessionId: effectiveSessionId,
-          options: {
-            cwd: resolvedProjectPath,
-            projectPath: resolvedProjectPath,
-            sessionId: effectiveSessionId,
-            resume: Boolean(effectiveSessionId),
-            model: opencodeModel,
-            sessionSummary,
-          },
-        });
-      } else {
-        sendMessage({
-          type: 'claude-command',
-          command: messageContent,
-          options: {
-            projectPath: resolvedProjectPath,
-            cwd: resolvedProjectPath,
-            sessionId: effectiveSessionId,
-            resume: Boolean(effectiveSessionId),
-            toolsSettings,
-            permissionMode,
-            model: claudeModel,
-            sessionSummary,
-            images: uploadedImages,
-          },
-        });
-      }
+      // One message shape for every provider. The backend resolves the
+      // provider, project path, and provider-native resume id from the
+      // session row; `options` only carries composer-level preferences.
+      sendMessage({
+        type: 'chat.send',
+        sessionId: targetSessionId,
+        content: messageContent,
+        options: {
+          model,
+          // Codex has no plan mode; downgrade rather than sending an
+          // unsupported value to its runtime.
+          permissionMode: provider === 'codex' && permissionMode === 'plan' ? 'default' : permissionMode,
+          toolsSettings,
+          skipPermissions: toolsSettings?.skipPermissions || false,
+          sessionSummary,
+          images: uploadedImages,
+        },
+      });
 
       setInput('');
       inputValueRef.current = '';
@@ -749,7 +739,6 @@ export function useChatComposerState({
       setUploadingImages(new Map());
       setImageErrors(new Map());
       setIsTextareaExpanded(false);
-      setThinkingMode('none');
 
       if (textareaRef.current) {
         textareaRef.current.style.height = 'auto';
@@ -768,22 +757,17 @@ export function useChatComposerState({
       geminiModel,
       opencodeModel,
       isLoading,
-      onSessionActive,
       onSessionProcessing,
-      pendingViewSessionRef,
+      onSessionEstablished,
       permissionMode,
       provider,
       resetCommandMenuState,
       scrollToBottom,
       selectedProject,
       sendMessage,
-      setCanAbortSession,
       addMessage,
-      setClaudeStatus,
-      setIsLoading,
       setIsUserScrolledUp,
       slashCommands,
-      thinkingMode,
     ],
   );
 
@@ -937,29 +921,19 @@ export function useChatComposerState({
       return;
     }
 
-    const cursorSessionId =
-      typeof window !== 'undefined' ? sessionStorage.getItem('cursorSessionId') : null;
-
-    const candidateSessionIds = [
-      currentSessionId,
-      provider === 'cursor' ? cursorSessionId : null,
-      selectedSession?.id || null,
-    ];
-
-    const targetSessionId =
-      candidateSessionIds.find((sessionId) => Boolean(sessionId)) || null;
-
+    const targetSessionId = selectedSession?.id || currentSessionId || null;
     if (!targetSessionId) {
-      console.warn('Abort requested but no concrete session ID is available yet.');
+      console.warn('Abort requested but no session ID is available.');
       return;
     }
 
+    // The backend resolves the provider from the session row, so no provider
+    // field is needed here.
     sendMessage({
-      type: 'abort-session',
+      type: 'chat.abort',
       sessionId: targetSessionId,
-      provider,
     });
-  }, [canAbortSession, currentSessionId, provider, selectedSession?.id, sendMessage]);
+  }, [canAbortSession, currentSessionId, selectedSession?.id, sendMessage]);
 
   const handleGrantToolPermission = useCallback(
     (suggestion: { entry: string; toolName: string }) => {
@@ -984,7 +958,7 @@ export function useChatComposerState({
 
       validIds.forEach((requestId) => {
         sendMessage({
-          type: 'claude-permission-response',
+          type: 'chat.permission-response',
           requestId,
           allow: Boolean(decision?.allow),
           updatedInput: decision?.updatedInput,
@@ -993,15 +967,11 @@ export function useChatComposerState({
         });
       });
 
-      setPendingPermissionRequests((previous) => {
-        const next = previous.filter((request) => !validIds.includes(request.requestId));
-        if (next.length === 0) {
-          setClaudeStatus(null);
-        }
-        return next;
-      });
+      setPendingPermissionRequests((previous) =>
+        previous.filter((request) => !validIds.includes(request.requestId)),
+      );
     },
-    [sendMessage, setClaudeStatus, setPendingPermissionRequests],
+    [sendMessage, setPendingPermissionRequests],
   );
 
   const [isInputFocused, setIsInputFocused] = useState(false);
@@ -1020,8 +990,6 @@ export function useChatComposerState({
     textareaRef,
     inputHighlightRef,
     isTextareaExpanded,
-    thinkingMode,
-    setThinkingMode,
     slashCommandsCount,
     filteredCommands,
     frequentCommands,
@@ -1059,5 +1027,6 @@ export function useChatComposerState({
     isInputFocused,
     commandModalPayload,
     closeCommandModal,
+    showCostModal,
   };
 }

src/components/chat/hooks/useChatMessages.ts

@@ -7,6 +7,12 @@ import type { NormalizedMessage } from '../../../stores/useSessionStore';
 import type { ChatMessage, SubagentChildTool } from '../types/types';
 import { decodeHtmlEntities, unescapeWithMathProtection, formatUsageLimitText } from '../utils/chatFormatting';
 
+function formatToolResultContent(content: unknown): string {
+  const text = typeof content === 'string' ? content : JSON.stringify(content);
+  const toolUseErrorMatch = /^<tool_use_error>([\s\S]*)<\/tool_use_error>$/.exec(text.trim());
+  return toolUseErrorMatch ? toolUseErrorMatch[1] : text;
+}
+
 /**
  * Convert NormalizedMessage[] from the session store into ChatMessage[]
  * that the existing UI components expect.
@@ -20,7 +26,12 @@ export function normalizedToChatMessages(messages: NormalizedMessage[]): ChatMes
 
   // First pass: collect tool results for attachment
   const toolResultMap = new Map<string, NormalizedMessage>();
+  const toolUseIds = new Set<string>();
   for (const msg of messages) {
+    if (msg.kind === 'tool_use' && msg.toolId) {
+      toolUseIds.add(msg.toolId);
+    }
+
     if (msg.kind === 'tool_result' && msg.toolId) {
       toolResultMap.set(msg.toolId, msg);
     }
@@ -97,7 +108,7 @@ export function normalizedToChatMessages(messages: NormalizedMessage[]): ChatMes
 
         const toolResult = tr
           ? {
-              content: typeof tr.content === 'string' ? tr.content : JSON.stringify(tr.content),
+              content: formatToolResultContent(tr.content),
               isError: Boolean(tr.isError),
               toolUseResult: (tr as any).toolUseResult,
             }
@@ -191,8 +202,25 @@ export function normalizedToChatMessages(messages: NormalizedMessage[]): ChatMes
         break;
 
       // tool_result is handled via attachment to tool_use above
-      case 'tool_result':
+      case 'tool_result': {
+        if (msg.toolId && toolUseIds.has(msg.toolId)) {
+          break;
+        }
+
+        const content = formatToolResultContent(msg.content || '');
+        if (!content.trim()) {
+          break;
+        }
+
+        converted.push({
+          type: msg.isError ? 'error' : 'assistant',
+          content,
+          timestamp: msg.timestamp,
+          toolId: msg.toolId,
+          ...sharedMetadata,
+        });
         break;
+      }
 
       default:
         break;

src/components/chat/hooks/useChatProviderState.ts

@@ -17,17 +17,35 @@ const FALLBACK_DEFAULT_MODEL: Record<LLMProvider, string> = {
   opencode: 'anthropic/claude-sonnet-4-5',
 };
 
-const getPermissionModesForProvider = (provider: LLMProvider): PermissionMode[] => {
-  if (provider === 'codex') {
-    return ['default', 'acceptEdits', 'bypassPermissions'];
-  }
-  if (provider === 'claude') {
-    return ['default', 'auto', 'acceptEdits', 'bypassPermissions', 'plan'];
-  }
-  if (provider === 'opencode') {
-    return ['default'];
-  }
-  return ['default', 'acceptEdits', 'bypassPermissions', 'plan'];
+/**
+ * Fallback permission-mode matrix used only until the backend capability
+ * matrix (`GET /api/providers/capabilities`) has loaded. The backend is the
+ * source of truth; this mirror exists so the composer renders sensibly on
+ * first paint and when the capabilities request fails.
+ */
+const FALLBACK_PERMISSION_MODES: Record<LLMProvider, PermissionMode[]> = {
+  claude: ['default', 'auto', 'acceptEdits', 'bypassPermissions', 'plan'],
+  cursor: ['default', 'acceptEdits', 'bypassPermissions', 'plan'],
+  codex: ['default', 'acceptEdits', 'bypassPermissions'],
+  gemini: ['default', 'acceptEdits', 'bypassPermissions', 'plan'],
+  opencode: ['default'],
+};
+
+type ProviderCapabilities = {
+  provider: LLMProvider;
+  permissionModes: string[];
+  defaultPermissionMode: string;
+  supportsImages: boolean;
+  supportsAbort: boolean;
+  supportsPermissionRequests: boolean;
+  supportsTokenUsage: boolean;
+};
+
+type ProviderCapabilitiesApiResponse = {
+  success?: boolean;
+  data?: {
+    providers?: ProviderCapabilities[];
+  };
 };
 
 interface UseChatProviderStateArgs {
@@ -76,6 +94,17 @@ export function useChatProviderState({ selectedSession, selectedProject }: UseCh
     return localStorage.getItem('opencode-model') || FALLBACK_DEFAULT_MODEL.opencode;
   });
 
+  /**
+   * Backend-owned capability matrix keyed by provider. Drives the permission
+   * mode picker (and is the extension point for future per-provider UI
+   * differences) so the frontend stays free of hardcoded provider branching.
+   * Null until `/api/providers/capabilities` resolves; the static fallback
+   * map covers that window.
+   */
+  const [providerCapabilities, setProviderCapabilities] = useState<
+    Partial<Record<LLMProvider, ProviderCapabilities>> | null
+  >(null);
+
   const [providerModelCatalog, setProviderModelCatalog] = useState<
     Partial<Record<LLMProvider, ProviderModelsDefinition>>
   >({});
@@ -181,6 +210,41 @@ export function useChatProviderState({ selectedSession, selectedProject }: UseCh
     void loadProviderModels();
   }, [loadProviderModels]);
 
+  useEffect(() => {
+    let cancelled = false;
+
+    const loadCapabilities = async () => {
+      try {
+        const response = await authenticatedFetch('/api/providers/capabilities');
+        const body = (await response.json()) as ProviderCapabilitiesApiResponse;
+        if (cancelled || !body.success || !Array.isArray(body.data?.providers)) {
+          return;
+        }
+
+        const byProvider: Partial<Record<LLMProvider, ProviderCapabilities>> = {};
+        for (const capabilities of body.data.providers) {
+          byProvider[capabilities.provider] = capabilities;
+        }
+        setProviderCapabilities(byProvider);
+      } catch (error) {
+        console.error('Error loading provider capabilities:', error);
+      }
+    };
+
+    void loadCapabilities();
+    return () => {
+      cancelled = true;
+    };
+  }, []);
+
+  const getPermissionModesForProvider = useCallback((targetProvider: LLMProvider): PermissionMode[] => {
+    const capabilityModes = providerCapabilities?.[targetProvider]?.permissionModes;
+    if (capabilityModes && capabilityModes.length > 0) {
+      return capabilityModes as PermissionMode[];
+    }
+    return FALLBACK_PERMISSION_MODES[targetProvider] ?? ['default'];
+  }, [providerCapabilities]);
+
   const pickStoredOrCurrent = (
     storageKey: string,
     current: string,
@@ -269,7 +333,7 @@ export function useChatProviderState({ selectedSession, selectedProject }: UseCh
     const savedMode = localStorage.getItem(`permissionMode-${selectedSession.id}`) as PermissionMode | null;
     const validModes = getPermissionModesForProvider(provider);
     setPermissionMode(savedMode && validModes.includes(savedMode) ? savedMode : 'default');
-  }, [selectedSession?.id, provider]);
+  }, [selectedSession?.id, provider, getPermissionModesForProvider]);
 
   useEffect(() => {
     if (!selectedSession?.__provider || selectedSession.__provider === provider) {
@@ -327,7 +391,7 @@ export function useChatProviderState({ selectedSession, selectedProject }: UseCh
     if (selectedSession?.id) {
       localStorage.setItem(`permissionMode-${selectedSession.id}`, nextMode);
     }
-  }, [permissionMode, provider, selectedSession?.id]);
+  }, [permissionMode, provider, selectedSession?.id, getPermissionModesForProvider]);
 
   const selectProviderModel = useCallback(async (
     targetProvider: LLMProvider,

src/components/chat/hooks/useChatRealtimeHandlers.ts

@@ -1,71 +1,34 @@
-import { useEffect, useRef } from 'react';
+import { useEffect } from 'react';
 import type { Dispatch, MutableRefObject, SetStateAction } from 'react';
 
-import { usePaletteOps } from '../../../contexts/PaletteOpsContext';
-import type { PendingPermissionRequest, SessionNavigationOptions } from '../types/types';
+import type { ServerEvent } from '../../../contexts/WebSocketContext';
+import { showCompletionTitleIndicator } from '../../../utils/pageTitleNotification';
+import { playChatCompletionSound } from '../../../utils/notificationSound';
+import type { MarkSessionIdle, MarkSessionProcessing } from '../../../hooks/useSessionProtection';
+import type { PendingPermissionRequest } from '../types/types';
 import type { ProjectSession, LLMProvider } from '../../../types/app';
 import type { SessionStore, NormalizedMessage } from '../../../stores/useSessionStore';
 
-type PendingViewSession = {
-  startedAt: number;
-};
-
-type LatestChatMessage = {
-  type?: string;
-  kind?: string;
-  data?: any;
-  message?: any;
-  delta?: string;
-  sessionId?: string;
-  session_id?: string;
-  requestId?: string;
-  toolName?: string;
-  input?: unknown;
-  context?: unknown;
-  error?: string;
-  tool?: any;
-  toolId?: string;
-  result?: any;
-  exitCode?: number;
-  isProcessing?: boolean;
-  actualSessionId?: string;
-  event?: string;
-  status?: any;
-  isNewSession?: boolean;
-  resultText?: string;
-  isError?: boolean;
-  success?: boolean;
-  reason?: string;
-  provider?: string;
-  content?: string;
-  text?: string;
-  tokens?: number;
-  canInterrupt?: boolean;
-  tokenBudget?: unknown;
-  newSessionId?: string;
-  aborted?: boolean;
-  [key: string]: any;
-};
-
 interface UseChatRealtimeHandlersArgs {
-  latestMessage: LatestChatMessage | null;
+  subscribe: (listener: (event: ServerEvent) => void) => () => void;
   provider: LLMProvider;
   selectedSession: ProjectSession | null;
   currentSessionId: string | null;
-  setCurrentSessionId: (sessionId: string | null) => void;
-  setIsLoading: (loading: boolean) => void;
-  setCanAbortSession: (canAbort: boolean) => void;
-  setClaudeStatus: (status: { text: string; tokens: number; can_interrupt: boolean } | null) => void;
   setTokenBudget: (budget: Record<string, unknown> | null) => void;
   setPendingPermissionRequests: Dispatch<SetStateAction<PendingPermissionRequest[]>>;
-  pendingViewSessionRef: MutableRefObject<PendingViewSession | null>;
   streamTimerRef: MutableRefObject<number | null>;
   accumulatedStreamRef: MutableRefObject<string>;
-  onSessionInactive?: (sessionId?: string | null) => void;
-  onSessionActive?: (sessionId?: string | null) => void;
-  onSessionProcessing?: (sessionId?: string | null) => void;
-  onSessionNotProcessing?: (sessionId?: string | null) => void;
-  onNavigateToSession?: (sessionId: string, options?: SessionNavigationOptions) => void;
+  /**
+   * Highest live `seq` observed per session. Essential for reconnect catch-up:
+   * `chat.subscribe` sends this value as `lastSeq` so the server replays only
+   * the events this client actually missed. Written here on every sequenced
+   * frame; read wherever a `chat.subscribe` is sent (session open, reconnect).
+   */
+  lastSeqRef: MutableRefObject<Map<string, number>>;
+  /** When each session's `chat.subscribe` was last sent; guards stale idle acks. */
+  statusCheckSentAtRef: MutableRefObject<Map<string, number>>;
+  onSessionProcessing?: MarkSessionProcessing;
+  onSessionIdle?: MarkSessionIdle;
   onWebSocketReconnect?: () => void;
   sessionStore: SessionStore;
 }
@@ -74,321 +37,259 @@ interface UseChatRealtimeHandlersArgs {
 /*  Hook                                                              */
 /* ------------------------------------------------------------------ */
 
+/**
+ * Routes server events into the session store and processing-state map.
+ *
+ * This is intentionally a thin reducer over the unified `kind`-based
+ * protocol: every frame is keyed by the stable app session id, so there is
+ * no session-id handoff, no provider branching, and no navigation here.
+ * Sidebar events (`session_upserted`, `loading_progress`) are handled by
+ * `useProjectsState`, not in this hook.
+ */
 export function useChatRealtimeHandlers({
-  latestMessage,
+  subscribe,
   provider,
   selectedSession,
   currentSessionId,
-  setCurrentSessionId,
-  setIsLoading,
-  setCanAbortSession,
-  setClaudeStatus,
   setTokenBudget,
   setPendingPermissionRequests,
-  pendingViewSessionRef,
   streamTimerRef,
   accumulatedStreamRef,
-  onSessionInactive,
-  onSessionActive,
+  lastSeqRef,
+  statusCheckSentAtRef,
   onSessionProcessing,
-  onSessionNotProcessing,
-  onNavigateToSession,
+  onSessionIdle,
   onWebSocketReconnect,
   sessionStore,
 }: UseChatRealtimeHandlersArgs) {
-  const paletteOps = usePaletteOps();
-  const lastProcessedMessageRef = useRef<LatestChatMessage | null>(null);
-
   useEffect(() => {
-    if (!latestMessage) return;
-    if (lastProcessedMessageRef.current === latestMessage) return;
-    lastProcessedMessageRef.current = latestMessage;
+    const handleEvent = (msg: ServerEvent) => {
+      if (!msg.kind) {
+        return;
+      }
 
-    const activeViewSessionId =
-      selectedSession?.id || currentSessionId || null;
+      const activeViewSessionId = selectedSession?.id || currentSessionId || null;
+      const sid = (typeof msg.sessionId === 'string' && msg.sessionId) || activeViewSessionId;
 
