Merge branch 'main' into refactor/remove-token-usage-logic

refactor: removed the token calculator logic
The information given was not accurate for the model providers
2026-04-11 16:11:31 +00:00 · 2026-04-10 15:30:43 +02:00 · 2026-04-10 16:21:01 +03:00
38 changed files with 875 additions and 573 deletions
--- a/public/api-docs.html
+++ b/public/api-docs.html
@@ -585,7 +585,7 @@
                            <p>Server-sent events (SSE) format with real-time updates. Content-Type: <code>text/event-stream</code></p>

                            <h4>Response (Non-Streaming)</h4>
-                            <p>JSON object containing session details, assistant messages only (filtered), and token usage summary. Content-Type: <code>application/json</code></p>
+                            <p>JSON object containing session details and assistant messages only (filtered). Content-Type: <code>application/json</code></p>

                            <h4>Error Response</h4>
                            <p>Returns error details with appropriate HTTP status code.</p>
@@ -674,21 +674,10 @@ data: {"type":"done"}</code></pre>
            "type": "text",
            "text": "I've completed the task..."
          }
-        ],
-        "usage": {
-          "input_tokens": 150,
-          "output_tokens": 50
-        }
+        ]
      }
    }
  ],
-  "tokens": {
-    "inputTokens": 150,
-    "outputTokens": 50,
-    "cacheReadTokens": 0,
-    "cacheCreationTokens": 0,
-    "totalTokens": 200
-  },
  "projectPath": "/path/to/project",
  "branch": {
    "name": "fix-authentication-bug-abc123",
--- a/server/claude-sdk.js
+++ b/server/claude-sdk.js
@@ -274,46 +274,6 @@ function transformMessage(sdkMessage) {
  return sdkMessage;
 }

-/**
- * Extracts token usage from SDK result messages
- * @param {Object} resultMessage - SDK result message
- * @returns {Object|null} Token budget object or null
- */
-function extractTokenBudget(resultMessage) {
-  if (resultMessage.type !== 'result' || !resultMessage.modelUsage) {
-    return null;
-  }
-
-  // Get the first model's usage data
-  const modelKey = Object.keys(resultMessage.modelUsage)[0];
-  const modelData = resultMessage.modelUsage[modelKey];
-
-  if (!modelData) {
-    return null;
-  }
-
-  // Use cumulative tokens if available (tracks total for the session)
-  // Otherwise fall back to per-request tokens
-  const inputTokens = modelData.cumulativeInputTokens || modelData.inputTokens || 0;
-  const outputTokens = modelData.cumulativeOutputTokens || modelData.outputTokens || 0;
-  const cacheReadTokens = modelData.cumulativeCacheReadInputTokens || modelData.cacheReadInputTokens || 0;
-  const cacheCreationTokens = modelData.cumulativeCacheCreationInputTokens || modelData.cacheCreationInputTokens || 0;
-
-  // Total used = input + output + cache tokens
-  const totalUsed = inputTokens + outputTokens + cacheReadTokens + cacheCreationTokens;
-
-  // Use configured context window budget from environment (default 160000)
-  // This is the user's budget limit, not the model's context window
-  const contextWindow = parseInt(process.env.CONTEXT_WINDOW) || 160000;
-
-  // Token calc logged via token-budget WS event
-
-  return {
-    used: totalUsed,
-    total: contextWindow
-  };
-}
-
 /**
 * Handles image processing for SDK queries
 * Saves base64 images to temporary files and returns modified prompt with file paths
@@ -657,18 +617,6 @@ async function queryClaudeSDK(command, options = {}, ws) {
        }
        ws.send(msg);
      }
-
-      // Extract and send token budget updates from result messages
-      if (message.type === 'result') {
-        const models = Object.keys(message.modelUsage || {});
-        if (models.length > 0) {
-          // Model info available in result message
-        }
-        const tokenBudgetData = extractTokenBudget(message);
-        if (tokenBudgetData) {
-          ws.send(createNormalizedMessage({ kind: 'status', text: 'token_budget', tokenBudget: tokenBudgetData, sessionId: capturedSessionId || sessionId || null, provider: 'claude' }));
-        }
-      }
    }

