2 Commits

Author	SHA1	Message	Date
Haile	374e9de719	feat: add opencode support (#762) ... * feat: add opencode support * fix: stabilize opencode session startup * fix: /models * fix: improveUI for commands * fix: format commands.js * feat: load models through provider adapters Provider model selection had outgrown a single hardcoded service. The old service mixed shared caching with provider catalogs and CLI lookup details. That made stale model lists more likely as providers changed on separate schedules. Move model discovery behind each provider so lookup lives next to the integration. The shared service now focuses on provider resolution, caching, persistence, and dedupe. Return cache metadata and add bypassCache because model availability changes outside the app. The UI and /models command can show freshness and let users force a provider refresh. Surface model descriptions while keeping fallback catalogs for unavailable CLIs or SDKs. * feat(models): resolve active session models through provider adapters The model inventory command was showing a mix of catalog defaults and composer-local state instead of the model that is actually active for a real provider session. That made /models, /cost, and /status misleading once a session had already started, especially for providers whose effective runtime model can differ from the optimistic model value held in the UI. Introduce an explicit getCurrentActiveModel() contract on IProviderModels so model resolution lives next to each provider's catalog logic and uses the provider-native source of truth: - Claude reads the init event from a resumed stream-json run - Codex reads model from ~/.codex/config.toml - Cursor reads lastUsedModel from the chat store.db - OpenCode reads the persisted session model from opencode.db - Gemini intentionally returns its default because the CLI does not provide a reliable active-session lookup Keep the returned shape intentionally minimal ({ model }). The goal is to expose only what downstream command consumers need and avoid leaking provider-specific metadata into a shared transport shape that would create extra UI coupling and future cleanup cost. Also make command behavior session-aware: when there is no concrete session id, do not spawn provider processes or inspect provider session storage just to answer /models, /cost, or /status. In a new-session view the correct answer is simply the provider default, and doing more work there adds latency and unnecessary side effects for no user value. As part of this, centralize two supporting concerns: - add a shared helper for building the default current-model result from a provider catalog so fallbacks stay aligned with DEFAULT - move leaf-directory validation into shared utils so Cursor session readers and model lookup code enforce the same path-safety rule Tests were expanded to cover both the new service delegation path and the sessionless command behavior, while keeping cache-sensitive tests isolated from persisted host cache state. Why this change: - command output should reflect the model actually driving a session - new-session views should stay fast and side-effect free - provider-specific active-model lookup should not be scattered across routes or UI code - fallback behavior should be explicit, consistent, and limited to the provider default when no true active model can be resolved * feat: support session-scoped model overrides Model selection was acting like a provider-level preference. That made resumed sessions drift back to a default or request-time model. Users expect /models changes made inside a conversation to affect that session. Store explicit session choices in app-owned ~/.cloudcli state. This avoids editing provider transcripts or native provider config. Resolve the effective model before launching each provider runtime. Claude, Cursor, Codex, Gemini, and OpenCode now honor stored resume choices. Expose a backend active-model change endpoint for existing sessions. The models modal can now distinguish default changes from session overrides. It also shows when a selected model will apply on the next response. For Claude, stop probing active model state by resuming with a dummy prompt. Read the indexed JSONL transcript from the end instead. This preserves provider history while honoring /model stdout or model fields. Add service tests for adapter delegation and resume-model precedence. The tests keep cache state, override state, and requested fallback separate. * feat: make command modal more compact * fix: preserve opencode session creation events OpenCode emits the real session id asynchronously on its first JSON output. The runner registered that id from a helper that could not see the spawned process because the process reference was scoped inside the model-resolution callback. That ReferenceError was swallowed by the generic JSON parse fallback, so the client never received session_created. Without that event, a new OpenCode chat stayed on / and the assistant stream was not attached to the new session view. Keep the process reference in the outer spawn scope so registration can update the active-process map and websocket writer as soon as OpenCode announces the session id. Split JSON parsing from event processing so malformed non-JSON output can still stream as raw text, while registration or adapter failures are surfaced as real errors instead of being hidden as assistant content. Add a fake opencode executable regression test to lock in the expected lifecycle ordering: session_created must be sent before live assistant messages, and the same session id must carry through stream_end and complete. * fix: clarify model refresh and onboarding providers OpenCode is now a supported chat provider, but first-run onboarding still only offered Claude, Cursor, Codex, and Gemini. That made OpenCode harder to discover and forced users to finish setup before finding the provider in settings or chat. Adding it to onboarding keeps first-run setup aligned with the providers the application already supports elsewhere. The model refresh control was also doing too much visual work. In the new chat model picker, the previous Hard Refresh label looked like the dialog heading, which made the primary task unclear. Users open that dialog to choose a model; refreshing catalogs is only a secondary maintenance action for stale cached provider model lists. Rename and reposition the refresh affordance so the model picker reads as a model picker