-    /* ---------------------------------------------------------------- */
-    /*  Legacy messages (no `kind` field) — handle and return           */
-    /* ---------------------------------------------------------------- */
+      // Record replay progress for every sequenced live event.
+      if (sid && typeof msg.seq === 'number') {
+        const known = lastSeqRef.current.get(sid) ?? 0;
+        if (msg.seq > known) {
+          lastSeqRef.current.set(sid, msg.seq);
+        }
+      }
 
-    const msg = latestMessage as any;
-
-    if (!msg.kind) {
-      const messageType = String(msg.type || '');
-
-      switch (messageType) {
-        case 'websocket-reconnected':
+      switch (msg.kind) {
+        case 'websocket_reconnected':
           onWebSocketReconnect?.();
           return;
 
-        case 'pending-permissions-response': {
-          const permSessionId = msg.sessionId;
-          const isCurrentPermSession =
-            permSessionId === currentSessionId || (selectedSession && permSessionId === selectedSession.id);
-          if (permSessionId && !isCurrentPermSession) return;
-          setPendingPermissionRequests(msg.data || []);
-          return;
-        }
-
-        case 'session-status': {
-          const statusSessionId = msg.sessionId;
-          if (!statusSessionId) return;
-
-          const status = msg.status;
-          if (status) {
-            const statusInfo = {
-              text: status.text || 'Working...',
-              tokens: status.tokens || 0,
-              can_interrupt: status.can_interrupt !== undefined ? status.can_interrupt : true,
-            };
-            setClaudeStatus(statusInfo);
-            setIsLoading(true);
-            setCanAbortSession(statusInfo.can_interrupt);
-            return;
-          }
-
-          // Legacy isProcessing format from check-session-status
-          const isCurrentSession =
-            statusSessionId === currentSessionId || (selectedSession && statusSessionId === selectedSession.id);
+        case 'chat_subscribed': {
+          // Ack for chat.subscribe: authoritative processing state plus any
+          // pending tool-permission prompts for the run.
+          if (!sid) return;
 
           if (msg.isProcessing) {
-            onSessionActive?.(statusSessionId);
-            onSessionProcessing?.(statusSessionId);
-            if (isCurrentSession) { setIsLoading(true); setCanAbortSession(true); }
-            return;
+            onSessionProcessing?.(sid);
+          } else {
+            // Idle ack: ignore it if a newer request started after the
+            // subscribe was sent — the ack describes the older state.
+            onSessionIdle?.(sid, {
+              ifStartedBefore: statusCheckSentAtRef.current.get(sid),
+            });
           }
 
-          onSessionInactive?.(statusSessionId);
-          onSessionNotProcessing?.(statusSessionId);
-          if (isCurrentSession) {
-            setIsLoading(false);
-            setCanAbortSession(false);
-            setClaudeStatus(null);
+          const isViewedSession = sid === activeViewSessionId;
+          if (isViewedSession && Array.isArray(msg.pendingPermissions)) {
+            setPendingPermissionRequests(msg.pendingPermissions as PendingPermissionRequest[]);
           }
           return;
         }
 
+        case 'protocol_error': {
+          console.error('[Chat] Protocol error:', msg.code, msg.error);
+          if (sid) {
+            // Surface the failure in the conversation and stop the spinner —
+            // the run never started (or was rejected), so no `complete` follows.
+            onSessionIdle?.(sid);
+            sessionStore.appendRealtime(sid, {
+              id: `protocol_error_${Date.now()}`,
+              sessionId: sid,
+              timestamp: new Date().toISOString(),
+              provider,
+              kind: 'error',
+              content: String(msg.error || 'Request failed'),
+            } as NormalizedMessage);
+          }
+          return;
+        }
+
+        // Sidebar/global events — owned by useProjectsState.
+        case 'session_upserted':
+        case 'loading_progress':
+          return;
+
         default:
-          // Unknown legacy message type — ignore
-          return;
+          break;
       }
-    }
 
-    /* ---------------------------------------------------------------- */
-    /*  NormalizedMessage handling (has `kind` field)                    */
-    /* ---------------------------------------------------------------- */
+      /* -------------------------------------------------------------- */
+      /*  Provider NormalizedMessage handling                            */
+      /* -------------------------------------------------------------- */
 
-    const sid = msg.sessionId || activeViewSessionId;
-
-    // --- Streaming: buffer for performance ---
-    if (msg.kind === 'stream_delta') {
-      const text = msg.content || '';
-      if (!text) return;
-      accumulatedStreamRef.current += text;
-      if (!streamTimerRef.current) {
-        streamTimerRef.current = window.setTimeout(() => {
-          streamTimerRef.current = null;
-          if (sid) {
-            sessionStore.updateStreaming(sid, accumulatedStreamRef.current, provider);
-          }
-        }, 100);
-      }
-      // Also route to store for non-active sessions
-      if (sid && sid !== activeViewSessionId) {
-        sessionStore.appendRealtime(sid, msg as NormalizedMessage);
-      }
-      return;
-    }
-
-    if (msg.kind === 'stream_end') {
-      if (streamTimerRef.current) {
-        clearTimeout(streamTimerRef.current);
-        streamTimerRef.current = null;
-      }
-      if (sid) {
-        if (accumulatedStreamRef.current) {
-          sessionStore.updateStreaming(sid, accumulatedStreamRef.current, provider);
+      // --- Streaming: buffer for performance ---
+      if (msg.kind === 'stream_delta') {
+        const text = (msg.content as string) || '';
+        if (!text) return;
+        accumulatedStreamRef.current += text;
+        if (!streamTimerRef.current) {
+          streamTimerRef.current = window.setTimeout(() => {
+            streamTimerRef.current = null;
+            if (sid) {
+              sessionStore.updateStreaming(sid, accumulatedStreamRef.current, provider);
+            }
+          }, 100);
         }
-        sessionStore.finalizeStreaming(sid);
-      }
-      accumulatedStreamRef.current = '';
-      return;
-    }
-
-    // --- All other messages: route to store ---
-    const shouldPersist =
-      msg.kind !== 'session_created'
-      && msg.kind !== 'complete'
-      && msg.kind !== 'status'
-      && msg.kind !== 'permission_request'
-      && msg.kind !== 'permission_cancelled';
-
-    if (sid && shouldPersist) {
-      sessionStore.appendRealtime(sid, msg as NormalizedMessage);
-    }
-
-    // --- UI side effects for specific kinds ---
-    switch (msg.kind) {
-      case 'session_created': {
-        const newSessionId = msg.newSessionId;
-        if (!newSessionId) break;
-
-        // We no longer synthesize client-side placeholder IDs. Until the provider
-        // announces `session_created`, the active id is expected to be null.
-        if (!currentSessionId) {
-          console.log('Session created with ID:', newSessionId);
-          console.log('Existing session ID:', currentSessionId);
-          setCurrentSessionId(newSessionId);
-          setPendingPermissionRequests((prev) =>
-            prev.map((r) => (r.sessionId ? r : { ...r, sessionId: newSessionId })),
-          );
+        // Also route to store for non-active sessions
+        if (sid && sid !== activeViewSessionId) {
+          sessionStore.appendRealtime(sid, msg as unknown as NormalizedMessage);
         }
-        pendingViewSessionRef.current = null;
-        onSessionActive?.(newSessionId);
-        onSessionProcessing?.(newSessionId);
-        setIsLoading(true);
-        setCanAbortSession(true);
-        setClaudeStatus({
-          text: 'Processing',
-          tokens: 0,
-          can_interrupt: true,
-        });
-        onNavigateToSession?.(newSessionId);
-        break;
+        return;
       }
 
-      case 'complete': {
-        // Flush any remaining streaming state
+      if (msg.kind === 'stream_end') {
         if (streamTimerRef.current) {
           clearTimeout(streamTimerRef.current);
           streamTimerRef.current = null;
         }
-        if (sid && accumulatedStreamRef.current) {
-          sessionStore.updateStreaming(sid, accumulatedStreamRef.current, provider);
+        if (sid) {
+          if (accumulatedStreamRef.current) {
+            sessionStore.updateStreaming(sid, accumulatedStreamRef.current, provider);
+          }
           sessionStore.finalizeStreaming(sid);
         }
         accumulatedStreamRef.current = '';
+        return;
+      }
 
-        setIsLoading(false);
-        setCanAbortSession(false);
-        setClaudeStatus(null);
-        setPendingPermissionRequests([]);
-        onSessionInactive?.(sid);
-        onSessionNotProcessing?.(sid);
-        pendingViewSessionRef.current = null;
+      // --- All other messages: route to store ---
+      const shouldPersist =
+        msg.kind !== 'complete'
+        && msg.kind !== 'status'
+        && msg.kind !== 'permission_request'
+        && msg.kind !== 'permission_cancelled';
+
+      if (sid && shouldPersist) {
+        sessionStore.appendRealtime(sid, msg as unknown as NormalizedMessage);
+      }
+
+      // --- UI side effects for specific kinds ---
+      switch (msg.kind) {
+        case 'complete': {
+          // Flush any remaining streaming state
+          if (streamTimerRef.current) {
+            clearTimeout(streamTimerRef.current);
+            streamTimerRef.current = null;
+          }
+          if (sid && accumulatedStreamRef.current) {
+            sessionStore.updateStreaming(sid, accumulatedStreamRef.current, provider);
+            sessionStore.finalizeStreaming(sid);
+          }
+          accumulatedStreamRef.current = '';
+
+          // `complete` is the unified terminal event — every provider run ends
+          // with exactly one, regardless of success, failure, or abort. The
+          // indicator derives from the processing map, so deleting the entry
+          // hides it immediately and atomically.
+          onSessionIdle?.(sid);
+          setPendingPermissionRequests([]);
+
+          if (msg.aborted) {
+            // Abort was requested — the complete event confirms it. No
+            // further UI action is needed beyond clearing the entry above.
+            break;
+          }
+
+          // Celebrate only successful runs (failed runs end with success: false).
+          if (msg.success !== false) {
+            showCompletionTitleIndicator();
+            void playChatCompletionSound();
+          }
+
+          // The session id is stable for the whole conversation (allocated
+          // before the first send), so the only follow-up is syncing the
+          // viewed conversation with the now-persisted transcript.
+          if (sid && sid === activeViewSessionId) {
+            void sessionStore.refreshFromServer(sid);
+          }
 
-        // Handle aborted case
-        if (msg.aborted) {
-          // Abort was requested — the complete event confirms it
-          // No special UI action needed beyond clearing loading state above
-          // The backend already sent any abort-related messages
           break;
         }
 
-        const actualSessionId =
-          typeof msg.actualSessionId === 'string' && msg.actualSessionId.trim().length > 0
-            ? msg.actualSessionId
-            : null;
-        const isVisibleSession =
-          Boolean(
-            sid
-            && sid === activeViewSessionId,
-          );
+        // 'error' is an informational message row, not a terminal event —
+        // providers emit it for mid-run stderr output too. Run teardown is
+        // always signalled by the unified 'complete' that follows.
 
-        if (actualSessionId && sid && actualSessionId !== sid) {
-          sessionStore.replaceSessionId(sid, actualSessionId);
-
-          if (isVisibleSession) {
-            setCurrentSessionId(actualSessionId);
-          }
-
-          if (isVisibleSession) {
-            onNavigateToSession?.(actualSessionId, { replace: true });
-            setTimeout(() => { void paletteOps.refreshProjects(); }, 500);
-          }
-          break;
-        }
-
-        break;
-      }
-
-      case 'error': {
-        setIsLoading(false);
-        setCanAbortSession(false);
-        setClaudeStatus(null);
-        onSessionInactive?.(sid);
-        onSessionNotProcessing?.(sid);
-        pendingViewSessionRef.current = null;
-        break;
-      }
-
-      case 'permission_request': {
-        if (!msg.requestId) break;
-        setPendingPermissionRequests((prev) => {
-          if (prev.some((r: PendingPermissionRequest) => r.requestId === msg.requestId)) return prev;
-          return [...prev, {
-            requestId: msg.requestId,
-            toolName: msg.toolName || 'UnknownTool',
-            input: msg.input,
-            context: msg.context,
-            sessionId: sid || null,
-            receivedAt: new Date(),
-          }];
-        });
-        setIsLoading(true);
-        setCanAbortSession(true);
-        setClaudeStatus({ text: 'Waiting for permission', tokens: 0, can_interrupt: true });
-        break;
-      }
-
-      case 'permission_cancelled': {
-        if (msg.requestId) {
-          setPendingPermissionRequests((prev) => prev.filter((r: PendingPermissionRequest) => r.requestId !== msg.requestId));
-        }
-        break;
-      }
-
-      case 'status': {
-        if (msg.text === 'token_budget' && msg.tokenBudget) {
-          setTokenBudget(msg.tokenBudget as Record<string, unknown>);
-        } else if (msg.text) {
-          setClaudeStatus({
-            text: msg.text,
-            tokens: msg.tokens || 0,
-            can_interrupt: msg.canInterrupt !== undefined ? msg.canInterrupt : true,
+        case 'permission_request': {
+          if (!msg.requestId) break;
+          setPendingPermissionRequests((prev) => {
+            if (prev.some((r: PendingPermissionRequest) => r.requestId === msg.requestId)) return prev;
+            return [...prev, {
+              requestId: msg.requestId as string,
+              toolName: (msg.toolName as string) || 'UnknownTool',
+              input: msg.input,
+              context: msg.context,
+              sessionId: sid || null,
+              receivedAt: new Date(),
+            }];
           });
-          setIsLoading(true);
-          setCanAbortSession(msg.canInterrupt !== false);
+          if (sid) {
+            onSessionProcessing?.(sid);
+          }
+          break;
         }
-        break;
-      }
 
-      // text, tool_use, tool_result, thinking, interactive_prompt, task_notification
-      // → already routed to store above, no UI side effects needed
-      default:
-        break;
-    }
+        case 'permission_cancelled': {
+          if (msg.requestId) {
+            setPendingPermissionRequests((prev) => prev.filter((r: PendingPermissionRequest) => r.requestId !== msg.requestId));
+          }
+          break;
+        }
+
+        case 'status': {
+          if (msg.text === 'token_budget' && msg.tokenBudget) {
+            setTokenBudget(msg.tokenBudget as Record<string, unknown>);
+          } else if (msg.text && sid) {
+            onSessionProcessing?.(sid, {
+              statusText: msg.text as string,
+              canInterrupt: msg.canInterrupt !== false,
+            });
+          }
+          break;
+        }
+
+        // text, tool_use, tool_result, thinking, interactive_prompt, task_notification
+        // → already routed to store above, no UI side effects needed
+        default:
+          break;
+      }
+    };
+
+    return subscribe(handleEvent);
   }, [
-    latestMessage,
+    subscribe,
     provider,
     selectedSession,
     currentSessionId,
-    setCurrentSessionId,
-    setIsLoading,
-    setCanAbortSession,
-    setClaudeStatus,
     setTokenBudget,
     setPendingPermissionRequests,
-    pendingViewSessionRef,
     streamTimerRef,
     accumulatedStreamRef,
-    onSessionInactive,
-    onSessionActive,
+    lastSeqRef,
+    statusCheckSentAtRef,
     onSessionProcessing,
-    onSessionNotProcessing,
-    onNavigateToSession,
+    onSessionIdle,
     onWebSocketReconnect,
     sessionStore,
-    paletteOps,
   ]);
 }

src/components/chat/hooks/useChatSessionState.ts

@@ -2,6 +2,7 @@ import { useCallback, useEffect, useLayoutEffect, useMemo, useRef, useState } fr
 import type { MutableRefObject } from 'react';
 
 import { authenticatedFetch } from '../../../utils/api';
+import type { MarkSessionIdle, SessionActivityMap } from '../../../hooks/useSessionProtection';
 import type { Project, ProjectSession, LLMProvider } from '../../../types/app';
 import type { SessionStore, NormalizedMessage } from '../../../stores/useSessionStore';
 import type { ChatMessage, Provider } from '../types/types';
@@ -12,10 +13,6 @@ import { normalizedToChatMessages } from './useChatMessages';
 const MESSAGES_PER_PAGE = 20;
 const INITIAL_VISIBLE_MESSAGES = 100;
 
-type PendingViewSession = {
-  startedAt: number;
-};
-
 interface UseChatSessionStateArgs {
   selectedProject: Project | null;
   selectedSession: ProjectSession | null;
@@ -24,9 +21,13 @@ interface UseChatSessionStateArgs {
   autoScrollToBottom?: boolean;
   externalMessageUpdate?: number;
   newSessionTrigger?: number;
-  processingSessions?: Set<string>;
+  processingSessions?: SessionActivityMap;
+  onSessionIdle?: MarkSessionIdle;
   resetStreamingState: () => void;
-  pendingViewSessionRef: MutableRefObject<PendingViewSession | null>;
+  /** When each session's `chat.subscribe` was last sent; guards stale idle acks. */
+  statusCheckSentAtRef: MutableRefObject<Map<string, number>>;
+  /** Highest live seq observed per session; sent as `lastSeq` on subscribe. */
+  lastSeqRef: MutableRefObject<Map<string, number>>;
   sessionStore: SessionStore;
 }
 