    // Clean up session on completion
--- a/server/index.js
+++ b/server/index.js
@@ -1984,6 +1984,155 @@ function handleShellConnection(ws) {
        console.error('[ERROR] Shell WebSocket error:', error);
    });
 }
+// Audio transcription endpoint
+app.post('/api/transcribe', authenticateToken, async (req, res) => {
+    try {
+        const multer = (await import('multer')).default;
+        const upload = multer({ storage: multer.memoryStorage() });
+
+        // Handle multipart form data
+        upload.single('audio')(req, res, async (err) => {
+            if (err) {
+                return res.status(400).json({ error: 'Failed to process audio file' });
+            }
+
+            if (!req.file) {
+                return res.status(400).json({ error: 'No audio file provided' });
+            }
+
+            const apiKey = process.env.OPENAI_API_KEY;
+            if (!apiKey) {
+                return res.status(500).json({ error: 'OpenAI API key not configured. Please set OPENAI_API_KEY in server environment.' });
+            }
+
+            try {
+                // Create form data for OpenAI
+                const FormData = (await import('form-data')).default;
+                const formData = new FormData();
+                formData.append('file', req.file.buffer, {
+                    filename: req.file.originalname,
+                    contentType: req.file.mimetype
+                });
+                formData.append('model', 'whisper-1');
+                formData.append('response_format', 'json');
+                formData.append('language', 'en');
+
+                // Make request to OpenAI
+                const response = await fetch('https://api.openai.com/v1/audio/transcriptions', {
+                    method: 'POST',
+                    headers: {
+                        'Authorization': `Bearer ${apiKey}`,
+                        ...formData.getHeaders()
+                    },
+                    body: formData
+                });
+
+                if (!response.ok) {
+                    const errorData = await response.json().catch(() => ({}));
+                    throw new Error(errorData.error?.message || `Whisper API error: ${response.status}`);
+                }
+
+                const data = await response.json();
+                let transcribedText = data.text || '';
+
+                // Check if enhancement mode is enabled
+                const mode = req.body.mode || 'default';
+
+                // If no transcribed text, return empty
+                if (!transcribedText) {
+                    return res.json({ text: '' });
+                }
+
+                // If default mode, return transcribed text without enhancement
+                if (mode === 'default') {
+                    return res.json({ text: transcribedText });
+                }
+
+                // Handle different enhancement modes
+                try {
+                    const OpenAI = (await import('openai')).default;
+                    const openai = new OpenAI({ apiKey });
+
+                    let prompt, systemMessage, temperature = 0.7, maxTokens = 800;
+
+                    switch (mode) {
+                        case 'prompt':
+                            systemMessage = 'You are an expert prompt engineer who creates clear, detailed, and effective prompts.';
+                            prompt = `You are an expert prompt engineer. Transform the following rough instruction into a clear, detailed, and context-aware AI prompt.
+
+Your enhanced prompt should:
+1. Be specific and unambiguous
+2. Include relevant context and constraints
+3. Specify the desired output format
+4. Use clear, actionable language
+5. Include examples where helpful
+6. Consider edge cases and potential ambiguities
+
+Transform this rough instruction into a well-crafted prompt:
+"${transcribedText}"
+
+Enhanced prompt:`;
+                            break;
+
+                        case 'vibe':
+                        case 'instructions':
+                        case 'architect':
+                            systemMessage = 'You are a helpful assistant that formats ideas into clear, actionable instructions for AI agents.';
+                            temperature = 0.5; // Lower temperature for more controlled output
+                            prompt = `Transform the following idea into clear, well-structured instructions that an AI agent can easily understand and execute.
+
+IMPORTANT RULES:
+- Format as clear, step-by-step instructions
+- Add reasonable implementation details based on common patterns
+- Only include details directly related to what was asked
+- Do NOT add features or functionality not mentioned
+- Keep the original intent and scope intact
+- Use clear, actionable language an agent can follow
+
+Transform this idea into agent-friendly instructions:
+"${transcribedText}"
+
+Agent instructions:`;
+                            break;
+
+                        default:
+                            // No enhancement needed
+                            break;
+                    }
+
+                    // Only make GPT call if we have a prompt
+                    if (prompt) {
+                        const completion = await openai.chat.completions.create({
+                            model: 'gpt-4o-mini',
+                            messages: [
+                                { role: 'system', content: systemMessage },
+                                { role: 'user', content: prompt }
+                            ],
+                            temperature: temperature,
+                            max_tokens: maxTokens
+                        });
+
+                        transcribedText = completion.choices[0].message.content || transcribedText;
+                    }
+
+                } catch (gptError) {
+                    console.error('GPT processing error:', gptError);
+                    // Fall back to original transcription if GPT fails
+                }
+
+                res.json({ text: transcribedText });
+
+            } catch (error) {
+                console.error('Transcription error:', error);
+                res.status(500).json({ error: error.message });
+            }
+        });
+    } catch (error) {
+        console.error('Endpoint error:', error);
+        res.status(500).json({ error: 'Internal server error' });
+    }
+});
+
 // Image upload endpoint
 app.post('/api/projects/:projectName/upload-images', authenticateToken, async (req, res) => {
    try {
@@ -2069,194 +2218,6 @@ app.post('/api/projects/:projectName/upload-images', authenticateToken, async (r
    }
 });