first. The copy now explains why catalogs are cached, when a refresh is useful, and that the refresh checks every provider. The /models modal gets the same clarification so both model-selection surfaces describe the cache behavior consistently. * fix: format opencode model catalog labels OpenCode returns provider-prefixed ids directly from the CLI. Passing those ids through as labels made the model picker hard to scan: users saw values like anthropic/claude-3-5-sonnet-20241022 or lowercased, hyphen-split text instead of readable model names. Keep the exact OpenCode id as the option value because that is what the CLI expects, but derive a presentation label for the frontend. The formatter is intentionally generic rather than a catalog of known providers. It handles common identifier structure such as provider/model, hyphen-delimited words, v-prefixed versions, adjacent numeric version tokens, and 8-digit date suffixes. This keeps OpenCode usable as its model list expands across many upstream providers without requiring code changes for every new provider or model family. The description keeps the raw provider-prefixed id visible so users can still confirm the precise model being selected. * feat: add more fallback models for cursor * docs: move model catalog out of shared The model catalog is no longer a frontend/backend runtime contract. Keeping it under shared made ownership misleading. It implied the catalog was application code shared by runtime consumers, even though it now only supports README links and public API documentation. Move the catalog into public so it lives beside the docs surfaces that need it. This gives the API docs a stable, served module and gives README readers a linkable source without suggesting frontend or backend runtime dependency. Render the API docs model list from the exported provider registry instead of a hardcoded Claude/Cursor/Codex subset. That keeps Gemini and OpenCode visible and makes future provider documentation changes flow through one docs-specific file. Update README links, provider maintenance notes, and package files so published artifacts include the standalone docs page and model catalog without relying on the old shared path. * fix: simplify empty-state model selector Keep the provider empty state focused on the setup action users need there: choosing a model. The refresh control, cache timestamp, and refresh explanation made the dialog feel like a cache-management surface. That extra action is out of place in the empty state, where the goal is to start a chat with the selected provider and model. Remove the refresh-specific UI from ProviderSelectionEmptyState and drop the now-unused refresh/cache props from the ChatMessagesPane pass-through. Refresh behavior remains available in the dedicated command result flow.	2026-05-28 10:50:41 +02:00
Haile	49dd3cfb23	Refactor provider runtimes for sessions, auth, and MCP management (#666) ... * feat: implement MCP provider registry and service - Add provider registry to manage LLM providers (Claude, Codex, Cursor, Gemini). - Create provider routes for MCP server operations (list, upsert, delete, run). - Implement MCP service for handling server operations and validations. - Introduce abstract provider class and MCP provider base for shared functionality. - Add tests for MCP server operations across different providers and scopes. - Define shared interfaces and types for MCP functionality. - Implement utility functions for handling JSON config files and API responses. * chore: remove dead code related to MCP server * refactor: put /api/providers in index.js and remove /providers prefix from provider.routes.ts * refactor(settings): move MCP server management into provider module Extract MCP server settings out of the settings controller and agents tab into a dedicated frontend MCP module. The settings UI now delegates MCP rendering and behavior to a single module that only needs the selected provider and current projects. Changes: - Add `src/components/mcp` as the single frontend MCP module - Move MCP server list rendering into `McpServers` - Move MCP add/edit modal into `McpServerFormModal` - Move MCP API/state logic into `useMcpServers` - Move MCP form state/validation logic into `useMcpServerForm` - Add provider-specific MCP constants, types, and formatting helpers - Use the unified `/api/providers/:provider/mcp/servers` API for all providers - Support MCP management for Claude, Cursor, Codex, and Gemini - Remove old settings-owned Claude/Codex MCP modal components - Remove old provider-specific `McpServersContent` branching from settings - Strip MCP server state, fetch, save, delete, and modal ownership from `useSettingsController` - Simplify agents settings props so MCP only receives `selectedProvider` and `currentProjects` - Keep Claude working-directory unsupported while preserving cwd support for Cursor, Codex, and Gemini - Add progressive MCP loading: - render user/global scope first - load project/local scopes in the background - append project results as they resolve - cache MCP lists briefly to avoid slow tab-switch refetches - ignore stale async responses after provider switches Verification: - `npx eslint src/components/mcp` - `npm run typecheck` - `npm run build:client` * fix(mcp): form with multiline text handling for args, env, headers, and envVars * feat(mcp): add global MCP server creation flow Add a separate global MCP add path in the settings MCP module so users can create one shared MCP server configuration across Claude, Cursor, Codex, and Gemini from the same screen. The provider-specific add flow is still kept next to it because these two actions have different intent. A global MCP server must be constrained to the subset of configuration that every provider can accept, while a provider-specific server can still use that provider's own supported scopes, transports, and fields. Naming the buttons as "Add Global MCP Server" and "Add <Provider> MCP Server" makes that distinction explicit without forcing users to infer it from the selected tab. This also moves the explanatory copy to button hover text to keep the MCP toolbar compact while still documenting the difference between global and provider-only adds at the point of action. Implementation details: - Add global MCP form mode with shared user/project scopes and stdio/http transports. - Submit global creates through `/api/providers/mcp/servers/global`. - Reuse the existing MCP form modal with configurable