@@ -99,21 +100,20 @@ export function useChatSessionState({
   externalMessageUpdate,
   newSessionTrigger,
   processingSessions,
+  onSessionIdle,
   resetStreamingState,
-  pendingViewSessionRef,
+  statusCheckSentAtRef,
+  lastSeqRef,
   sessionStore,
 }: UseChatSessionStateArgs) {
-  const [isLoading, setIsLoading] = useState(false);
   const [currentSessionId, setCurrentSessionId] = useState<string | null>(selectedSession?.id || null);
   const [isLoadingSessionMessages, setIsLoadingSessionMessages] = useState(false);
   const [isLoadingMoreMessages, setIsLoadingMoreMessages] = useState(false);
   const [hasMoreMessages, setHasMoreMessages] = useState(false);
   const [totalMessages, setTotalMessages] = useState(0);
-  const [canAbortSession, setCanAbortSession] = useState(false);
   const [isUserScrolledUp, setIsUserScrolledUp] = useState(false);
   const [tokenBudget, setTokenBudget] = useState<Record<string, unknown> | null>(null);
   const [visibleMessageCount, setVisibleMessageCount] = useState(INITIAL_VISIBLE_MESSAGES);
-  const [claudeStatus, setClaudeStatus] = useState<{ text: string; tokens: number; can_interrupt: boolean } | null>(null);
   const [allMessagesLoaded, setAllMessagesLoaded] = useState(false);
   const [isLoadingAllMessages, setIsLoadingAllMessages] = useState(false);
   const [loadAllJustFinished, setLoadAllJustFinished] = useState(false);
@@ -170,13 +170,8 @@ export function useChatSessionState({
      * - No coupling to unrelated external update signals.
      */
     resetStreamingState();
-    pendingViewSessionRef.current = null;
-    setClaudeStatus(null);
-    setCanAbortSession(false);
-    setIsLoading(false);
     setCurrentSessionId(null);
     setPendingUserMessage(null);
-    sessionStorage.removeItem('cursorSessionId');
     messagesOffsetRef.current = 0;
     setHasMoreMessages(false);
     setTotalMessages(0);
@@ -204,13 +199,30 @@ export function useChatSessionState({
       clearTimeout(loadAllFinishedTimerRef.current);
       loadAllFinishedTimerRef.current = null;
     }
-  }, [newSessionTrigger, pendingViewSessionRef, resetStreamingState]);
+  }, [newSessionTrigger, onSessionIdle, resetStreamingState]);
+
+  /* ---------------------------------------------------------------- */
+  /*  Derive processing state for the viewed session                  */
+  /* ---------------------------------------------------------------- */
+
+  const activeSessionId = selectedSession?.id || currentSessionId || null;
+
+  // The activity indicator always reflects the latest status of the session
+  // being viewed — never stale local UI state from the last time it was
+  // open. Session ids are concrete before any send, so no pending
+  // placeholder entry exists anymore.
+  const sessionActivity = (activeSessionId && processingSessions?.get(activeSessionId)) || null;
+  const isProcessing = sessionActivity !== null;
+  const canAbortSession = isProcessing && sessionActivity.canInterrupt;
+
+  // Ref mirror so effects can read the latest map without re-running on
+  // every activity transition.
+  const processingSessionsRef = useRef(processingSessions);
+  processingSessionsRef.current = processingSessions;
 
   /* ---------------------------------------------------------------- */
   /*  Derive chatMessages from the store                              */
   /* ---------------------------------------------------------------- */
-
-  const activeSessionId = selectedSession?.id || currentSessionId || null;
   const [pendingUserMessage, setPendingUserMessage] = useState<ChatMessage | null>(null);
   const flushedPendingUserMessageRef = useRef<ChatMessage | null>(null);
 
@@ -383,30 +395,61 @@ export function useChatSessionState({
     setIsUserScrolledUp(false);
   }, [selectedProject?.projectId, selectedSession?.id]);
 
-  // Initial scroll to bottom
+  // Initial scroll to bottom — robust to lazy content reflow.
+  // The previous implementation fired one scrollToBottom() at +200ms and
+  // cleared the pending flag. When markdown blocks, code highlighting, or
+  // images finished rendering after that window, scrollHeight grew but
+  // nothing re-anchored the viewport, leaving the chat tab visually
+  // "scrolled way up" with the latest assistant message off-screen.
+  //
+  // This version re-scrolls every animation frame while scrollHeight is
+  // still growing, capped at ~1s (60 frames) or 3 consecutive stable
+  // frames. Cancels cleanly on session change via the pending flag.
   useEffect(() => {
     if (!pendingInitialScrollRef.current || !scrollContainerRef.current || isLoadingSessionMessages) return;
     if (chatMessages.length === 0) { pendingInitialScrollRef.current = false; return; }
-    pendingInitialScrollRef.current = false;
-    if (!searchScrollActiveRef.current) setTimeout(() => scrollToBottom(), 200);
+    if (searchScrollActiveRef.current) { pendingInitialScrollRef.current = false; return; }
+
+    const container = scrollContainerRef.current;
+    let frame = 0;
+    let lastHeight = 0;
+    let stableCount = 0;
+    let rafId = 0;
+
+    const tick = () => {
+      if (!pendingInitialScrollRef.current || !scrollContainerRef.current) return;
+      container.scrollTop = container.scrollHeight;
+      if (container.scrollHeight === lastHeight) {
+        stableCount++;
+      } else {
+        stableCount = 0;
+        lastHeight = container.scrollHeight;
+      }
+      frame++;
+      if (stableCount < 3 && frame < 60) {
+        rafId = requestAnimationFrame(tick);
+      } else {
+        pendingInitialScrollRef.current = false;
+      }
+    };
+    rafId = requestAnimationFrame(tick);
+    return () => {
+      if (rafId) cancelAnimationFrame(rafId);
+    };
   }, [chatMessages.length, isLoadingSessionMessages, scrollToBottom]);
 
   // Main session loading effect — store-based
   useEffect(() => {
     if (!selectedSession || !selectedProject) {
-      // A new provider run can be in flight before the router has a canonical
-      // selectedSession. Keep the processing banner alive until complete/error.
-      if (pendingViewSessionRef.current) {
+      // A freshly created session can be mid-run before the router has a
+      // canonical selectedSession (the URL effect synthesizes one on the
+      // next render). Keep the active view intact instead of wiping it.
+      if (currentSessionId && processingSessionsRef.current?.has(currentSessionId)) {
         return;
       }
 
       resetStreamingState();
-      pendingViewSessionRef.current = null;
-      setClaudeStatus(null);
-      setCanAbortSession(false);
-      setIsLoading(false);
       setCurrentSessionId(null);
-      sessionStorage.removeItem('cursorSessionId');
       messagesOffsetRef.current = 0;
       setHasMoreMessages(false);
       setTotalMessages(0);
@@ -426,9 +469,6 @@ export function useChatSessionState({
     const sessionChanged = currentSessionId !== null && currentSessionId !== selectedSession.id;
     if (sessionChanged) {
       resetStreamingState();
-      pendingViewSessionRef.current = null;
-      setClaudeStatus(null);
-      setCanAbortSession(false);
     }
 
     // Reset pagination/scroll state
@@ -447,17 +487,24 @@ export function useChatSessionState({
 
     if (sessionChanged) {
       setTokenBudget(null);
-      setIsLoading(false);
     }
 
     setCurrentSessionId(selectedSession.id);
-    if (provider === 'cursor') {
-      sessionStorage.setItem('cursorSessionId', selectedSession.id);
-    }
 
-    // Check session status
+    // Subscribe to the session's live run (if any): the ack reconciles the
+    // processing indicator, re-attaches a mid-flight stream to this socket,
+    // and replays any live events missed since `lastSeq`. Recording the send
+    // time lets the ack handler discard idle acks that a newer request has
+    // since outdated.
     if (ws) {
-      sendMessage({ type: 'check-session-status', sessionId: selectedSession.id, provider });
+      statusCheckSentAtRef.current.set(selectedSession.id, Date.now());
+      sendMessage({
+        type: 'chat.subscribe',
+        sessions: [{
+          sessionId: selectedSession.id,
+          lastSeq: lastSeqRef.current.get(selectedSession.id) ?? 0,
+        }],
+      });
     }
 
     lastLoadedSessionKeyRef.current = sessionKey;
@@ -481,11 +528,12 @@ export function useChatSessionState({
       setIsLoadingSessionMessages(false);
     });
   }, [
-    pendingViewSessionRef,
     resetStreamingState,
     selectedProject,
     selectedSession?.id,
     sendMessage,
+    statusCheckSentAtRef,
+    lastSeqRef,
     ws,
     sessionStore,
   ]);
@@ -499,7 +547,7 @@ export function useChatSessionState({
         const provider = (localStorage.getItem('selected-provider') as Provider) || 'claude';
 
         // Skip store refresh during active streaming
-        if (!isLoading) {
+        if (!isProcessing) {
           await sessionStore.refreshFromServer(selectedSession.id, {
             provider: (selectedSession.__provider || provider) as LLMProvider,
             projectId: selectedProject.projectId,
@@ -524,7 +572,7 @@ export function useChatSessionState({
     selectedProject,
     selectedSession,
     sessionStore,
-    isLoading,
+    isProcessing,
   ]);
 
   // Search navigation target
@@ -691,16 +739,6 @@ export function useChatSessionState({
     return () => container.removeEventListener('scroll', handleScroll);
   }, [handleScroll]);
 
-  useEffect(() => {
-    const activeViewSessionId = selectedSession?.id || currentSessionId;
-    if (!activeViewSessionId || !processingSessions) return;
-    const shouldBeProcessing = processingSessions.has(activeViewSessionId);
-    if (shouldBeProcessing && !isLoading) {
-      setIsLoading(true);
-      setCanAbortSession(true);
-    }
-  }, [currentSessionId, isLoading, processingSessions, selectedSession?.id]);
-
   // "Load all" overlay
   const prevLoadingRef = useRef(false);
   useEffect(() => {
@@ -782,16 +820,15 @@ export function useChatSessionState({
     addMessage,
     clearMessages,
     rewindMessages,
-    isLoading,
-    setIsLoading,
+    sessionActivity,
+    isProcessing,
+    canAbortSession,
     currentSessionId,
     setCurrentSessionId,
     isLoadingSessionMessages,
     isLoadingMoreMessages,
     hasMoreMessages,
     totalMessages,
-    canAbortSession,
-    setCanAbortSession,
     isUserScrolledUp,
     setIsUserScrolledUp,
     tokenBudget,
@@ -804,8 +841,6 @@ export function useChatSessionState({
     isLoadingAllMessages,
     loadAllJustFinished,
     showLoadAllOverlay,
-    claudeStatus,
-    setClaudeStatus,
     createDiff,
     scrollContainerRef,
     scrollToBottom,

src/components/chat/hooks/useSlashCommands.ts

@@ -393,7 +393,8 @@ export function useSlashCommands({
         return;
       }
 
-      const slashPattern = /^\/(\S*)$/;
+      // Match / at start of input OR after whitespace, capturing the /word up to cursor.
+      const slashPattern = /(?:^|\s)(\/\S*)$/;
       const match = textBeforeCursor.match(slashPattern);
 
       if (!match) {
@@ -401,8 +402,9 @@ export function useSlashCommands({
         return;
       }
 
-      const slashPos = 0;
-      const query = match[1];
+      // Compute actual position of / in the full input string.
+      const slashPos = match.index! + (match[0].length - match[1].length);
+      const query = match[1].slice(1); // strip leading /
 
       setSlashPosition(slashPos);
       setShowCommandMenu(true);

src/components/chat/tools/configs/toolConfigs.ts

@@ -564,11 +564,15 @@ export function shouldHideToolResult(toolName: string, toolResult: any): boolean
 
   if (!config.result) return false;
 
+  // Hidden/success-only configs suppress noisy successful output, but errors
+  // still need to be visible so failed tool calls are diagnosable.
+  if (toolResult?.isError) return false;
+
   // Always hidden
   if (config.result.hidden) return true;
 
   // Hide on success only
-  if (config.result.hideOnSuccess && toolResult && !toolResult.isError) {
+  if (config.result.hideOnSuccess && toolResult) {
     return true;
   }
 

src/components/chat/types/types.ts

@@ -1,4 +1,9 @@
 import type { Project, ProjectSession, LLMProvider } from '../../../types/app';
+import type {
+  MarkSessionIdle,
+  MarkSessionProcessing,
+  SessionActivityMap,
+} from '../../../hooks/useSessionProtection';
 
 export type Provider = LLMProvider;
 
@@ -102,20 +107,24 @@ export type SessionNavigationOptions = {
   replace?: boolean;
 };
 
+export type SessionEstablishedContext = {
+  provider: LLMProvider;
+  project: Project;
+  summary?: string | null;
+};
+
 export interface ChatInterfaceProps {
   selectedProject: Project | null;
   selectedSession: ProjectSession | null;
   ws: WebSocket | null;
   sendMessage: (message: unknown) => void;
-  latestMessage: any;
   onFileOpen?: (filePath: string, diffInfo?: any) => void;
   onInputFocusChange?: (focused: boolean) => void;
-  onSessionActive?: (sessionId?: string | null) => void;
-  onSessionInactive?: (sessionId?: string | null) => void;
-  onSessionProcessing?: (sessionId?: string | null) => void;
-  onSessionNotProcessing?: (sessionId?: string | null) => void;
-  processingSessions?: Set<string>;
+  onSessionProcessing?: MarkSessionProcessing;
+  onSessionIdle?: MarkSessionIdle;
+  processingSessions?: SessionActivityMap;
   onNavigateToSession?: (targetSessionId: string, options?: SessionNavigationOptions) => void;
+  onSessionEstablished?: (sessionId: string, context: SessionEstablishedContext) => void;
   onShowSettings?: () => void;
   autoExpandTools?: boolean;
   showRawParameters?: boolean;

src/components/chat/view/ChatInterface.tsx

@@ -2,6 +2,7 @@ import React, { useCallback, useEffect, useMemo, useRef } from 'react';
 import { useTranslation } from 'react-i18next';
 
 import { useTasksSettings } from '../../../contexts/TasksSettingsContext';
+import { useWebSocket } from '../../../contexts/WebSocketContext';
 import PermissionContext from '../../../contexts/PermissionContext';
 import { QuickSettingsPanel } from '../../quick-settings-panel';
 import type { ChatInterfaceProps, Provider  } from '../types/types';
@@ -17,24 +18,18 @@ import ChatComposer from './subcomponents/ChatComposer';
 import CommandResultModal from './subcomponents/CommandResultModal';
 
 
-type PendingViewSession = {
-  startedAt: number;
-};
-
 function ChatInterface({
   selectedProject,
   selectedSession,
   ws,
   sendMessage,
-  latestMessage,
   onFileOpen,
   onInputFocusChange,
-  onSessionActive,
-  onSessionInactive,
   onSessionProcessing,
-  onSessionNotProcessing,
+  onSessionIdle,
   processingSessions,
   onNavigateToSession,
+  onSessionEstablished,
   onShowSettings,
   autoExpandTools,
   showRawParameters,
@@ -46,12 +41,19 @@ function ChatInterface({
   onShowAllTasks,
 }: ChatInterfaceProps) {
   const { tasksEnabled, isTaskMasterInstalled } = useTasksSettings();
+  const { subscribe } = useWebSocket();
   const { t } = useTranslation('chat');
 
   const sessionStore = useSessionStore();
   const streamTimerRef = useRef<number | null>(null);
   const accumulatedStreamRef = useRef('');
-  const pendingViewSessionRef = useRef<PendingViewSession | null>(null);
+  // When each session's `chat.subscribe` was last sent; idle acks older than
+  // a later local request are discarded as stale.
+  const statusCheckSentAtRef = useRef(new Map<string, number>());
+  // Highest live `seq` observed per session. Written by the realtime handler
+  // on every sequenced frame, read whenever a `chat.subscribe` is sent so the
+  // server replays only the events this client actually missed.
+  const lastSeqRef = useRef(new Map<string, number>());
 
   const resetStreamingState = useCallback(() => {
     if (streamTimerRef.current) {
@@ -92,16 +94,15 @@ function ChatInterface({
   const {
     chatMessages,
     addMessage,
-    isLoading,
-    setIsLoading,
+    sessionActivity,
+    isProcessing,
+    canAbortSession,
     currentSessionId,
     setCurrentSessionId,
     isLoadingSessionMessages,
     isLoadingMoreMessages,
     hasMoreMessages,
     totalMessages,
-    canAbortSession,
-    setCanAbortSession,
     isUserScrolledUp,
     setIsUserScrolledUp,
     tokenBudget,
@@ -114,8 +115,6 @@ function ChatInterface({
     isLoadingAllMessages,
     loadAllJustFinished,
     showLoadAllOverlay,
-    claudeStatus,
-    setClaudeStatus,
     createDiff,
     scrollContainerRef,
     scrollToBottom,
@@ -130,19 +129,28 @@ function ChatInterface({
     externalMessageUpdate,
     newSessionTrigger,
     processingSessions,
+    onSessionIdle,
     resetStreamingState,
-    pendingViewSessionRef,
+    statusCheckSentAtRef,
+    lastSeqRef,
     sessionStore,
   });
 
+  // Brand-new conversation: the composer allocated a stable session id via
+  // the session gateway before the first send. Record it locally and put it
+  // in the URL — this id never changes again, so there is no later handoff.
+  const handleSessionEstablished = useCallback<NonNullable<ChatInterfaceProps['onSessionEstablished']>>((sessionId, context) => {
+    setCurrentSessionId(sessionId);
+    onSessionEstablished?.(sessionId, context);
+    onNavigateToSession?.(sessionId);
+  }, [setCurrentSessionId, onSessionEstablished, onNavigateToSession]);
+
   const {
     input,
     setInput,
     textareaRef,
     inputHighlightRef,
     isTextareaExpanded,
-    thinkingMode,
-    setThinkingMode,
     slashCommandsCount,
     filteredCommands,
     frequentCommands,
@@ -180,6 +188,7 @@ function ChatInterface({
     isInputFocused: _isInputFocused,
     commandModalPayload,
     closeCommandModal,
+    showCostModal,
   } = useChatComposerState({
     selectedProject,
     selectedSession,
@@ -192,66 +201,67 @@ function ChatInterface({
     codexModel,
     geminiModel,
     opencodeModel,
-    isLoading,
+    isLoading: isProcessing,
     canAbortSession,
     tokenBudget,
     sendMessage,
     sendByCtrlEnter,
-    onSessionActive,
     onSessionProcessing,
+    onSessionEstablished: handleSessionEstablished,
     onInputFocusChange,
     onFileOpen,
     onShowSettings,
-    pendingViewSessionRef,
     scrollToBottom,
     addMessage,
-    setIsLoading,
-    setCanAbortSession,
-    setClaudeStatus,
     setIsUserScrolledUp,
     setPendingPermissionRequests,
   });
 
-  // On WebSocket reconnect, re-fetch the current session's messages from the server
-  // so missed streaming events are shown. Also reset isLoading.
+  // On WebSocket reconnect, re-fetch the current session's messages from the
+  // server so missed streaming events are shown, then re-subscribe — the
+  // `chat_subscribed` ack restores or clears the activity indicator, replays
+  // missed live events, and re-attaches a still-running stream to this socket.
   const handleWebSocketReconnect = useCallback(async () => {
     if (!selectedProject || !selectedSession) return;
-    const providerVal = (localStorage.getItem('selected-provider') as LLMProvider) || 'claude';
+    const providerVal =
+      selectedSession.__provider
+      || (localStorage.getItem('selected-provider') as LLMProvider)
+      || 'claude';
     await sessionStore.refreshFromServer(selectedSession.id, {
-      provider: (selectedSession.__provider || providerVal) as LLMProvider,
+      provider: providerVal as LLMProvider,
       // Use DB projectId; legacy folder-derived projectName is no longer accepted here.
       projectId: selectedProject.projectId,
       projectPath: selectedProject.fullPath || selectedProject.path || '',
     });
-    setIsLoading(false);
-    setCanAbortSession(false);
-  }, [selectedProject, selectedSession, sessionStore, setIsLoading, setCanAbortSession]);
+    statusCheckSentAtRef.current.set(selectedSession.id, Date.now());
+    sendMessage({
+      type: 'chat.subscribe',
+      sessions: [{
+        sessionId: selectedSession.id,
+        lastSeq: lastSeqRef.current.get(selectedSession.id) ?? 0,
+      }],
+    });
+  }, [selectedProject, selectedSession, sendMessage, sessionStore]);
 
   useChatRealtimeHandlers({
-    latestMessage,
+    subscribe,
     provider,
     selectedSession,
     currentSessionId,
-    setCurrentSessionId,
-    setIsLoading,
-    setCanAbortSession,
-    setClaudeStatus,
     setTokenBudget,
     setPendingPermissionRequests,
-    pendingViewSessionRef,
     streamTimerRef,
     accumulatedStreamRef,
-    onSessionInactive,
-    onSessionActive,
+    lastSeqRef,
+    statusCheckSentAtRef,
     onSessionProcessing,
-    onSessionNotProcessing,
-    onNavigateToSession,
+    onSessionIdle,
     onWebSocketReconnect: handleWebSocketReconnect,
     sessionStore,
   });
 
   useEffect(() => {
-    if (!isLoading || !canAbortSession) {
+    if (!canAbortSession) {
       return;
     }
 