-// Get token usage for a specific session
-app.get('/api/projects/:projectName/sessions/:sessionId/token-usage', authenticateToken, async (req, res) => {
-    try {
-        const { projectName, sessionId } = req.params;
-        const { provider = 'claude' } = req.query;
-        const homeDir = os.homedir();
-
-        // Allow only safe characters in sessionId
-        const safeSessionId = String(sessionId).replace(/[^a-zA-Z0-9._-]/g, '');
-        if (!safeSessionId || safeSessionId !== String(sessionId)) {
-            return res.status(400).json({ error: 'Invalid sessionId' });
-        }
-
-        // Handle Cursor sessions - they use SQLite and don't have token usage info
-        if (provider === 'cursor') {
-            return res.json({
-                used: 0,
-                total: 0,
-                breakdown: { input: 0, cacheCreation: 0, cacheRead: 0 },
-                unsupported: true,
-                message: 'Token usage tracking not available for Cursor sessions'
-            });
-        }
-
-        // Handle Gemini sessions - they are raw logs in our current setup
-        if (provider === 'gemini') {
-            return res.json({
-                used: 0,
-                total: 0,
-                breakdown: { input: 0, cacheCreation: 0, cacheRead: 0 },
-                unsupported: true,
-                message: 'Token usage tracking not available for Gemini sessions'
-            });
-        }
-
-        // Handle Codex sessions
-        if (provider === 'codex') {
-            const codexSessionsDir = path.join(homeDir, '.codex', 'sessions');
-
-            // Find the session file by searching for the session ID
-            const findSessionFile = async (dir) => {
-                try {
-                    const entries = await fsPromises.readdir(dir, { withFileTypes: true });
-                    for (const entry of entries) {
-                        const fullPath = path.join(dir, entry.name);
-                        if (entry.isDirectory()) {
-                            const found = await findSessionFile(fullPath);
-                            if (found) return found;
-                        } else if (entry.name.includes(safeSessionId) && entry.name.endsWith('.jsonl')) {
-                            return fullPath;
-                        }
-                    }
-                } catch (error) {
-                    // Skip directories we can't read
-                }
-                return null;
-            };
-
-            const sessionFilePath = await findSessionFile(codexSessionsDir);
-
-            if (!sessionFilePath) {
-                return res.status(404).json({ error: 'Codex session file not found', sessionId: safeSessionId });
-            }
-
-            // Read and parse the Codex JSONL file
-            let fileContent;
-            try {
-                fileContent = await fsPromises.readFile(sessionFilePath, 'utf8');
-            } catch (error) {
-                if (error.code === 'ENOENT') {
-                    return res.status(404).json({ error: 'Session file not found', path: sessionFilePath });
-                }
-                throw error;
-            }
-            const lines = fileContent.trim().split('\n');
-            let totalTokens = 0;
-            let contextWindow = 200000; // Default for Codex/OpenAI
-
-            // Find the latest token_count event with info (scan from end)
-            for (let i = lines.length - 1; i >= 0; i--) {
-                try {
-                    const entry = JSON.parse(lines[i]);
-
-                    // Codex stores token info in event_msg with type: "token_count"
-                    if (entry.type === 'event_msg' && entry.payload?.type === 'token_count' && entry.payload?.info) {
-                        const tokenInfo = entry.payload.info;
-                        if (tokenInfo.total_token_usage) {
-                            totalTokens = tokenInfo.total_token_usage.total_tokens || 0;
-                        }
-                        if (tokenInfo.model_context_window) {