scopes, transports, labels, and descriptions instead of duplicating form logic. - Disable provider-only fields for the global flow because those fields cannot be safely written to every provider. - Clear the MCP server cache globally after a global add because every provider tab may have changed. - Surface partial global add failures with provider-specific error messages. Validation: - npx eslint src/components/mcp/view/McpServers.tsx - npm run typecheck - npm run build:client * feat: implement platform-specific provider visibility for cursor agent * refactor(providers): centralize message handling in provider module Move provider-specific normalizeMessage and fetchHistory logic out of the legacy server/providers adapters and into the refactored provider classes so callers can depend on the main provider contract instead of parallel adapter plumbing. Add a providers service to resolve concrete providers through the registry and delegate message normalization/history loading from realtime handlers and the unified messages route. Add shared TypeScript message/history types and normalized message helpers so provider implementations and callers use the same contract. Remove the old adapter registry/files now that Claude, Codex, Cursor, and Gemini implement the required behavior directly. * refactor(providers): move auth status checks into provider runtimes Move provider authentication status logic out of the CLI auth route so auth checks live with the provider implementations that understand each provider's install and credential model. Add provider-specific auth runtime classes for Claude, Codex, Cursor, and Gemini, and expose them through the shared provider contract as `provider.auth`. Add a provider auth service that resolves providers through the registry and delegates status checks via `auth.getStatus()`. Keep the existing `/api/cli/<provider>/status` endpoints, but make them thin route adapters over the new provider auth service. This removes duplicated route-local credential parsing and makes auth status a first-class provider capability beside MCP and message handling. * refactor(providers): clarify provider auth and MCP naming Rename provider auth/MCP contracts to remove the overloaded Runtime suffix so the shared interfaces read as stable provider capabilities instead of execution implementation details. Add a consistent provider-first auth class naming convention by renaming ClaudeAuthProvider, CodexAuthProvider, CursorAuthProvider, and GeminiAuthProvider to ClaudeProviderAuth, CodexProviderAuth, CursorProviderAuth, and GeminiProviderAuth. This keeps the provider module API easier to scan and aligns auth naming with the main provider ownership model. * refactor(providers): move session message delegation into sessions service Move provider-backed session history and message normalization calls out of the generic providers service so the service name reflects the behavior it owns. Add a dedicated sessions service for listing session-capable providers, normalizing live provider events, and fetching persisted session history through the provider registry. Update realtime handlers and the unified messages route to depend on `sessionsService` instead of `providersService`. This separates session message operations from other provider concerns such as auth and MCP, keeping the provider services easier to navigate as the module grows. * refactor(providers): move auth status routes under provider API Move provider authentication status endpoints out of the legacy `/api/cli` route namespace so auth status is exposed through the same provider module that owns provider auth and MCP behavior. Add `GET /api/providers/:provider/auth/status` to the provider router and route it through the provider auth service. Remove the old `cli-auth` route file and `/api/cli` mount now that provider auth status is handled by the unified provider API. Update the frontend provider auth endpoint map to call the new provider-scoped routes and rename the endpoint constant to reflect that it is no longer CLI specific. * chore(api): remove unused backend endpoints after MCP audit Remove legacy backend routes that no longer have frontend or internal callers, including the old Claude/Codex MCP APIs, unused Cursor and Codex helper endpoints, stale TaskMaster detection/next/initialize routes, and unused command/project helpers. This reduces duplicated MCP behavior now handled by the provider-based MCP API, shrinks the exposed backend surface, and removes probe/service code that only existed for deleted endpoints. Add an MCP settings API audit document to capture the route-usage analysis and explain why the legacy MCP endpoints were considered safe to remove. * refactor(providers): remove debug logging from Claude authentication status checks * refactor(cursor): lazy-load better-sqlite3 and remove unused type definitions * refactor(cursor): remove SSE from CursorMcpProvider constructor and error message * refactor(auth): standardize API response structure and remove unused error handling * refactor: make providers use dedicated session handling classes * refactor: remove legacy provider selection UI and logic * fix(server/providers): harden and correct session history normalization/pagination Address correctness and safety issues in provider session adapters while preserving existing normalized message shapes. Claude sessions: - Ensure user text content parts generate unique normalized message ids. - Replace duplicate `${baseId}_text` ids with index-suffixed ids to avoid collisions when one user message contains multiple text segments. Cursor sessions: - Add session id sanitization before constructing SQLite paths to prevent path traversal via crafted session ids. - Enforce containment by resolving the computed DB path and asserting it stays under ~/.cursor/chats/<cwdId>. - Refactor blob parsing to a two-pass flow: first build blobMap and collect JSON blobs, then parse binary parent refs against the fully populated map. - Fix pagination semantics so limit=0 returns an empty page instead of full history, with consistent total/hasMore/offset/limit metadata. Gemini sessions: - Honor FetchHistoryOptions pagination by reading limit/offset and slicing normalized history accordingly. - Return consistent hasMore/offset/limit metadata for paged responses. Validation: - eslint passed for touched files. - server TypeScript check passed (tsc --noEmit -p server/tsconfig.json). ---------	2026-04-21 14:38:51 +02:00