@@ -268,7 +278,7 @@ function ChatInterface({
     return () => {
       document.removeEventListener('keydown', handleGlobalEscape, { capture: true });
     };
-  }, [canAbortSession, handleAbortSession, isLoading]);
+  }, [canAbortSession, handleAbortSession]);
 
   useEffect(() => {
     return () => {
@@ -315,6 +325,7 @@ function ChatInterface({
           onWheel={handleScroll}
           onTouchMove={handleScroll}
           isLoadingSessionMessages={isLoadingSessionMessages}
+          isProcessing={isProcessing}
           chatMessages={chatMessages}
           selectedSession={selectedSession}
           currentSessionId={currentSessionId}
@@ -363,15 +374,13 @@ function ChatInterface({
           pendingPermissionRequests={pendingPermissionRequests}
           handlePermissionDecision={handlePermissionDecision}
           handleGrantToolPermission={handleGrantToolPermission}
-          claudeStatus={claudeStatus}
-          isLoading={isLoading}
+          activity={sessionActivity}
+          isLoading={isProcessing}
           onAbortSession={handleAbortSession}
-          provider={provider}
           permissionMode={permissionMode}
           onModeSwitch={cyclePermissionMode}
-          thinkingMode={thinkingMode}
-          setThinkingMode={setThinkingMode}
           tokenBudget={tokenBudget}
+          onShowTokenUsage={showCostModal}
           slashCommandsCount={slashCommandsCount}
           onToggleCommandMenu={handleToggleCommandMenu}
           hasInput={Boolean(input.trim())}

src/components/chat/view/subcomponents/ActivityIndicator.tsx

@@ -0,0 +1,80 @@
+import { useEffect, useState } from 'react';
+import { useTranslation } from 'react-i18next';
+
+import { Shimmer } from '../../../../shared/view/ui';
+import type { SessionActivity } from '../../../../hooks/useSessionProtection';
+
+type ActivityIndicatorProps = {
+  activity: SessionActivity | null;
+  onAbort?: () => void;
+};
+
+const ACTION_KEYS = [
+  'claudeStatus.actions.thinking',
+  'claudeStatus.actions.processing',
+  'claudeStatus.actions.analyzing',
+  'claudeStatus.actions.working',
+  'claudeStatus.actions.computing',
+  'claudeStatus.actions.reasoning',
+];
+const DEFAULT_ACTION_WORDS = ['Thinking', 'Processing', 'Analyzing', 'Working', 'Computing', 'Reasoning'];
+
+/**
+ * Minimal response-in-progress indicator, in the spirit of the inline status
+ * lines in Claude Code / Codex / OpenCode: a shimmering activity label, the
+ * elapsed time, and an interrupt affordance. Rendered only while the viewed
+ * session has an entry in the processing map; it disappears the instant that
+ * entry is removed.
+ */
+export default function ActivityIndicator({ activity, onAbort }: ActivityIndicatorProps) {
+  const { t } = useTranslation('chat');
+  const startedAt = activity?.startedAt ?? null;
+  const [elapsedSeconds, setElapsedSeconds] = useState(0);
+
+  useEffect(() => {
+    if (startedAt === null) return;
+    const update = () => setElapsedSeconds(Math.max(0, Math.floor((Date.now() - startedAt) / 1000)));
+    update();
+    const timer = setInterval(update, 1000);
+    return () => clearInterval(timer);
+  }, [startedAt]);
+
+  if (!activity) return null;
+
+  const actionWords = ACTION_KEYS.map((key, i) => t(key, { defaultValue: DEFAULT_ACTION_WORDS[i] }));
+  const label = (activity.statusText || actionWords[Math.floor(elapsedSeconds / 4) % actionWords.length])
+    .replace(/\.+$/, '');
+
+  const minutes = Math.floor(elapsedSeconds / 60);
+  const seconds = elapsedSeconds % 60;
+  const elapsedLabel = minutes < 1
+    ? t('claudeStatus.elapsed.seconds', { count: seconds, defaultValue: '{{count}}s' })
+    : t('claudeStatus.elapsed.minutesSeconds', { minutes, seconds, defaultValue: '{{minutes}}m {{seconds}}s' });
+
+  return (
+    <div className="animate-in fade-in mb-2 w-full duration-300">
+      <div className="mx-auto flex max-w-4xl items-center gap-2 px-1">
+        <span className="h-1.5 w-1.5 shrink-0 animate-pulse rounded-full bg-primary" aria-hidden />
+        <Shimmer className="text-xs font-medium">{`${label}…`}</Shimmer>
+        <span className="text-xs tabular-nums text-muted-foreground/60">{elapsedLabel}</span>
+
+        {activity.canInterrupt && onAbort && (
+          <button
+            type="button"
+            onClick={onAbort}
+            className="ml-auto flex items-center gap-1.5 rounded-md px-2 py-0.5 text-xs text-muted-foreground transition-colors hover:bg-destructive/10 hover:text-destructive"
+            aria-label={t('claudeStatus.stop', { defaultValue: 'Stop' })}
+          >
+            <svg className="h-2.5 w-2.5 fill-current" viewBox="0 0 24 24" aria-hidden>
+              <rect x="5" y="5" width="14" height="14" rx="2" />
+            </svg>
+            <span>{t('claudeStatus.stop', { defaultValue: 'Stop' })}</span>
+            <kbd className="hidden rounded border border-border/60 px-1 text-[10px] text-muted-foreground/70 sm:inline-block">
+              esc
+            </kbd>
+          </button>
+        )}
+      </div>
+    </div>
+  );
+}

src/components/chat/view/subcomponents/ChatComposer.tsx

@@ -2,23 +2,17 @@ import { useTranslation } from 'react-i18next';
 import type {
   ChangeEvent,
   ClipboardEvent,
-  Dispatch,
   FormEvent,
   KeyboardEvent,
   MouseEvent,
   ReactNode,
   RefObject,
-  SetStateAction,
   TouchEvent,
 } from 'react';
 import { ImageIcon, MessageSquareIcon, XIcon, ArrowDownIcon } from 'lucide-react';
-import type { PendingPermissionRequest, PermissionMode, Provider } from '../../types/types';
-import CommandMenu from './CommandMenu';
-import ClaudeStatus from './ClaudeStatus';
-import ImageAttachment from './ImageAttachment';
-import PermissionRequestsBanner from './PermissionRequestsBanner';
-import ThinkingModeSelector from './ThinkingModeSelector';
-import TokenUsageSummary from './TokenUsageSummary';
+
+import type { SessionActivity } from '../../../../hooks/useSessionProtection';
+import type { PendingPermissionRequest, PermissionMode } from '../../types/types';
 import {
   PromptInput,
   PromptInputHeader,
@@ -30,6 +24,12 @@ import {
   PromptInputSubmit,
 } from '../../../../shared/view/ui';
 
+import CommandMenu from './CommandMenu';
+import ActivityIndicator from './ActivityIndicator';
+import ImageAttachment from './ImageAttachment';
+import PermissionRequestsBanner from './PermissionRequestsBanner';
+import TokenUsageSummary from './TokenUsageSummary';
+
 interface MentionableFile {
   name: string;
   path: string;
@@ -52,15 +52,13 @@ interface ChatComposerProps {
     decision: { allow?: boolean; message?: string; rememberEntry?: string | null; updatedInput?: unknown },
   ) => void;
   handleGrantToolPermission: (suggestion: { entry: string; toolName: string }) => { success: boolean };
-  claudeStatus: { text: string; tokens: number; can_interrupt: boolean } | null;
+  activity: SessionActivity | null;
   isLoading: boolean;
   onAbortSession: () => void;
-  provider: Provider | string;
   permissionMode: PermissionMode | string;
   onModeSwitch: () => void;
-  thinkingMode: string;
-  setThinkingMode: Dispatch<SetStateAction<string>>;
   tokenBudget: Record<string, unknown> | null;
+  onShowTokenUsage: () => void;
   slashCommandsCount: number;
   onToggleCommandMenu: () => void;
   hasInput: boolean;
@@ -107,15 +105,13 @@ export default function ChatComposer({
   pendingPermissionRequests,
   handlePermissionDecision,
   handleGrantToolPermission,
-  claudeStatus,
+  activity,
   isLoading,
   onAbortSession,
-  provider,
   permissionMode,
   onModeSwitch,
-  thinkingMode,
-  setThinkingMode,
   tokenBudget,
+  onShowTokenUsage,
   slashCommandsCount,
   onToggleCommandMenu,
   hasInput,
@@ -176,12 +172,7 @@ export default function ChatComposer({
   return (
     <div className="flex-shrink-0 p-2 pb-2 sm:p-4 sm:pb-4 md:p-4 md:pb-6">
       {!hasPendingPermissions && (
-        <ClaudeStatus
-          status={claudeStatus}
-          isLoading={isLoading}
-          onAbort={onAbortSession}
-          provider={provider}
-        />
+        <ActivityIndicator activity={activity} onAbort={onAbortSession} />
       )}
 
       {pendingPermissionRequests.length > 0 && (
@@ -295,6 +286,7 @@ export default function ChatComposer({
 
             <PromptInputTextarea
               ref={textareaRef}
+              dir="auto"
               value={input}
               onChange={onInputChange}
               onClick={onTextareaClick}
@@ -357,11 +349,7 @@ export default function ChatComposer({
               </div>
             </button>
 
-            {provider === 'claude' && (
-              <ThinkingModeSelector selectedMode={thinkingMode} onModeChange={setThinkingMode} onClose={() => {}} className="" />
-            )}
-
-            <TokenUsageSummary usage={tokenBudget} />
+            <TokenUsageSummary usage={tokenBudget} onClick={onShowTokenUsage} />
 
             <PromptInputButton
               tooltip={{ content: t('input.showAllCommands') }}
@@ -382,7 +370,7 @@ export default function ChatComposer({
               <PromptInputButton
                 tooltip={{ content: t('input.clearInput', { defaultValue: 'Clear input' }) }}
                 onClick={onClearInput}
-                className="hidden sm:No-flex"
+                className="hidden sm:flex"
               >
                 <XIcon />
               </PromptInputButton>
@@ -399,7 +387,8 @@ export default function ChatComposer({
               {sendByCtrlEnter ? t('input.hintText.ctrlEnter') : t('input.hintText.enter')}
             </div>
             <PromptInputSubmit
-              disabled={!input.trim() || isLoading}
+              onClick={isLoading ? onAbortSession : undefined}
+              disabled={!isLoading && !input.trim()}
               className="h-10 w-10 sm:h-10 sm:w-10"
             />
           </div>

src/components/chat/view/subcomponents/ChatMessagesPane.tsx

@@ -19,6 +19,8 @@ interface ChatMessagesPaneProps {
   onWheel: () => void;
   onTouchMove: () => void;
   isLoadingSessionMessages: boolean;
+  /** True while the viewed session has an active provider run in flight. */
+  isProcessing?: boolean;
   chatMessages: ChatMessage[];
   selectedSession: ProjectSession | null;
   currentSessionId: string | null;
@@ -68,6 +70,7 @@ export default function ChatMessagesPane({
   onWheel,
   onTouchMove,
   isLoadingSessionMessages,
+  isProcessing = false,
   chatMessages,
   selectedSession,
   currentSessionId,
@@ -147,7 +150,7 @@ export default function ChatMessagesPane({
       onTouchMove={onTouchMove}
       className="relative flex-1 space-y-3 overflow-y-auto overflow-x-hidden px-0 py-3 sm:space-y-4 sm:p-4"
     >
-      {isLoadingSessionMessages && chatMessages.length === 0 ? (
+      {(isLoadingSessionMessages || isProcessing) && chatMessages.length === 0 ? (
         <div className="mt-8 text-center text-gray-500 dark:text-gray-400">
           <div className="flex items-center justify-center space-x-2">
             <div className="h-4 w-4 animate-spin rounded-full border-b-2 border-gray-400" />

src/components/chat/view/subcomponents/ClaudeStatus.tsx

@@ -1,130 +0,0 @@
-import { useEffect, useState } from 'react';
-import { useTranslation } from 'react-i18next';
-import { cn } from '../../../../lib/utils';
-import SessionProviderLogo from '../../../llm-logo-provider/SessionProviderLogo';
-
-type ClaudeStatusProps = {
-  status: {
-    text?: string;
-    tokens?: number;
-    can_interrupt?: boolean;
-  } | null;
-  onAbort?: () => void;
-  isLoading: boolean;
-  provider?: string;
-};
-
-const ACTION_KEYS = [
-  'claudeStatus.actions.thinking',
-  'claudeStatus.actions.processing',
-  'claudeStatus.actions.analyzing',
-  'claudeStatus.actions.working',
-  'claudeStatus.actions.computing',
-  'claudeStatus.actions.reasoning',
-];
-const DEFAULT_ACTION_WORDS = ['Thinking', 'Processing', 'Analyzing', 'Working', 'Computing', 'Reasoning'];
-
-const PROVIDER_LABEL_KEYS: Record<string, string> = {
-  claude: 'messageTypes.claude',
-  codex: 'messageTypes.codex',
-  cursor: 'messageTypes.cursor',
-  gemini: 'messageTypes.gemini',
-  opencode: 'messageTypes.opencode',
-};
-
-function formatElapsedTime(totalSeconds: number) {
-  const mins = Math.floor(totalSeconds / 60);
-  const secs = totalSeconds % 60;
-  return mins < 1 ? `${secs}s` : `${mins}m ${secs}s`;
-}
-
-export default function ClaudeStatus({
-  status,
-  onAbort,
-  isLoading,
-  provider = 'claude',
-}: ClaudeStatusProps) {
-  const { t } = useTranslation('chat');
-  const [elapsedTime, setElapsedTime] = useState(0);
-  const [dots, setDots] = useState('');
-
-  useEffect(() => {
-    if (!isLoading) {
-      setElapsedTime(0);
-      return;
-    }
-    const startTime = Date.now();
-    const timer = setInterval(() => {
-      setElapsedTime(Math.floor((Date.now() - startTime) / 1000));
-    }, 1000);
-    const dotTimer = setInterval(() => {
-      setDots((prev) => (prev.length >= 3 ? '' : prev + '.'));
-    }, 500);
-
-    return () => {
-      clearInterval(timer);
-      clearInterval(dotTimer);
-    };
-  }, [isLoading]);
-
-  if (!isLoading && !status) return null;
-
-  const actionWords = ACTION_KEYS.map((key, i) => t(key, { defaultValue: DEFAULT_ACTION_WORDS[i] }));
-  const statusText = (status?.text || actionWords[Math.floor(elapsedTime / 3) % actionWords.length]).replace(/[.]+$/, '');
-
-  const providerLabel = t(PROVIDER_LABEL_KEYS[provider] || 'claudeStatus.providers.assistant', { defaultValue: 'Assistant' });
-
-  return (
-    <div className="animate-in fade-in slide-in-from-bottom-2 mb-3 w-full duration-500">
-      <div className="mx-auto flex max-w-4xl items-center justify-between gap-3 overflow-hidden rounded-full border border-border/50 bg-slate-100 px-3 py-1.5 shadow-sm backdrop-blur-md dark:bg-slate-900">
-
-        {/* Left Side: Identity & Status */}
-        <div className="flex min-w-0 items-center gap-2.5">
-          <div className="relative flex h-6 w-6 shrink-0 items-center justify-center rounded-full bg-primary/20 ring-1 ring-primary/10">
-            <SessionProviderLogo provider={provider} className="h-3.5 w-3.5" />
-            {isLoading && (
-              <span className="absolute inset-0 animate-pulse rounded-full ring-2 ring-emerald-500/20" />
-            )}
-          </div>
-
-          <div className="flex min-w-0 flex-col sm:flex-row sm:items-center sm:gap-2">
-            <span className="text-[10px] font-bold uppercase tracking-wider text-muted-foreground/70">
-              {providerLabel}
-            </span>
-            <div className="flex items-center gap-1.5">
-              <span className={cn("h-1.5 w-1.5 rounded-full", isLoading ? "bg-emerald-500 animate-pulse" : "bg-amber-500")} />
-              <p className="truncate text-xs font-medium text-foreground">
-                {statusText}<span className="inline-block w-4 text-primary">{isLoading ? dots : ''}</span>
-              </p>
-            </div>
-          </div>
-        </div>
-
-        {/* Right Side: Metrics & Actions */}
-        <div className="flex items-center gap-2">
-          {isLoading && status?.can_interrupt !== false && onAbort && (
-            <>
-              <div className="hidden items-center rounded-md bg-muted/50 px-2 py-0.5 text-[10px] font-medium tabular-nums text-muted-foreground sm:flex">
-                {formatElapsedTime(elapsedTime)}
-              </div>
-
-              <button
-                type="button"
-                onClick={onAbort}
-                className="group flex items-center gap-1.5 rounded-full bg-destructive/10 px-2.5 py-1 text-[10px] font-bold text-destructive transition-all hover:bg-destructive hover:text-destructive-foreground"
-              >
-                <svg className="h-3 w-3 fill-current" viewBox="0 0 24 24">
-                  <path d="M6 6h12v12H6z" />
-                </svg>
-                <span className="hidden sm:inline">STOP</span>
-                <kbd className="hidden rounded bg-black/10 px-1 text-[9px] group-hover:bg-white/20 sm:block">
-                  ESC
-                </kbd>
-              </button>
-            </>
-          )}
-        </div>
-      </div>
-    </div>
-  );
-}

src/components/chat/view/subcomponents/MessageComponent.tsx

@@ -1,5 +1,6 @@
 import { memo, useEffect, useMemo, useRef, useState } from 'react';
 import { useTranslation } from 'react-i18next';
+
 import SessionProviderLogo from '../../../llm-logo-provider/SessionProviderLogo';
 import type {
   ChatMessage,
@@ -8,10 +9,10 @@ import type {
   Provider,
 } from '../../types/types';
 import { formatUsageLimitText } from '../../utils/chatFormatting';
-import { getClaudePermissionSuggestion } from '../../utils/chatPermissions';
 import type { Project } from '../../../../types/app';
 import { ToolRenderer, shouldHideToolResult } from '../../tools';
 import { Reasoning, ReasoningTrigger, ReasoningContent } from '../../../../shared/view/ui';
+
 import { Markdown } from './Markdown';
 import MessageCopyControl from './MessageCopyControl';
 