-                            contextWindow = tokenInfo.model_context_window;
-                        }
-                        break; // Stop after finding the latest token count
-                    }
-                } catch (parseError) {
-                    // Skip lines that can't be parsed
-                    continue;
-                }
-            }
-
-            return res.json({
-                used: totalTokens,
-                total: contextWindow
-            });
-        }
-
-        // Handle Claude sessions (default)
-        // Extract actual project path
-        let projectPath;
-        try {
-            projectPath = await extractProjectDirectory(projectName);
-        } catch (error) {
-            console.error('Error extracting project directory:', error);
-            return res.status(500).json({ error: 'Failed to determine project path' });
-        }
-
-        // Construct the JSONL file path
-        // Claude stores session files in ~/.claude/projects/[encoded-project-path]/[session-id].jsonl
-        // The encoding replaces any non-alphanumeric character (except -) with -
-        const encodedPath = projectPath.replace(/[^a-zA-Z0-9-]/g, '-');
-        const projectDir = path.join(homeDir, '.claude', 'projects', encodedPath);
-
-        const jsonlPath = path.join(projectDir, `${safeSessionId}.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' });
-        }
-
-        // Read and parse the JSONL file
-        let fileContent;
-        try {
-            fileContent = await fsPromises.readFile(jsonlPath, 'utf8');
-        } catch (error) {
-            if (error.code === 'ENOENT') {
-                return res.status(404).json({ error: 'Session file not found', path: jsonlPath });
-            }
-            throw error; // Re-throw other errors to be caught by outer try-catch
-        }
-        const lines = fileContent.trim().split('\n');
-
-        const parsedContextWindow = parseInt(process.env.CONTEXT_WINDOW, 10);
-        const contextWindow = Number.isFinite(parsedContextWindow) ? parsedContextWindow : 160000;
-        let inputTokens = 0;
-        let cacheCreationTokens = 0;
-        let cacheReadTokens = 0;
-
-        // Find the latest assistant message with usage data (scan from end)
-        for (let i = lines.length - 1; i >= 0; i--) {
-            try {
-                const entry = JSON.parse(lines[i]);
-
-                // Only count assistant messages which have usage data
-                if (entry.type === 'assistant' && entry.message?.usage) {
-                    const usage = entry.message.usage;
-
-                    // Use token counts from latest assistant message only
-                    inputTokens = usage.input_tokens || 0;
-                    cacheCreationTokens = usage.cache_creation_input_tokens || 0;
-                    cacheReadTokens = usage.cache_read_input_tokens || 0;
-
-                    break; // Stop after finding the latest assistant message
-                }
-            } catch (parseError) {
-                // Skip lines that can't be parsed
-                continue;
-            }
-        }
-
-        // Calculate total context usage (excluding output_tokens, as per ccusage)
-        const totalUsed = inputTokens + cacheCreationTokens + cacheReadTokens;
-
-        res.json({
-            used: totalUsed,
-            total: contextWindow,
-            breakdown: {
-                input: inputTokens,
-                cacheCreation: cacheCreationTokens,
-                cacheRead: cacheReadTokens
-            }
-        });
-    } catch (error) {
-        console.error('Error reading session token usage:', error);
-        res.status(500).json({ error: 'Failed to read session token usage' });
-    }
-});
-
 // Serve React app for all other routes (excluding static files)
 app.get('*', (req, res) => {
    // Skip requests for static assets (files with extensions)
--- a/server/openai-codex.js
+++ b/server/openai-codex.js
@@ -129,8 +129,7 @@ function transformCodexEvent(event) {