@@ -41,10 +42,9 @@ type InteractiveOption = {
   isSelected: boolean;
 };
 
-type PermissionGrantState = 'idle' | 'granted' | 'error';
 const COPY_HIDDEN_TOOL_NAMES = new Set(['Bash', 'Edit', 'Write', 'ApplyPatch']);
 
-const MessageComponent = memo(({ message, prevMessage, createDiff, onFileOpen, onShowSettings, onGrantToolPermission, autoExpandTools, showRawParameters, showThinking, selectedProject, provider }: MessageComponentProps) => {
+const MessageComponent = memo(({ message, prevMessage, createDiff, onFileOpen, autoExpandTools, showRawParameters, showThinking, selectedProject, provider }: MessageComponentProps) => {
   const { t } = useTranslation('chat');
   const isGrouped = prevMessage && prevMessage.type === message.type &&
     ((prevMessage.type === 'assistant') ||
@@ -53,8 +53,6 @@ const MessageComponent = memo(({ message, prevMessage, createDiff, onFileOpen, o
       (prevMessage.type === 'error'));
   const messageRef = useRef<HTMLDivElement | null>(null);
   const [isExpanded, setIsExpanded] = useState(false);
-  const permissionSuggestion = getClaudePermissionSuggestion(message, provider);
-  const [permissionGrantState, setPermissionGrantState] = useState<PermissionGrantState>('idle');
   const userCopyContent = String(message.content || '');
   const formattedMessageContent = useMemo(
     () => formatUsageLimitText(String(message.content || '')),
@@ -73,10 +71,6 @@ const MessageComponent = memo(({ message, prevMessage, createDiff, onFileOpen, o
     !message.isThinking;
 
 
-  useEffect(() => {
-    setPermissionGrantState('idle');
-  }, [permissionSuggestion?.entry, message.toolId]);
-
   useEffect(() => {
     const node = messageRef.current;
     if (!autoExpandTools || !node || !message.isToolUse) return;
@@ -120,7 +114,7 @@ const MessageComponent = memo(({ message, prevMessage, createDiff, onFileOpen, o
         /* User message bubble on the right */
         <div className="flex w-full items-end space-x-0 sm:w-auto sm:max-w-[85%] sm:space-x-3 md:max-w-md lg:max-w-lg xl:max-w-xl">
           <div className="group flex-1 rounded-2xl rounded-br-md bg-blue-600 px-3 py-2 text-white shadow-sm sm:flex-initial sm:px-4">
-            <div className="whitespace-pre-wrap break-words text-sm">
+            <div dir="auto" className="whitespace-pre-wrap break-words text-sm">
               {message.content}
             </div>
             {message.images && message.images.length > 0 && (
@@ -241,55 +235,6 @@ const MessageComponent = memo(({ message, prevMessage, createDiff, onFileOpen, o
                         <Markdown className="prose prose-sm prose-red max-w-none dark:prose-invert">
                           {String(message.toolResult.content || '')}
                         </Markdown>
-                        {permissionSuggestion && (
-                          <div className="mt-4 border-t border-red-200/60 pt-3 dark:border-red-800/60">
-                            <div className="flex flex-wrap items-center gap-2">
-                              <button
-                                type="button"
-                                onClick={() => {
-                                  if (!onGrantToolPermission) return;
-                                  const result = onGrantToolPermission(permissionSuggestion);
-                                  if (result?.success) {
-                                    setPermissionGrantState('granted');
-                                  } else {
-                                    setPermissionGrantState('error');
-                                  }
-                                }}
-                                disabled={permissionSuggestion.isAllowed || permissionGrantState === 'granted'}
-                                className={`inline-flex items-center gap-2 rounded-md border px-3 py-1.5 text-xs font-medium transition-colors ${permissionSuggestion.isAllowed || permissionGrantState === 'granted'
-                                  ? 'cursor-default border-green-300/70 bg-green-100 text-green-800 dark:border-green-800/60 dark:bg-green-900/30 dark:text-green-200'
-                                  : 'border-red-300/70 bg-white/80 text-red-700 hover:bg-white dark:border-red-800/60 dark:bg-gray-900/40 dark:text-red-200 dark:hover:bg-gray-900/70'
-                                  }`}
-                              >
-                                {permissionSuggestion.isAllowed || permissionGrantState === 'granted'
-                                  ? t('permissions.added')
-                                  : t('permissions.grant', { tool: permissionSuggestion.toolName })}
-                              </button>
-                              {onShowSettings && (
-                                <button
-                                  type="button"
-                                  onClick={(e) => { e.stopPropagation(); onShowSettings(); }}
-                                  className="text-xs text-red-700 underline hover:text-red-800 dark:text-red-200 dark:hover:text-red-100"
-                                >
-                                  {t('permissions.openSettings')}
-                                </button>
-                              )}
-                            </div>
-                            <div className="mt-2 text-xs text-red-700/90 dark:text-red-200/80">
-                              {t('permissions.addTo', { entry: permissionSuggestion.entry })}
-                            </div>
-                            {permissionGrantState === 'error' && (
-                              <div className="mt-2 text-xs text-red-700 dark:text-red-200">
-                                {t('permissions.error')}
-                              </div>
-                            )}
-                            {(permissionSuggestion.isAllowed || permissionGrantState === 'granted') && (
-                              <div className="mt-2 text-xs text-green-700 dark:text-green-200">
-                                {t('permissions.retry')}
-                              </div>
-                            )}
-                          </div>
-                        )}
                       </div>
                     </div>
                   ) : (
@@ -405,7 +350,7 @@ const MessageComponent = memo(({ message, prevMessage, createDiff, onFileOpen, o
                 </ReasoningContent>
               </Reasoning>
             ) : (
-              <div className="text-sm text-gray-700 dark:text-gray-300">
+              <div dir="auto" className="text-sm text-gray-700 dark:text-gray-300">
                 {/* Reasoning accordion */}
                 {showThinking && message.reasoning && (
                   <Reasoning className="mb-3" defaultOpen={false}>

src/components/chat/view/subcomponents/ProviderSelectionEmptyState.tsx

@@ -277,11 +277,15 @@ export default function ProviderSelectionEmptyState({
                           >
                             <div className="min-w-0 flex-1">
                               <div className="truncate">{model.label}</div>
-                              {model.description && (
+                              {/* 
+                              // * Temporarly commented out because the description of models from claude 
+                              // * was a bit inconsistent.  Will return it back when it becomes more consistent.
+                              */}
+                              {/* {model.description && (
                                 <div className="truncate text-xs text-muted-foreground">
                                   {model.description}
                                 </div>
-                              )}
+                              )} */}
                             </div>
                             {isSelected && (
                               <Check className="ml-auto h-4 w-4 shrink-0 text-primary" />
@@ -321,6 +325,7 @@ export default function ProviderSelectionEmptyState({
 
           <p className="mt-3 flex items-center justify-center gap-1.5 text-center text-xs text-muted-foreground/60">
             <Trans
+              ns="chat"
               i18nKey="providerSelection.pressToSearch"
               values={{ shortcut: MOD_KEY === "⌘" ? "⌘K" : "Ctrl+K" }}
               components={{

src/components/chat/view/subcomponents/ThinkingModeSelector.tsx

@@ -1,244 +0,0 @@
-import { useState, useRef, useEffect, useCallback, type CSSProperties } from 'react';
-import { createPortal } from 'react-dom';
-import { Brain, X } from 'lucide-react';
-import { useTranslation } from 'react-i18next';
-import { thinkingModes } from '../../constants/thinkingModes';
-
-type ThinkingModeSelectorProps = {
-  selectedMode: string;
-  onModeChange: (modeId: string) => void;
-  onClose?: () => void;
-  className?: string;
-};
-
-function ThinkingModeSelector({ selectedMode, onModeChange, onClose, className = '' }: ThinkingModeSelectorProps) {
-  const { t } = useTranslation('chat');
-  const [isOpen, setIsOpen] = useState(false);
-  const containerRef = useRef<HTMLDivElement>(null);
-  const triggerRef = useRef<HTMLButtonElement>(null);
-  const dropdownRef = useRef<HTMLDivElement>(null);
-  const [dropdownStyle, setDropdownStyle] = useState<CSSProperties | null>(null);
-
-  // Mapping from mode ID to translation key
-  const modeKeyMap: Record<string, string> = {
-    'think-hard': 'thinkHard',
-    'think-harder': 'thinkHarder'
-  };
-  // Create translated modes for display
-  const translatedModes = thinkingModes.map(mode => {
-    const modeKey = modeKeyMap[mode.id] || mode.id;
-    return {
-      ...mode,
-      name: t(`thinkingMode.modes.${modeKey}.name`),
-      description: t(`thinkingMode.modes.${modeKey}.description`),
-      prefix: t(`thinkingMode.modes.${modeKey}.prefix`)
-    };
-  });
-
-  const closeDropdown = useCallback(() => {
-    setIsOpen(false);
-    onClose?.();
-  }, [onClose]);
-
-  const updateDropdownPosition = useCallback(() => {
-    const trigger = triggerRef.current;
-    const dropdown = dropdownRef.current;
-    if (!trigger || !dropdown || typeof window === 'undefined') {
-      return;
-    }
-
-    const triggerRect = trigger.getBoundingClientRect();
-    const viewportPadding = window.innerWidth < 640 ? 12 : 16;
-    const spacing = 8;
-    const width = Math.min(window.innerWidth - viewportPadding * 2, window.innerWidth < 640 ? 320 : 256);
-    let left = triggerRect.left + triggerRect.width / 2 - width / 2;
-    left = Math.max(viewportPadding, Math.min(left, window.innerWidth - width - viewportPadding));
-
-    const measuredHeight = dropdown.offsetHeight || 0;
-    const spaceBelow = window.innerHeight - triggerRect.bottom - spacing - viewportPadding;
-    const spaceAbove = triggerRect.top - spacing - viewportPadding;
-    const openBelow = spaceBelow >= Math.min(measuredHeight || 320, 320) || spaceBelow >= spaceAbove;
-    const availableHeight = Math.min(
-      window.innerHeight - viewportPadding * 2,
-      Math.max(180, openBelow ? spaceBelow : spaceAbove),
-    );
-    const panelHeight = Math.min(measuredHeight || availableHeight, availableHeight);
-    const top = openBelow
-      ? Math.min(triggerRect.bottom + spacing, window.innerHeight - viewportPadding - panelHeight)
-      : Math.max(viewportPadding, triggerRect.top - spacing - panelHeight);
-
-    setDropdownStyle({
-      position: 'fixed',
-      top,
-      left,
-      width,
-      maxHeight: availableHeight,
-      zIndex: 80,
-    });
-  }, []);
-
-  useEffect(() => {
-    if (!isOpen) {
-      setDropdownStyle(null);
-      return;
-    }
-
-    const rafId = window.requestAnimationFrame(updateDropdownPosition);
-    const handleViewportChange = () => updateDropdownPosition();
-
-    window.addEventListener('resize', handleViewportChange);
-    window.addEventListener('scroll', handleViewportChange, true);
-
-    return () => {
-      window.cancelAnimationFrame(rafId);
-      window.removeEventListener('resize', handleViewportChange);
-      window.removeEventListener('scroll', handleViewportChange, true);
-    };
-  }, [isOpen, updateDropdownPosition]);
-
-  useEffect(() => {
-    if (!isOpen) {
-      return;
-    }
-
-    const handlePointerDown = (event: PointerEvent) => {
-      const target = event.target;
-      if (!(target instanceof Node)) {
-        return;
-      }
-
-      if (containerRef.current?.contains(target) || dropdownRef.current?.contains(target)) {
-        return;
-      }
-
-      closeDropdown();
-    };
-
-    const handleKeyDown = (event: KeyboardEvent) => {
-      if (event.key === 'Escape') {
-        closeDropdown();
-      }
-    };
-
-    document.addEventListener('pointerdown', handlePointerDown, true);
-    document.addEventListener('keydown', handleKeyDown);
-
-    return () => {
-      document.removeEventListener('pointerdown', handlePointerDown, true);
-      document.removeEventListener('keydown', handleKeyDown);
-    };
-  }, [isOpen, closeDropdown]);
-
-  const currentMode = translatedModes.find(mode => mode.id === selectedMode) || translatedModes[0];
-  const IconComponent = currentMode.icon || Brain;
-
-  return (
-    <div className={`relative ${className}`} ref={containerRef}>
-      <button
-        ref={triggerRef}
-        type="button"
-        onClick={() => {
-          if (isOpen) {
-            closeDropdown();
-            return;
-          }
-
-          setIsOpen(true);
-        }}
-        className={`flex h-10 w-10 items-center justify-center rounded-full transition-all duration-200 sm:h-10 sm:w-10 ${selectedMode === 'none'
-            ? 'bg-gray-100 hover:bg-gray-200 dark:bg-gray-700 dark:hover:bg-gray-600'
-            : 'bg-blue-100 hover:bg-blue-200 dark:bg-blue-900 dark:hover:bg-blue-800'
-          }`}
-        title={t('thinkingMode.buttonTitle', { mode: currentMode.name })}
-        aria-haspopup="dialog"
-        aria-expanded={isOpen}
-      >
-        <IconComponent className={`h-5 w-5 ${currentMode.color}`} />
-      </button>
-
-      {isOpen && typeof document !== 'undefined' && createPortal(
-        <div
-          ref={dropdownRef}
-          style={dropdownStyle || { position: 'fixed', top: 0, left: 0, visibility: 'hidden' }}
-          className="flex flex-col overflow-hidden rounded-xl border border-gray-200 bg-white shadow-xl dark:border-gray-700 dark:bg-gray-800"
-          role="dialog"
-          aria-modal="false"
-        >
-          <div className="border-b border-gray-200 p-3 dark:border-gray-700">
-            <div className="flex items-center justify-between">
-              <h3 className="text-sm font-semibold text-gray-900 dark:text-white">
-                {t('thinkingMode.selector.title')}
-              </h3>
-              <button
-                type="button"
-                onClick={closeDropdown}
-                className="rounded p-1 hover:bg-gray-100 dark:hover:bg-gray-700"
-              >
-                <X className="h-4 w-4 text-gray-500" />
-              </button>
-            </div>
-            <p className="mt-1 text-xs text-gray-500 dark:text-gray-400">
-              {t('thinkingMode.selector.description')}
-            </p>
-          </div>
-
-          <div className="min-h-0 overflow-y-auto py-1">
-            {translatedModes.map((mode) => {
-              const ModeIcon = mode.icon;
-              const isSelected = mode.id === selectedMode;
-
-              return (
-                <button
-                  key={mode.id}
-                  type="button"
-                  onClick={() => {
-                    onModeChange(mode.id);
-                    closeDropdown();
-                  }}
-                  className={`w-full px-4 py-3 text-left transition-colors hover:bg-gray-50 dark:hover:bg-gray-700 ${isSelected ? 'bg-gray-50 dark:bg-gray-700' : ''
-                    }`}
-                >
-                  <div className="flex items-start gap-3">
-                    <div className={`mt-0.5 ${mode.icon ? mode.color : 'text-gray-400'}`}>
-                      {ModeIcon ? <ModeIcon className="h-5 w-5" /> : <div className="h-5 w-5" />}
-                    </div>
-                    <div className="min-w-0 flex-1">
-                      <div className="flex items-center gap-2">
-                        <span className={`text-sm font-medium ${isSelected ? 'text-gray-900 dark:text-white' : 'text-gray-700 dark:text-gray-300'
-                          }`}>
-                          {mode.name}
-                        </span>
-                        {isSelected && (
-                          <span className="rounded bg-blue-100 px-2 py-0.5 text-xs text-blue-700 dark:bg-blue-900 dark:text-blue-300">
-                            {t('thinkingMode.selector.active')}
-                          </span>
-                        )}
-                      </div>
-                      <p className="mt-0.5 text-xs text-gray-500 dark:text-gray-400">
-                        {mode.description}
-                      </p>
-                      {mode.prefix && (
-                        <code className="mt-1 inline-block rounded bg-gray-100 px-1.5 py-0.5 text-xs dark:bg-gray-700">
-                          {mode.prefix}
-                        </code>
-                      )}
-                    </div>
-                  </div>
-                </button>
-              );
-            })}
-          </div>
-
-          <div className="border-t border-gray-200 bg-gray-50 p-3 dark:border-gray-700 dark:bg-gray-900">
-            <p className="text-xs text-gray-600 dark:text-gray-400">
-              <strong>Tip:</strong> {t('thinkingMode.selector.tip')}
-            </p>
-          </div>
-        </div>,
-        document.body
-      )}
-    </div>
-  );
-}
-
-export default ThinkingModeSelector;

src/components/chat/view/subcomponents/TokenUsageSummary.tsx

@@ -2,6 +2,7 @@ import { ActivityIcon } from 'lucide-react';
 
 type TokenUsageSummaryProps = {
   usage: Record<string, unknown> | null;
+  onClick?: () => void;
 };
 
 const formatTokenCount = (value: number) => {
@@ -29,7 +30,7 @@ const readUsageNumber = (value: unknown) => {
   return Number.isFinite(parsed) ? parsed : 0;
 };
 