    case 'turn.completed':
      return {
-        type: 'turn_complete',
-        usage: event.usage
+        type: 'turn_complete'
      };

    case 'turn.failed':
@@ -279,12 +278,6 @@ export async function queryCodex(command, options = {}, ws) {
          error: terminalFailure
        });
      }
-
-      // Extract and send token usage if available (normalized to match Claude format)
-      if (event.type === 'turn.completed' && event.usage) {
-        const totalTokens = (event.usage.input_tokens || 0) + (event.usage.output_tokens || 0);
-        sendMessage(ws, createNormalizedMessage({ kind: 'status', text: 'token_budget', tokenBudget: { used: totalTokens, total: 200000 }, sessionId: currentSessionId, provider: 'codex' }));
-      }
    }

    // Send completion event
--- a/server/projects.js
+++ b/server/projects.js
@@ -1618,7 +1618,6 @@ async function getCodexSessionMessages(sessionId, limit = null, offset = 0) {
    }

    const messages = [];
-    let tokenUsage = null;
    const fileStream = fsSync.createReadStream(sessionFilePath);
    const rl = readline.createInterface({
      input: fileStream,
@@ -1647,17 +1646,6 @@ async function getCodexSessionMessages(sessionId, limit = null, offset = 0) {
        try {
          const entry = JSON.parse(line);

-          // Extract token usage from token_count events (keep latest)
-          if (entry.type === 'event_msg' && entry.payload?.type === 'token_count' && entry.payload?.info) {
-            const info = entry.payload.info;
-            if (info.total_token_usage) {
-              tokenUsage = {
-                used: info.total_token_usage.total_tokens || 0,
-                total: info.model_context_window || 200000
-              };
-            }
-          }
-          
          // Use event_msg.user_message for user-visible inputs.
          if (entry.type === 'event_msg' && isVisibleCodexUserMessage(entry.payload)) {
            messages.push({
@@ -1820,11 +1808,10 @@ async function getCodexSessionMessages(sessionId, limit = null, offset = 0) {
        hasMore,
        offset,
        limit,
-        tokenUsage
      };
    }

-    return { messages, tokenUsage };
+    return { messages };

  } catch (error) {
    console.error(`Error reading Codex session messages for ${sessionId}:`, error);
--- a/server/providers/codex/adapter.js
+++ b/server/providers/codex/adapter.js
@@ -214,7 +214,6 @@ export const codexAdapter = {
    const rawMessages = Array.isArray(result) ? result : (result.messages || []);
    const total = Array.isArray(result) ? rawMessages.length : (result.total || 0);
    const hasMore = Array.isArray(result) ? false : Boolean(result.hasMore);
-    const tokenUsage = result.tokenUsage || null;

    const normalized = [];
    for (const raw of rawMessages) {
@@ -242,7 +241,6 @@ export const codexAdapter = {
      hasMore,
      offset,
      limit,
-      tokenUsage,
    };
  },
 };
--- a/server/providers/gemini/adapter.js
+++ b/server/providers/gemini/adapter.js
@@ -53,14 +53,7 @@ export function normalizeMessage(raw, sessionId) {
  }

  if (raw.type === 'result') {
-    const msgs = [createNormalizedMessage({ sessionId, timestamp: ts, provider: PROVIDER, kind: 'stream_end' })];
-    if (raw.stats?.total_tokens) {
-      msgs.push(createNormalizedMessage({
-        sessionId, timestamp: ts, provider: PROVIDER,
-        kind: 'status', text: 'Complete', tokens: raw.stats.total_tokens, canInterrupt: false,
-      }));
-    }
-    return msgs;
+    return [createNormalizedMessage({ sessionId, timestamp: ts, provider: PROVIDER, kind: 'stream_end' })];
  }