-export default function TokenUsageSummary({ usage }: TokenUsageSummaryProps) {
+export default function TokenUsageSummary({ usage, onClick }: TokenUsageSummaryProps) {
   const breakdown =
     usage?.breakdown && typeof usage.breakdown === 'object'
       ? usage.breakdown as Record<string, unknown>
@@ -39,15 +40,18 @@ export default function TokenUsageSummary({ usage }: TokenUsageSummaryProps) {
   const usedTokens = readUsageNumber(usage?.used) || inputTokens + outputTokens;
 
   return (
-    <div
-      className="inline-flex h-9 items-center gap-1.5 rounded-lg border border-border/70 bg-background/70 px-2 text-xs text-muted-foreground shadow-sm transition-colors hover:border-primary/25 hover:text-foreground sm:gap-2 sm:px-2.5"
+    <button
+      type="button"
+      onClick={onClick}
+      className="inline-flex h-9 items-center gap-1.5 rounded-lg border border-border/70 bg-background/70 px-2 text-xs text-muted-foreground shadow-sm transition-colors hover:border-primary/25 hover:text-foreground focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-ring focus-visible:ring-offset-2 sm:gap-2 sm:px-2.5"
       title={`${usedTokens.toLocaleString()} tokens used`}
+      aria-label="Show token usage"
     >
       <span className="grid h-5 w-5 place-items-center rounded-md bg-primary/10 text-primary">
         <ActivityIcon className="h-3.5 w-3.5" />
       </span>
       <span className="font-medium text-foreground">{formatTokenCount(usedTokens)}</span>
       <span className="hidden text-muted-foreground/70 sm:inline">tokens</span>
-    </div>
+    </button>
   );
 }

src/components/code-editor/view/EditorSidebar.tsx

@@ -102,7 +102,7 @@ export default function EditorSidebar({
   const useFlexLayout = editorExpanded || (fillSpace && !hasManualWidth);
 
   return (
-    <div ref={containerRef} className={`flex h-full min-w-0 flex-shrink-0 ${editorExpanded ? 'flex-1' : ''}`}>
+    <div ref={containerRef} className={`flex h-full min-w-0 ${editorExpanded ? 'flex-1' : ''}`}>
       {!editorExpanded && (
         <div
           ref={resizeHandleRef}

src/components/file-tree/constants/constants.ts

@@ -6,6 +6,14 @@ export const FILE_TREE_DEFAULT_VIEW_MODE: FileTreeViewMode = 'detailed';
 
 export const FILE_TREE_VIEW_MODES: FileTreeViewMode[] = ['simple', 'compact', 'detailed'];
 
+export const MAX_FILE_UPLOAD_SIZE_MB = 200;
+
+export const MAX_FILE_UPLOAD_SIZE_BYTES = MAX_FILE_UPLOAD_SIZE_MB * 1024 * 1024;
+
+export const MAX_FILE_UPLOAD_SIZE_LABEL = `${MAX_FILE_UPLOAD_SIZE_MB}MB`;
+
+export const MAX_FILE_UPLOAD_COUNT = 20;
+
 export const IMAGE_FILE_EXTENSIONS = new Set([
   'png',
   'jpg',

src/components/file-tree/hooks/useFileTreeUpload.ts

@@ -1,6 +1,13 @@
-import { useCallback, useState, useRef } from 'react';
+import { useCallback, useEffect, useRef, useState } from 'react';
+import type { DragEvent } from 'react';
+
+import { IS_PLATFORM } from '../../../constants/config';
 import type { Project } from '../../../types/app';
-import { api } from '../../../utils/api';
+import {
+  MAX_FILE_UPLOAD_COUNT,
+  MAX_FILE_UPLOAD_SIZE_BYTES,
+  MAX_FILE_UPLOAD_SIZE_LABEL,
+} from '../constants/constants';
 
 type UseFileTreeUploadOptions = {
   selectedProject: Project | null;
@@ -8,6 +15,141 @@ type UseFileTreeUploadOptions = {
   showToast: (message: string, type: 'success' | 'error') => void;
 };
 
+export type FileTreeUploadProgressState = {
+  status: 'uploading' | 'complete' | 'error';
+  progress: number;
+  fileCount: number;
+  uploadedCount?: number;
+  fileName?: string;
+  targetPath?: string;
+  error?: string;
+};
+
+type UploadResponse = {
+  error?: string;
+  message?: string;
+  files?: unknown[];
+  uploadedCount?: number;
+  requestedFileCount?: number;
+};
+
+const COMPLETE_PROGRESS_CLEAR_DELAY_MS = 1400;
+const ERROR_PROGRESS_CLEAR_DELAY_MS = 3200;
+
+const pluralizeFiles = (count: number) => (count === 1 ? 'file' : 'files');
+
+const getRelativePath = (file: File) => {
+  const fileWithRelativePath = file as File & { webkitRelativePath?: string };
+  return fileWithRelativePath.webkitRelativePath || file.name;
+};
+
+const getFileDisplayName = (file: File) => {
+  const relativePath = getRelativePath(file);
+  return relativePath.split(/[\\/]/).pop() || file.name;
+};
+
+const validateFilesForUpload = (files: File[]): string | null => {
+  if (files.length > MAX_FILE_UPLOAD_COUNT) {
+    return `You can upload up to ${MAX_FILE_UPLOAD_COUNT} files at once.`;
+  }
+
+  const oversizedFile = files.find((file) => file.size > MAX_FILE_UPLOAD_SIZE_BYTES);
+  if (oversizedFile) {
+    return `${getFileDisplayName(oversizedFile)} is larger than ${MAX_FILE_UPLOAD_SIZE_LABEL}.`;
+  }
+
+  return null;
+};
+
+const parseUploadResponse = (xhr: XMLHttpRequest): UploadResponse => {
+  if (!xhr.responseText) {
+    return {};
+  }
+
+  try {
+    return JSON.parse(xhr.responseText) as UploadResponse;
+  } catch {
+    return {};
+  }
+};
+
+const formatUploadSuccessMessage = (uploadedCount: number, requestedFileCount: number) => {
+  if (uploadedCount !== requestedFileCount) {
+    return `Uploaded ${uploadedCount} of ${requestedFileCount} ${pluralizeFiles(requestedFileCount)}`;
+  }
+
+  return `Uploaded ${uploadedCount} ${pluralizeFiles(uploadedCount)} successfully`;
+};
+
+const buildUploadFormData = (files: File[], targetPath: string) => {
+  const formData = new FormData();
+  const relativePaths: string[] = [];
+
+  formData.append('targetPath', targetPath);
+  formData.append('requestedFileCount', String(files.length));
+
+  files.forEach((file) => {
+    const relativePath = getRelativePath(file);
+    const cleanFile = new File([file], relativePath.split(/[\\/]/).pop() || file.name, {
+      type: file.type,
+      lastModified: file.lastModified,
+    });
+
+    formData.append('files', cleanFile);
+    relativePaths.push(relativePath);
+  });
+
+  formData.append('relativePaths', JSON.stringify(relativePaths));
+
+  return formData;
+};
+
+const uploadFormDataWithProgress = (
+  projectId: string,
+  formData: FormData,
+  onProgress: (progress: number) => void,
+) =>
+  new Promise<UploadResponse>((resolve, reject) => {
+    const xhr = new XMLHttpRequest();
+
+    xhr.open('POST', `/api/projects/${encodeURIComponent(projectId)}/files/upload`);
+
+    const token = localStorage.getItem('auth-token');
+    if (!IS_PLATFORM && token) {
+      xhr.setRequestHeader('Authorization', `Bearer ${token}`);
+    }
+
+    xhr.upload.onprogress = (event) => {
+      if (!event.lengthComputable) {
+        return;
+      }
+
+      // Keep 100% for the server response so the UI can distinguish transfer
+      // completion from the final write/refresh step.
+      onProgress(Math.min(99, Math.round((event.loaded / event.total) * 100)));
+    };
+
+    xhr.onload = () => {
+      const refreshedToken = xhr.getResponseHeader('X-Refreshed-Token');
+      if (refreshedToken) {
+        localStorage.setItem('auth-token', refreshedToken);
+      }
+
+      const payload = parseUploadResponse(xhr);
+      if (xhr.status >= 200 && xhr.status < 300) {
+        resolve(payload);
+        return;
+      }
+
+      reject(new Error(payload.error || payload.message || `Upload failed with status ${xhr.status}`));
+    };
+
+    xhr.onerror = () => reject(new Error('Upload failed. Check your connection and try again.'));
+    xhr.onabort = () => reject(new Error('Upload canceled.'));
+
+    xhr.send(formData);
+  });
+
 // Helper function to read all files from a directory entry recursively
 const readAllDirectoryEntries = async (directoryEntry: FileSystemDirectoryEntry, basePath = ''): Promise<File[]> => {
   const files: File[] = [];
@@ -57,6 +199,48 @@ const readAllDirectoryEntries = async (directoryEntry: FileSystemDirectoryEntry,
   return files;
 };
 
+const collectDroppedFiles = async (dataTransfer: DataTransfer) => {
+  const files: File[] = [];
+
+  // Use DataTransferItemList for folder support
+  const { items } = dataTransfer;
+  if (items) {
+    for (const item of Array.from(items)) {
+      if (item.kind !== 'file') {
+        continue;
+      }
+
+      const entry = item.webkitGetAsEntry ? item.webkitGetAsEntry() : null;
+      if (!entry) {
+        const file = item.getAsFile();
+        if (file) {
+          files.push(file);
+        }
+        continue;
+      }
+
+      if (entry.isFile) {
+        const file = await new Promise<File>((resolve, reject) => {
+          (entry as FileSystemFileEntry).file(resolve, reject);
+        });
+        files.push(file);
+      } else if (entry.isDirectory) {
+        // Pass the directory name as basePath so files include the folder path
+        const dirFiles = await readAllDirectoryEntries(entry as FileSystemDirectoryEntry, entry.name);
+        files.push(...dirFiles);
+      }
+    }
+    return files;
+  }
+
+  // Fallback for browsers that don't support webkitGetAsEntry
+  for (const file of Array.from(dataTransfer.files)) {
+    files.push(file);
+  }
+
+  return files;
+};
+
 export const useFileTreeUpload = ({
   selectedProject,
   onRefresh,
@@ -65,20 +249,150 @@ export const useFileTreeUpload = ({
   const [isDragOver, setIsDragOver] = useState(false);
   const [dropTarget, setDropTarget] = useState<string | null>(null);
   const [operationLoading, setOperationLoading] = useState(false);
+  const [uploadProgress, setUploadProgress] = useState<FileTreeUploadProgressState | null>(null);
   const treeRef = useRef<HTMLDivElement>(null);
+  const clearProgressTimerRef = useRef<ReturnType<typeof setTimeout> | null>(null);
 
-  const handleDragEnter = useCallback((e: React.DragEvent) => {
+  const clearProgressTimer = useCallback(() => {
+    if (clearProgressTimerRef.current) {
+      clearTimeout(clearProgressTimerRef.current);
+      clearProgressTimerRef.current = null;
+    }
+  }, []);
+
+  const scheduleProgressClear = useCallback(
+    (delay: number) => {
+      clearProgressTimer();
+      clearProgressTimerRef.current = setTimeout(() => {
+        setUploadProgress(null);
+        clearProgressTimerRef.current = null;
+      }, delay);
+    },
+    [clearProgressTimer],
+  );
+
+  useEffect(() => clearProgressTimer, [clearProgressTimer]);
+
+  const setUploadError = useCallback(
+    (message: string, fileCount: number, targetPath = '', fileName?: string, progress = 0) => {
+      setUploadProgress({
+        status: 'error',
+        progress,
+        fileCount,
+        fileName,
+        targetPath,
+        error: message,
+      });
+      scheduleProgressClear(ERROR_PROGRESS_CLEAR_DELAY_MS);
+    },
+    [scheduleProgressClear],
+  );
+
+  const uploadFiles = useCallback(
+    async (files: File[], targetPath = '') => {
+      if (files.length === 0) {
+        setDropTarget(null);
+        return;
+      }
+
+      const fileName = files.length === 1 ? getFileDisplayName(files[0]) : undefined;
+
+      if (!selectedProject) {
+        const message = 'Select a project before uploading files.';
+        showToast(message, 'error');
+        setUploadError(message, files.length, targetPath, fileName);
+        return;
+      }
+
+      const validationError = validateFilesForUpload(files);
+      if (validationError) {
+        showToast(validationError, 'error');
+        setUploadError(validationError, files.length, targetPath, fileName);
+        return;
+      }
+
+      clearProgressTimer();
+      setOperationLoading(true);
+      setUploadProgress({
+        status: 'uploading',
+        progress: 0,
+        fileCount: files.length,
+        fileName,
+        targetPath,
+      });
+
+      let latestProgress = 0;
+
+      try {
+        const response = await uploadFormDataWithProgress(
+          selectedProject.projectId,
+          buildUploadFormData(files, targetPath),
+          (progress) => {
+            latestProgress = progress;
+            setUploadProgress((current) =>
+              current && current.status === 'uploading'
+                ? { ...current, progress }
+                : current,
+            );
+          },
+        );
+
+        const uploadedCount =
+          typeof response.uploadedCount === 'number' ? response.uploadedCount : response.files?.length ?? files.length;
+        const requestedFileCount =
+          typeof response.requestedFileCount === 'number' ? response.requestedFileCount : files.length;
+
+        setUploadProgress({
+          status: 'complete',
+          progress: 100,
+          fileCount: requestedFileCount,
+          uploadedCount,
+          fileName,
+          targetPath,
+        });
+
+        showToast(formatUploadSuccessMessage(uploadedCount, requestedFileCount), 'success');
+        scheduleProgressClear(COMPLETE_PROGRESS_CLEAR_DELAY_MS);
+        onRefresh();
+      } catch (err) {
+        const message = err instanceof Error ? err.message : 'Upload failed';
+        console.error('Upload error:', err);
+        showToast(message, 'error');
+        setUploadError(message, files.length, targetPath, fileName, latestProgress);
+      } finally {
+        setOperationLoading(false);
+        setDropTarget(null);
+      }
+    },
+    [
+      clearProgressTimer,
+      onRefresh,
+      scheduleProgressClear,
+      selectedProject,
+      setUploadError,
+      showToast,
+    ],
+  );
+
+  const handleFileSelect = useCallback(
+    async (fileList: FileList | File[]) => {
+      await uploadFiles(Array.from(fileList), '');
+    },
+    [uploadFiles],
+  );
+
+  const handleDragEnter = useCallback((e: DragEvent) => {
     e.preventDefault();
     e.stopPropagation();
     setIsDragOver(true);
   }, []);
 
-  const handleDragOver = useCallback((e: React.DragEvent) => {
+  const handleDragOver = useCallback((e: DragEvent) => {
     e.preventDefault();
     e.stopPropagation();
   }, []);
 
-  const handleDragLeave = useCallback((e: React.DragEvent) => {
+  const handleDragLeave = useCallback((e: DragEvent) => {
     e.preventDefault();
     e.stopPropagation();
     // Only set isDragOver to false if we're leaving the entire tree
@@ -88,103 +402,35 @@ export const useFileTreeUpload = ({
     }
   }, []);
 
-  const handleDrop = useCallback(async (e: React.DragEvent) => {
-    e.preventDefault();
-    e.stopPropagation();
-    setIsDragOver(false);
+  const handleDrop = useCallback(
+    async (e: DragEvent) => {
+      e.preventDefault();
+      e.stopPropagation();
+      setIsDragOver(false);
 
-    const targetPath = dropTarget || '';
-    setOperationLoading(true);
+      const targetPath = dropTarget || '';
 
-    try {
-      const files: File[] = [];
-
-      // Use DataTransferItemList for folder support
-      const items = e.dataTransfer.items;
-      if (items) {
-        for (const item of Array.from(items)) {
-          if (item.kind === 'file') {
-            const entry = item.webkitGetAsEntry ? item.webkitGetAsEntry() : null;
-
-            if (entry) {
-              if (entry.isFile) {
-                const file = await new Promise<File>((resolve, reject) => {
-                  (entry as FileSystemFileEntry).file(resolve, reject);
-                });
-                files.push(file);
-              } else if (entry.isDirectory) {
-                // Pass the directory name as basePath so files include the folder path
-                const dirFiles = await readAllDirectoryEntries(entry as FileSystemDirectoryEntry, entry.name);
-                files.push(...dirFiles);
-              }
-            }
-          }
-        }
-      } else {
-        // Fallback for browsers that don't support webkitGetAsEntry
-        const fileList = e.dataTransfer.files;
-        for (const file of Array.from(fileList)) {
-          files.push(file);
-        }
-      }
-
-      if (files.length === 0) {
-        setOperationLoading(false);
+      try {
+        const files = await collectDroppedFiles(e.dataTransfer);
+        await uploadFiles(files, targetPath);
+      } catch (err) {
+        const message = err instanceof Error ? err.message : 'Could not read dropped files';
+        console.error('Upload error:', err);
+        showToast(message, 'error');
+        setUploadError(message, 0, targetPath);
         setDropTarget(null);
-        return;
       }
+    },
+    [dropTarget, setUploadError, showToast, uploadFiles],
+  );
 
-      const formData = new FormData();
-      formData.append('targetPath', targetPath);
-
-      // Store relative paths separately since FormData strips path info from File.name
-      const relativePaths: string[] = [];
-      files.forEach((file) => {
-        // Create a new file with just the filename (without path) for FormData
-        // but store the relative path separately
-        const cleanFile = new File([file], file.name.split('/').pop()!, {
-          type: file.type,
-          lastModified: file.lastModified
-        });
-        formData.append('files', cleanFile);
-        relativePaths.push(file.name); // Keep the full relative path
-      });
-
-      // Send relative paths as a JSON array
-      formData.append('relativePaths', JSON.stringify(relativePaths));
-
-      const response = await api.post(
-        // File upload endpoint is keyed by DB projectId post-migration.
-        `/projects/${encodeURIComponent(selectedProject!.projectId)}/files/upload`,
-        formData
-      );
-
-      if (!response.ok) {
-        const data = await response.json();
-        throw new Error(data.error || 'Upload failed');
-      }
-
-      showToast(
-        `Uploaded ${files.length} file(s)`,
-        'success'
-      );
-      onRefresh();
-    } catch (err) {
-      console.error('Upload error:', err);
-      showToast(err instanceof Error ? err.message : 'Upload failed', 'error');
-    } finally {
-      setOperationLoading(false);
-      setDropTarget(null);
-    }
-  }, [dropTarget, selectedProject, onRefresh, showToast]);
-
-  const handleItemDragOver = useCallback((e: React.DragEvent, itemPath: string) => {
+  const handleItemDragOver = useCallback((e: DragEvent, itemPath: string) => {
     e.preventDefault();
     e.stopPropagation();
     setDropTarget(itemPath);
   }, []);
 
-  const handleItemDrop = useCallback((e: React.DragEvent, itemPath: string) => {
+  const handleItemDrop = useCallback((e: DragEvent, itemPath: string) => {
     e.preventDefault();
     e.stopPropagation();
     setDropTarget(itemPath);
@@ -194,7 +440,9 @@ export const useFileTreeUpload = ({
     isDragOver,
     dropTarget,
     operationLoading,
+    uploadProgress,
     treeRef,
+    handleFileSelect,
     handleDragEnter,
     handleDragOver,
     handleDragLeave,

src/components/file-tree/view/FileTree.tsx

@@ -1,6 +1,7 @@
 import { useCallback, useState, useEffect, useRef } from 'react';
 import { useTranslation } from 'react-i18next';
 import { AlertTriangle, Check, X, Loader2, Folder, Upload } from 'lucide-react';
+
 import { cn } from '../../../lib/utils';
 import { ICON_SIZE_CLASS, getFileIconData } from '../constants/fileIcons';
 import { useExpandedDirectories } from '../hooks/useExpandedDirectories';
@@ -13,10 +14,12 @@ import type { FileTreeImageSelection, FileTreeNode } from '../types/types';
 import { formatFileSize, formatRelativeTime, isImageFile } from '../utils/fileTreeUtils';
 import { Project } from '../../../types/app';
 import { ScrollArea, Input } from '../../../shared/view/ui';
+
 import FileTreeBody from './FileTreeBody';
 import FileTreeDetailedColumns from './FileTreeDetailedColumns';
 import FileTreeHeader from './FileTreeHeader';
 import FileTreeLoadingState from './FileTreeLoadingState';
+import FileTreeUploadProgress from './FileTreeUploadProgress';
 import ImageViewer from './ImageViewer';
 
 
@@ -66,6 +69,7 @@ export default function FileTree({ selectedProject, onFileOpen }: FileTreeProps)
     onRefresh: refreshFiles,
     showToast,
   });
+  const operationLoading = operations.operationLoading || upload.operationLoading;
 
   // Focus input when creating new item
   useEffect(() => {
@@ -146,14 +150,19 @@ export default function FileTree({ selectedProject, onFileOpen }: FileTreeProps)
         onViewModeChange={changeViewMode}
         searchQuery={searchQuery}
         onSearchQueryChange={setSearchQuery}
+        onUploadFiles={upload.handleFileSelect}
         onNewFile={() => operations.handleStartCreate('', 'file')}
         onNewFolder={() => operations.handleStartCreate('', 'directory')}
         onRefresh={refreshFiles}
         onCollapseAll={collapseAll}
         loading={loading}
-        operationLoading={operations.operationLoading}
+        operationLoading={operationLoading}
+        isUploading={upload.uploadProgress?.status === 'uploading'}
+        uploadProgress={upload.uploadProgress?.progress ?? null}
       />
 
+      <FileTreeUploadProgress upload={upload.uploadProgress} />
+
       {viewMode === 'detailed' && filteredFiles.length > 0 && <FileTreeDetailedColumns />}
 
       <ScrollArea className="flex-1 px-2 py-1">
@@ -184,7 +193,7 @@ export default function FileTree({ selectedProject, onFileOpen }: FileTreeProps)
                 }, 100);
               }}
               className="h-6 flex-1 text-sm"
-              disabled={operations.operationLoading}
+              disabled={operationLoading}
             />
           </div>
         )}
@@ -213,7 +222,7 @@ export default function FileTree({ selectedProject, onFileOpen }: FileTreeProps)
           handleConfirmRename={operations.handleConfirmRename}
           handleCancelRename={operations.handleCancelRename}
           renameInputRef={renameInputRef}
-          operationLoading={operations.operationLoading}
+          operationLoading={operationLoading}
         />
       </ScrollArea>
 
@@ -251,17 +260,17 @@ export default function FileTree({ selectedProject, onFileOpen }: FileTreeProps)
             <div className="flex justify-end gap-2">
               <button
                 onClick={operations.handleCancelDelete}
-                disabled={operations.operationLoading}
+                disabled={operationLoading}
                 className="rounded-md px-3 py-1.5 text-sm transition-colors hover:bg-accent"
               >
                 {t('common.cancel', 'Cancel')}
               </button>
               <button
                 onClick={operations.handleConfirmDelete}
-                disabled={operations.operationLoading}
+                disabled={operationLoading}
                 className="flex items-center gap-2 rounded-md bg-red-600 px-3 py-1.5 text-sm text-white transition-colors hover:bg-red-700 disabled:opacity-50"
               >
-                {operations.operationLoading && <Loader2 className="h-4 w-4 animate-spin" />}
+                {operationLoading && <Loader2 className="h-4 w-4 animate-spin" />}
                 {t('fileTree.delete.confirm', 'Delete')}
               </button>
             </div>

src/components/file-tree/view/FileTreeHeader.tsx

@@ -1,7 +1,11 @@
-import { ChevronDown, Eye, FileText, FolderPlus, List, RefreshCw, Search, TableProperties, X } from 'lucide-react';
+import { useRef } from 'react';
+import type { ChangeEvent } from 'react';
+import { ChevronDown, Eye, FileText, FolderPlus, List, Loader2, RefreshCw, Search, TableProperties, Upload, X } from 'lucide-react';
 import { useTranslation } from 'react-i18next';
+
 import { Button, Input } from '../../../shared/view/ui';
 import { cn } from '../../../lib/utils';
+import { MAX_FILE_UPLOAD_SIZE_LABEL } from '../constants/constants';
 import type { FileTreeViewMode } from '../types/types';
 
 type FileTreeHeaderProps = {
@@ -12,11 +16,14 @@ type FileTreeHeaderProps = {
   // Toolbar actions
   onNewFile?: () => void;
   onNewFolder?: () => void;
+  onUploadFiles?: (files: FileList) => void;
   onRefresh?: () => void;
   onCollapseAll?: () => void;
   // Loading state
   loading?: boolean;
   operationLoading?: boolean;
+  isUploading?: boolean;
+  uploadProgress?: number | null;
 };
 
 export default function FileTreeHeader({
@@ -26,12 +33,24 @@ export default function FileTreeHeader({
   onSearchQueryChange,
   onNewFile,
   onNewFolder,
+  onUploadFiles,
   onRefresh,
   onCollapseAll,
   loading,
   operationLoading,
+  isUploading,
+  uploadProgress,
 }: FileTreeHeaderProps) {
   const { t } = useTranslation();
+  const uploadInputRef = useRef<HTMLInputElement>(null);
+
+  const handleUploadInputChange = (event: ChangeEvent<HTMLInputElement>) => {
+    const { files } = event.target;
+    if (files && files.length > 0) {
+      onUploadFiles?.(files);
+    }
+    event.target.value = '';
+  };
 
   return (
     <div className="space-y-2 border-b border-border px-3 pb-2 pt-3">
@@ -40,6 +59,50 @@ export default function FileTreeHeader({
         <h3 className="text-sm font-medium text-foreground">{t('fileTree.files')}</h3>
         <div className="flex items-center gap-0.5">
           {/* Action buttons */}
+          {onUploadFiles && (
+            <>
+              <input
+                ref={uploadInputRef}
+                type="file"
+                multiple
+                className="hidden"
+                onChange={handleUploadInputChange}
+                tabIndex={-1}
+                aria-hidden="true"
+              />
+              <Button
+                variant="ghost"
+                size="sm"
+                className="relative h-7 w-7 p-0"
+                onClick={() => uploadInputRef.current?.click()}
+                title={
+                  isUploading
+                    ? t('fileTree.uploadingFiles', 'Uploading files')
+                    : t('fileTree.uploadFiles', 'Upload files (max {{size}} each)', {
+                        size: MAX_FILE_UPLOAD_SIZE_LABEL,
+                      })
+                }
+                aria-label={t('fileTree.uploadFiles', 'Upload files (max {{size}} each)', {
+                  size: MAX_FILE_UPLOAD_SIZE_LABEL,
+                })}
+                disabled={operationLoading}
+              >
+                {isUploading ? (
+                  <Loader2 className="h-3.5 w-3.5 animate-spin" />
+                ) : (
+                  <Upload className="h-3.5 w-3.5" />
+                )}
+                {isUploading && typeof uploadProgress === 'number' && (
+                  <span className="absolute bottom-0.5 left-1/2 h-0.5 w-4 -translate-x-1/2 overflow-hidden rounded-full bg-primary/20">
+                    <span
+                      className="block h-full rounded-full bg-primary transition-[width] duration-150"
+                      style={{ width: `${uploadProgress}%` }}
+                    />
+                  </span>
+                )}
+              </Button>
+            </>
+          )}
           {onNewFile && (
             <Button
               variant="ghost"

src/components/file-tree/view/FileTreeUploadProgress.tsx

@@ -0,0 +1,90 @@
+import { AlertCircle, CheckCircle2, Upload } from 'lucide-react';
+import { useTranslation } from 'react-i18next';
+
+import { cn } from '../../../lib/utils';
+import type { FileTreeUploadProgressState } from '../hooks/useFileTreeUpload';
+
+type FileTreeUploadProgressProps = {
+  upload: FileTreeUploadProgressState | null;
+};
+
+const clampProgress = (progress: number) => Math.min(100, Math.max(0, progress));
+
+const pluralizeFiles = (count: number) => (count === 1 ? 'file' : 'files');
+
+export default function FileTreeUploadProgress({ upload }: FileTreeUploadProgressProps) {
+  const { t } = useTranslation();
+
+  if (!upload) {
+    return null;
+  }
+
+  const progress = clampProgress(upload.progress);
+  const isUploading = upload.status === 'uploading';
+  const isComplete = upload.status === 'complete';
+  const isError = upload.status === 'error';
+  const fileSummary =
+    upload.fileCount === 1 && upload.fileName
+      ? upload.fileName
+      : `${upload.fileCount} ${pluralizeFiles(upload.fileCount)}`;
+
+  const title = isUploading
+    ? t('fileTree.uploadingFiles', 'Uploading files')
+    : isComplete
+    ? t('fileTree.uploadComplete', 'Upload complete')
+    : t('fileTree.uploadFailed', 'Upload failed');
+
+  const detail = isError
+    ? upload.error
+    : isComplete && typeof upload.uploadedCount === 'number'
+    ? t('fileTree.uploadedCount', 'Uploaded {{uploaded}} of {{total}} {{label}}', {
+        uploaded: upload.uploadedCount,
+        total: upload.fileCount,
+        label: pluralizeFiles(upload.fileCount),
+      })
+    : fileSummary;
+
+  const Icon = isError ? AlertCircle : isComplete ? CheckCircle2 : Upload;
+
+  return (
+    <div
+      className={cn(
+        'border-b px-3 py-2 transition-colors',
+        isError
+          ? 'border-red-500/20 bg-red-500/10 text-red-700 dark:text-red-300'
+          : isComplete
+          ? 'border-emerald-500/20 bg-emerald-500/10 text-emerald-700 dark:text-emerald-300'
+          : 'border-primary/20 bg-primary/10 text-foreground',
+      )}
+    >
+      <div className="flex min-h-[36px] items-center gap-2">
+        <div
+          className={cn(
+            'flex h-7 w-7 shrink-0 items-center justify-center rounded-md',
+            isError ? 'bg-red-500/15' : isComplete ? 'bg-emerald-500/15' : 'bg-primary/15',
+          )}
+        >
+          <Icon className={cn('h-3.5 w-3.5', isUploading && 'animate-pulse')} />
+        </div>
+        <div className="min-w-0 flex-1">
+          <div className="flex items-center justify-between gap-3">
+            <span className="truncate text-xs font-medium">{title}</span>
+            <span className="shrink-0 text-[11px] tabular-nums text-muted-foreground">
+              {isUploading ? `${progress}%` : isComplete ? t('common.done', 'Done') : t('common.failed', 'Failed')}
+            </span>
+          </div>
+          <div className="mt-1 truncate text-[11px] text-muted-foreground">{detail}</div>
+        </div>
+      </div>
+      <div className="mt-2 h-1.5 overflow-hidden rounded-full bg-background/80">
+        <div
+          className={cn(
+            'h-full rounded-full transition-[width] duration-200',
+            isError ? 'bg-red-500' : isComplete ? 'bg-emerald-500' : 'bg-primary',
+          )}
+          style={{ width: `${isError ? Math.max(progress, 8) : progress}%` }}
+        />
+      </div>
+    </div>
+  );
+}

src/components/main-content/types/types.ts

@@ -1,9 +1,12 @@
 import type { Dispatch, SetStateAction } from 'react';
 
 import type { AppTab, Project, ProjectSession } from '../../../types/app';
-import type { SessionNavigationOptions } from '../../chat/types/types';
-
-export type SessionLifecycleHandler = (sessionId?: string | null) => void;
+import type {
+  MarkSessionIdle,
+  MarkSessionProcessing,
+  SessionActivityMap,
+} from '../../../hooks/useSessionProtection';
+import type { SessionEstablishedContext, SessionNavigationOptions } from '../../chat/types/types';
 
 export type TaskMasterTask = {
   id: string | number;
@@ -41,17 +44,15 @@ export type MainContentProps = {
   setActiveTab: Dispatch<SetStateAction<AppTab>>;
   ws: WebSocket | null;
   sendMessage: (message: unknown) => void;
-  latestMessage: unknown;
   isMobile: boolean;
   onMenuClick: () => void;
   isLoading: boolean;
   onInputFocusChange: (focused: boolean) => void;
-  onSessionActive: SessionLifecycleHandler;
-  onSessionInactive: SessionLifecycleHandler;
-  onSessionProcessing: SessionLifecycleHandler;
-  onSessionNotProcessing: SessionLifecycleHandler;
-  processingSessions: Set<string>;
+  onSessionProcessing: MarkSessionProcessing;
+  onSessionIdle: MarkSessionIdle;
+  processingSessions: SessionActivityMap;
   onNavigateToSession: (targetSessionId: string, options?: SessionNavigationOptions) => void;
+  onSessionEstablished: (sessionId: string, context: SessionEstablishedContext) => void;
   onShowSettings: () => void;
   externalMessageUpdate: number;
   newSessionTrigger: number;

src/components/main-content/view/MainContent.tsx

@@ -37,17 +37,15 @@ function MainContent({
   setActiveTab,
   ws,
   sendMessage,
-  latestMessage,
   isMobile,
   onMenuClick,
   isLoading,
   onInputFocusChange,
-  onSessionActive,
-  onSessionInactive,
   onSessionProcessing,
-  onSessionNotProcessing,
+  onSessionIdle,
   processingSessions,
   onNavigateToSession,
+  onSessionEstablished,
   onShowSettings,
   externalMessageUpdate,
   newSessionTrigger,
@@ -128,15 +126,13 @@ function MainContent({
                 selectedSession={selectedSession}
                 ws={ws}
                 sendMessage={sendMessage}
-                latestMessage={latestMessage}
                 onFileOpen={handleFileOpen}
                 onInputFocusChange={onInputFocusChange}
-                onSessionActive={onSessionActive}
-                onSessionInactive={onSessionInactive}
                 onSessionProcessing={onSessionProcessing}
-                onSessionNotProcessing={onSessionNotProcessing}
+                onSessionIdle={onSessionIdle}
                 processingSessions={processingSessions}
                 onNavigateToSession={onNavigateToSession}
+                onSessionEstablished={onSessionEstablished}
                 onShowSettings={onShowSettings}
                 autoExpandTools={autoExpandTools}
                 showRawParameters={showRawParameters}

src/components/plugins/view/PluginSettingsTab.tsx

@@ -26,6 +26,7 @@ const STARTER_PLUGIN_URL = 'https://github.com/cloudcli-ai/cloudcli-plugin-start
 const TERMINAL_PLUGIN_URL = 'https://github.com/cloudcli-ai/cloudcli-plugin-terminal';
 const SCHEDULED_PROMPT_PLUGIN_URL = 'https://github.com/grostim/cloudcli-cron';
 const CLAUDE_WATCH_PLUGIN_URL = 'https://github.com/satsuki19980613/cloudcli-claude-watch';
+const PRISM_CLOUDCLI_PLUGIN_URL = 'https://github.com/jakeefr/cloudcli-plugin-prism';
 
 type PluginRecommendation = {
   id: string;
@@ -72,6 +73,14 @@ const UNOFFICIAL_PLUGIN_RECOMMENDATIONS: PluginRecommendation[] = [
     icon: Clock,
     source: 'unofficial',
   },
+  {
+    id: 'prism',
+    translationKey: 'prismCloudCLI',
+    repoUrl: PRISM_CLOUDCLI_PLUGIN_URL,
+    installedNames: ['prism'],
+    icon: Activity,
+    source: 'unofficial'
+  }
 ];
 
 function repoSlug(repoUrl: string) {

src/components/settings/hooks/useSettingsController.ts

@@ -1,6 +1,8 @@
 import { useCallback, useEffect, useRef, useState } from 'react';
+
 import { useTheme } from '../../../contexts/ThemeContext';
 import { authenticatedFetch } from '../../../utils/api';
+import { setNotificationSoundEnabled } from '../../../utils/notificationSound';
 import { useProviderAuthStatus } from '../../provider-auth/hooks/useProviderAuthStatus';
 import {
   DEFAULT_CODE_EDITOR_SETTINGS,
@@ -107,6 +109,7 @@ const createDefaultNotificationPreferences = (): NotificationPreferencesState =>
   channels: {
     inApp: true,
     webPush: false,
+    sound: true,
   },
   events: {
     actionRequired: true,
@@ -115,6 +118,25 @@ const createDefaultNotificationPreferences = (): NotificationPreferencesState =>
   },
 });
 
+const normalizeNotificationPreferences = (
+  preferences?: Partial<NotificationPreferencesState> | null,
+): NotificationPreferencesState => {
+  const defaults = createDefaultNotificationPreferences();
+
+  return {
+    channels: {
+      inApp: preferences?.channels?.inApp ?? defaults.channels.inApp,
+      webPush: preferences?.channels?.webPush ?? defaults.channels.webPush,
+      sound: preferences?.channels?.sound ?? defaults.channels.sound,
+    },
+    events: {
+      actionRequired: preferences?.events?.actionRequired ?? defaults.events.actionRequired,
+      stop: preferences?.events?.stop ?? defaults.events.stop,
+      error: preferences?.events?.error ?? defaults.events.error,
+    },
+  };
+};
+
 export function useSettingsController({ isOpen, initialTab }: UseSettingsControllerArgs) {
   const { isDarkMode, toggleDarkMode } = useTheme() as ThemeContextValue;
   const closeTimerRef = useRef<number | null>(null);
@@ -186,7 +208,7 @@ export function useSettingsController({ isOpen, initialTab }: UseSettingsControl
         if (notificationResponse.ok) {
           const notificationData = await toResponseJson<NotificationPreferencesResponse>(notificationResponse);
           if (notificationData.success && notificationData.preferences) {
-            setNotificationPreferences(notificationData.preferences);
+            setNotificationPreferences(normalizeNotificationPreferences(notificationData.preferences));
           } else {
             setNotificationPreferences(createDefaultNotificationPreferences());
           }
@@ -301,6 +323,10 @@ export function useSettingsController({ isOpen, initialTab }: UseSettingsControl
     void refreshProviderAuthStatuses();
   }, [initialTab, isOpen, loadSettings, refreshProviderAuthStatuses]);
 
+  useEffect(() => {
+    setNotificationSoundEnabled(notificationPreferences.channels.sound);
+  }, [notificationPreferences.channels.sound]);
+
   useEffect(() => {
     localStorage.setItem('codeEditorTheme', codeEditorSettings.theme);
     localStorage.setItem('codeEditorWordWrap', String(codeEditorSettings.wordWrap));

src/components/settings/types/types.ts

@@ -1,4 +1,5 @@
 import type { Dispatch, SetStateAction } from 'react';
+
 import type { LLMProvider } from '../../../types/app';
 import type { ProviderAuthStatus } from '../../provider-auth/types';
 
@@ -29,6 +30,7 @@ export type NotificationPreferencesState = {
   channels: {
     inApp: boolean;
     webPush: boolean;
+    sound: boolean;
   };
   events: {
     actionRequired: boolean;

src/components/settings/view/tabs/NotificationsSettingsTab.tsx

@@ -1,5 +1,8 @@
-import { Bell, BellOff, BellRing, Loader2 } from 'lucide-react';
+import { Bell, BellOff, BellRing, Loader2, Play, Volume2 } from 'lucide-react';
 import { useTranslation } from 'react-i18next';
+
+import { Button } from '../../../../shared/view/ui';
+import { playChatCompletionSound } from '../../../../utils/notificationSound';
 import type { NotificationPreferencesState } from '../../types/types';
 
 type NotificationsSettingsTabProps = {
@@ -82,6 +85,54 @@ export default function NotificationsSettingsTab({
         )}
       </div>
 
+      <div className="space-y-4 rounded-lg border border-border bg-card p-4">
+        <div className="flex flex-col gap-4 sm:flex-row sm:items-center sm:justify-between">
+          <div className="space-y-1">
+            <div className="flex items-center gap-2">
+              <Volume2 className="h-4 w-4 text-blue-600" />
+              <h4 className="font-medium text-foreground">
+                {t('notifications.sound.title', { defaultValue: 'Sound' })}
+              </h4>
+            </div>
+            <p className="text-sm text-muted-foreground">
+              {t('notifications.sound.description', {
+                defaultValue: 'Play a short tone when a chat run finishes.',
+              })}
+            </p>
+          </div>
+
+          <label className="flex shrink-0 items-center gap-2 text-sm text-foreground">
+            <input
+              type="checkbox"
+              checked={notificationPreferences.channels.sound}
+              onChange={(event) =>
+                onNotificationPreferencesChange({
+                  ...notificationPreferences,
+                  channels: {
+                    ...notificationPreferences.channels,
+                    sound: event.target.checked,
+                  },
+                })
+              }
+              className="h-4 w-4"
+            />
+            {t('notifications.sound.enabled', { defaultValue: 'Enabled' })}
+          </label>
+        </div>
+
+        <Button
+          type="button"
+          variant="outline"
+          size="sm"
+          onClick={() => {
+            void playChatCompletionSound({ force: true });
+          }}
+        >
+          <Play className="h-4 w-4" />
+          {t('notifications.sound.test', { defaultValue: 'Test sound' })}
+        </Button>
+      </div>
+
       <div className="space-y-4 bg-card border border-border rounded-lg p-4">
         <h4 className="font-medium text-foreground">{t('notifications.events.title')}</h4>
         <div className="space-y-3">

src/components/shell/hooks/useShellConnection.ts

@@ -2,6 +2,7 @@ import { useCallback, useEffect, useRef, useState } from 'react';
 import type { MutableRefObject } from 'react';
 import type { FitAddon } from '@xterm/addon-fit';
 import type { Terminal } from '@xterm/xterm';
+
 import type { Project, ProjectSession } from '../../../types/app';
 import { TERMINAL_INIT_DELAY_MS } from '../constants/constants';
 import { getShellWebSocketUrl, parseShellMessage, sendSocketMessage } from '../utils/socket';
@@ -31,8 +32,8 @@ type UseShellConnectionResult = {
   isConnected: boolean;
   isConnecting: boolean;
   closeSocket: () => void;
-  connectToShell: () => void;
-  disconnectFromShell: () => void;
+  connectToShell: (options?: { forceRestart?: boolean }) => void;
+  disconnectFromShell: (options?: { suppressAutoConnect?: boolean }) => void;
 };
 
 export function useShellConnection({
@@ -54,6 +55,8 @@ export function useShellConnection({
   const [isConnected, setIsConnected] = useState(false);
   const [isConnecting, setIsConnecting] = useState(false);
   const connectingRef = useRef(false);
+  const forceRestartOnInitRef = useRef(false);
+  const suppressAutoConnectRef = useRef(false);
 
   const handleProcessCompletion = useCallback(
     (output: string) => {
@@ -141,6 +144,8 @@ export function useShellConnection({
             }
 
             currentFitAddon.fit();
+            const forceRestart = forceRestartOnInitRef.current;
+            forceRestartOnInitRef.current = false;
 
             sendSocketMessage(socket, {
               type: 'init',
@@ -152,6 +157,7 @@ export function useShellConnection({
               rows: currentTerminal.rows,
               initialCommand: initialCommandRef.current,
               isPlainShell: isPlainShellRef.current,
+              forceRestart,
             });
           }, TERMINAL_INIT_DELAY_MS);
         };
@@ -177,6 +183,7 @@ export function useShellConnection({
         setIsConnected(false);
         setIsConnecting(false);
         connectingRef.current = false;
+        forceRestartOnInitRef.current = false;
       }
     },
     [
@@ -195,27 +202,40 @@ export function useShellConnection({
     ],
   );
 
-  const connectToShell = useCallback(() => {
+  const connectToShell = useCallback((options?: { forceRestart?: boolean }) => {
     if (!isInitialized || isConnected || isConnecting || connectingRef.current) {
       return;
     }
 
+    forceRestartOnInitRef.current = Boolean(options?.forceRestart);
+    suppressAutoConnectRef.current = false;
     connectingRef.current = true;
     setIsConnecting(true);
     connectWebSocket(true);
   }, [connectWebSocket, isConnected, isConnecting, isInitialized]);
 
-  const disconnectFromShell = useCallback(() => {
+  const disconnectFromShell = useCallback((options?: { suppressAutoConnect?: boolean }) => {
+    if (options?.suppressAutoConnect) {
+      suppressAutoConnectRef.current = true;
+    }
+
     closeSocket();
     clearTerminalScreen();
     setIsConnected(false);
     setIsConnecting(false);
     connectingRef.current = false;
+    forceRestartOnInitRef.current = false;
     setAuthUrl('');
   }, [clearTerminalScreen, closeSocket, setAuthUrl]);
 
   useEffect(() => {
-    if (!autoConnect || !isInitialized || isConnecting || isConnected) {
+    if (
+      !autoConnect ||
+      suppressAutoConnectRef.current ||
+      !isInitialized ||
+      isConnecting ||
+      isConnected
+    ) {
       return;
     }
 

src/components/shell/types/types.ts

@@ -1,6 +1,7 @@
 import type { MutableRefObject, RefObject } from 'react';
 import type { FitAddon } from '@xterm/addon-fit';
 import type { Terminal } from '@xterm/xterm';
+
 import type { Project, ProjectSession } from '../../../types/app';
 
 export type AuthCopyStatus = 'idle' | 'copied' | 'failed';
@@ -15,6 +16,7 @@ export type ShellInitMessage = {
   rows: number;
   initialCommand: string | null | undefined;
   isPlainShell: boolean;
+  forceRestart?: boolean;
 };
 
 export type ShellResizeMessage = {
@@ -69,8 +71,8 @@ export type UseShellRuntimeResult = {
   isConnecting: boolean;
   authUrl: string;
   authUrlVersion: number;
-  connectToShell: () => void;
-  disconnectFromShell: () => void;
+  connectToShell: (options?: { forceRestart?: boolean }) => void;
+  disconnectFromShell: (options?: { suppressAutoConnect?: boolean }) => void;
   openAuthUrlInBrowser: (url?: string) => boolean;
   copyAuthUrlToClipboard: (url?: string) => Promise<boolean>;
 };

src/components/shell/view/Shell.tsx

@@ -1,5 +1,6 @@
 import { useCallback, useEffect, useMemo, useRef, useState } from 'react';
 import { useTranslation } from 'react-i18next';
+
 import '@xterm/xterm/css/xterm.css';
 import type { Project, ProjectSession } from '../../../types/app';
 import {
@@ -13,6 +14,7 @@ import {
 import { useShellRuntime } from '../hooks/useShellRuntime';
 import { sendSocketMessage } from '../utils/socket';
 import { getSessionDisplayName } from '../utils/auth';
+
 import ShellConnectionOverlay from './subcomponents/ShellConnectionOverlay';
 import ShellEmptyState from './subcomponents/ShellEmptyState';
 import ShellHeader from './subcomponents/ShellHeader';
@@ -46,6 +48,8 @@ export default function Shell({
   const [isRestarting, setIsRestarting] = useState(false);
   const [cliPromptOptions, setCliPromptOptions] = useState<CliPromptOption[] | null>(null);
   const promptCheckTimer = useRef<ReturnType<typeof setTimeout> | null>(null);
+  const restartTimerRef = useRef<ReturnType<typeof setTimeout> | null>(null);
+  const restartAfterInitRef = useRef(false);
   const onOutputRef = useRef<(() => void) | null>(null);
 
   const {
@@ -140,6 +144,7 @@ export default function Shell({
   useEffect(() => {
     return () => {
       if (promptCheckTimer.current) clearTimeout(promptCheckTimer.current);
+      if (restartTimerRef.current) clearTimeout(restartTimerRef.current);
     };
   }, []);
 
@@ -190,12 +195,42 @@ export default function Shell({
   );
 
   const handleRestartShell = useCallback(() => {
+    restartAfterInitRef.current = true;
     setIsRestarting(true);
-    window.setTimeout(() => {
+    if (restartTimerRef.current) {
+      clearTimeout(restartTimerRef.current);
+    }
+    restartTimerRef.current = setTimeout(() => {
       setIsRestarting(false);
+      restartTimerRef.current = null;
     }, SHELL_RESTART_DELAY_MS);
   }, []);
 
+  const handleDisconnectShell = useCallback(() => {
+    restartAfterInitRef.current = false;
+    if (restartTimerRef.current) {
+      clearTimeout(restartTimerRef.current);
+      restartTimerRef.current = null;
+    }
+    setIsRestarting(false);
+    disconnectFromShell({ suppressAutoConnect: true });
+  }, [disconnectFromShell]);
+
+  useEffect(() => {
+    if (
+      !restartAfterInitRef.current ||
+      isRestarting ||
+      !isInitialized ||
+      isConnected ||
+      isConnecting
+    ) {
+      return;
+    }
+
+    restartAfterInitRef.current = false;
+    connectToShell({ forceRestart: true });
+  }, [connectToShell, isConnected, isConnecting, isInitialized, isRestarting]);
+
   if (!selectedProject) {
     return (
       <ShellEmptyState
@@ -254,7 +289,7 @@ export default function Shell({
         isRestarting={isRestarting}
         hasSession={Boolean(selectedSession)}
         sessionDisplayNameShort={sessionDisplayNameShort}
-        onDisconnect={disconnectFromShell}
+        onDisconnect={handleDisconnectShell}
         onRestart={handleRestartShell}
         statusNewSessionText={t('shell.status.newSession')}
         statusInitializingText={t('shell.status.initializing')}
@@ -263,7 +298,7 @@ export default function Shell({
         disconnectTitle={t('shell.actions.disconnectTitle')}
         restartLabel={t('shell.actions.restart')}
         restartTitle={t('shell.actions.restartTitle')}
-        disableRestart={isRestarting || isConnected}
+        disableRestart={isRestarting || !isInitialized}
       />
 
       <div className="relative flex-1 overflow-hidden p-2">
@@ -281,7 +316,7 @@ export default function Shell({
             connectLabel={t('shell.actions.connect')}
             connectTitle={t('shell.actions.connectTitle')}
             connectingLabel={t('shell.connecting')}
-            onConnect={connectToShell}
+            onConnect={handleRestartShell}
           />
         )}
 

src/components/shell/view/subcomponents/ShellConnectionOverlay.tsx

@@ -1,3 +1,5 @@
+import { Loader2, RotateCcw } from 'lucide-react';
+
 type ShellConnectionOverlayProps = {
   mode: 'loading' | 'connect' | 'connecting';
   description: string;
@@ -19,40 +21,42 @@ export default function ShellConnectionOverlay({
 }: ShellConnectionOverlayProps) {
   if (mode === 'loading') {
     return (
-      <div className="absolute inset-0 flex items-center justify-center bg-gray-900 bg-opacity-90">
-        <div className="text-white">{loadingLabel}</div>
+      <div className="absolute inset-0 z-20 flex items-center justify-center bg-gray-950/90">
+        <div className="inline-flex items-center gap-2 text-sm font-medium text-gray-100">
+          <Loader2 className="h-4 w-4 animate-spin text-blue-300" aria-hidden="true" />
+          <span>{loadingLabel}</span>
+        </div>
       </div>
     );
   }
 
   if (mode === 'connect') {
     return (
-      <div className="absolute inset-0 flex items-center justify-center bg-gray-900 bg-opacity-90 p-4">
-        <div className="w-full max-w-sm text-center">
+      <div className="absolute inset-0 z-20 flex items-center justify-center bg-gray-950/90 p-6">
+        <div className="flex w-full max-w-md flex-col items-center gap-3 text-center">
           <button
+            type="button"
             onClick={onConnect}
-            className="flex w-full items-center justify-center space-x-2 rounded-lg bg-green-600 px-6 py-3 text-base font-medium text-white transition-colors hover:bg-green-700 sm:w-auto"
+            className="pointer-events-auto inline-flex min-h-12 w-full max-w-xs cursor-pointer items-center justify-center gap-2 rounded-md bg-emerald-600 px-5 py-3 text-base font-semibold text-white shadow-lg shadow-emerald-950/30 transition-colors hover:bg-emerald-500 focus:outline-none focus:ring-2 focus:ring-emerald-300 focus:ring-offset-2 focus:ring-offset-gray-950 active:bg-emerald-700"
             title={connectTitle}
           >
-            <svg className="h-5 w-5" fill="none" stroke="currentColor" viewBox="0 0 24 24">
-              <path strokeLinecap="round" strokeLinejoin="round" strokeWidth={2} d="M13 10V3L4 14h7v7l9-11h-7z" />
-            </svg>
-            <span>{connectLabel}</span>
+            <RotateCcw className="h-4 w-4" aria-hidden="true" />
+            <span className="min-w-0 truncate">{connectLabel}</span>
           </button>
-          <p className="mt-3 px-2 text-sm text-gray-400">{description}</p>
+          <p className="max-w-md break-words px-2 text-sm leading-6 text-gray-300">{description}</p>
         </div>
       </div>
     );
   }
 
   return (
-    <div className="absolute inset-0 flex items-center justify-center bg-gray-900 bg-opacity-90 p-4">
-      <div className="w-full max-w-sm text-center">
-        <div className="flex items-center justify-center space-x-3 text-yellow-400">
-          <div className="h-6 w-6 animate-spin rounded-full border-2 border-yellow-400 border-t-transparent"></div>
+    <div className="absolute inset-0 z-20 flex items-center justify-center bg-gray-950/90 p-6">
+      <div className="flex w-full max-w-md flex-col items-center gap-3 text-center">
+        <div className="flex items-center justify-center gap-3 text-yellow-300">
+          <Loader2 className="h-5 w-5 animate-spin" aria-hidden="true" />
           <span className="text-base font-medium">{connectingLabel}</span>
         </div>
-        <p className="mt-3 px-2 text-sm text-gray-400">{description}</p>
+        <p className="max-w-md break-words px-2 text-sm leading-6 text-gray-300">{description}</p>
       </div>
     </div>
   );

src/components/shell/view/subcomponents/ShellHeader.tsx

@@ -1,3 +1,5 @@
+import { RotateCcw, X } from 'lucide-react';
+
 type ShellHeaderProps = {
   isConnected: boolean;
   isInitialized: boolean;
@@ -50,34 +52,27 @@ export default function ShellHeader({
           {isRestarting && <span className="text-xs text-blue-400">{statusRestartingText}</span>}
         </div>
 
-        <div className="flex items-center space-x-3">
+        <div className="flex items-center gap-2">
           {isConnected && (
             <button
+              type="button"
               onClick={onDisconnect}
-              className="flex items-center space-x-1 rounded bg-red-600 px-3 py-1 text-xs text-white hover:bg-red-700"
+              className="inline-flex h-8 items-center gap-1.5 rounded-md bg-red-600 px-3 text-xs font-medium text-white transition-colors hover:bg-red-700 focus:outline-none focus:ring-2 focus:ring-red-400/70 focus:ring-offset-2 focus:ring-offset-gray-800"
               title={disconnectTitle}
             >
-              <svg className="h-3 w-3" fill="none" stroke="currentColor" viewBox="0 0 24 24">
-                <path strokeLinecap="round" strokeLinejoin="round" strokeWidth={2} d="M6 18L18 6M6 6l12 12" />
-              </svg>
+              <X className="h-3.5 w-3.5" aria-hidden="true" />
               <span>{disconnectLabel}</span>
             </button>
           )}
 
           <button
+            type="button"
             onClick={onRestart}
             disabled={disableRestart}
-            className="flex items-center space-x-1 text-xs text-gray-400 hover:text-white disabled:cursor-not-allowed disabled:opacity-50"
+            className="inline-flex h-8 items-center gap-1.5 rounded-md border border-gray-600/80 bg-gray-700/70 px-3 text-xs font-medium text-gray-100 transition-colors hover:border-blue-400/70 hover:bg-blue-600/80 hover:text-white focus:outline-none focus:ring-2 focus:ring-blue-400/70 focus:ring-offset-2 focus:ring-offset-gray-800 disabled:cursor-not-allowed disabled:border-transparent disabled:bg-transparent disabled:text-gray-500 disabled:opacity-60"
             title={restartTitle}
           >
-            <svg className="h-3 w-3" fill="none" stroke="currentColor" viewBox="0 0 24 24">
-              <path
-                strokeLinecap="round"
-                strokeLinejoin="round"
-                strokeWidth={2}
-                d="M4 4v5h.582m15.356 2A8.001 8.001 0 004.582 9m0 0H9m11 11v-5h-.581m0 0a8.003 8.003 0 01-15.357-2m15.357 2H15"
-              />
-            </svg>
+            <RotateCcw className={`h-3.5 w-3.5 ${isRestarting ? 'animate-spin' : ''}`} aria-hidden="true" />
             <span>{restartLabel}</span>
           </button>
         </div>