  if (raw.type === 'error') {
--- a/server/providers/types.js
+++ b/server/providers/types.js
@@ -41,7 +41,7 @@
 * - stream_end:           (no extra fields)
 * - error:                content
 * - complete:             (no extra fields)
- * - status:               text, tokens?, canInterrupt?
+ * - status:               text, canInterrupt?
 * - permission_request:   requestId, toolName, input, context?
 * - permission_cancelled: requestId
 * - session_created:      newSessionId
@@ -66,7 +66,6 @@
 * @property {boolean} hasMore - Whether more messages exist before the current page
 * @property {number} offset - Current offset
 * @property {number|null} limit - Page size used
- * @property {object} [tokenUsage] - Token usage data (provider-specific)
 */

 // ─── Provider Adapter Interface ──────────────────────────────────────────────
--- a/server/routes/agent.js
+++ b/server/routes/agent.js
@@ -546,7 +546,12 @@ class ResponseCollector {
          const parsed = JSON.parse(msg);
          // Only include claude-response messages with assistant type
          if (parsed.type === 'claude-response' && parsed.data && parsed.data.type === 'assistant') {
-            assistantMessages.push(parsed.data);
+            const assistantMessage = { ...parsed.data };
+            if (assistantMessage.message?.usage) {
+              assistantMessage.message = { ...assistantMessage.message };
+              delete assistantMessage.message.usage;
+            }
+            assistantMessages.push(assistantMessage);
          }
        } catch (e) {
          // Not JSON, skip
@@ -556,49 +561,6 @@ class ResponseCollector {

    return assistantMessages;
  }
-
-  /**
-   * Calculate total tokens from all messages
-   */
-  getTotalTokens() {
-    let totalInput = 0;
-    let totalOutput = 0;
-    let totalCacheRead = 0;
-    let totalCacheCreation = 0;
-
-    for (const msg of this.messages) {
-      let data = msg;
-
-      // Parse if string
-      if (typeof msg === 'string') {
-        try {
-          data = JSON.parse(msg);
-        } catch (e) {
-          continue;
-        }
-      }
-
-      // Extract usage from claude-response messages
-      if (data && data.type === 'claude-response' && data.data) {
-        const msgData = data.data;
-        if (msgData.message && msgData.message.usage) {
-          const usage = msgData.message.usage;
-          totalInput += usage.input_tokens || 0;
-          totalOutput += usage.output_tokens || 0;
-          totalCacheRead += usage.cache_read_input_tokens || 0;
-          totalCacheCreation += usage.cache_creation_input_tokens || 0;
-        }
-      }
-    }
-
-    return {
-      inputTokens: totalInput,
-      outputTokens: totalOutput,
-      cacheReadTokens: totalCacheRead,
-      cacheCreationTokens: totalCacheCreation,
-      totalTokens: totalInput + totalOutput + totalCacheRead + totalCacheCreation
-    };
-  }
 }

 // ===============================
@@ -789,13 +751,6 @@ class ResponseCollector {
 *     success: true,
 *     sessionId: "session-123",
 *     messages: [...],        // Assistant messages only (filtered)
- *     tokens: {
- *       inputTokens: 150,
- *       outputTokens: 50,
- *       cacheReadTokens: 0,
- *       cacheCreationTokens: 0,
- *       totalTokens: 200
- *     },
 *     projectPath: "/path/to/project",
 *     branch: {               // Only if createBranch=true
 *       name: "feature/xyz",
@@ -1173,15 +1128,13 @@ router.post('/', validateExternalApiKey, async (req, res) => {
      // Streaming mode: end the SSE stream
      writer.end();
    } else {
-      // Non-streaming mode: send filtered messages and token summary as JSON
+      // Non-streaming mode: send filtered messages as JSON
      const assistantMessages = writer.getAssistantMessages();
-      const tokenSummary = writer.getTotalTokens();

      const response = {
        success: true,
        sessionId: writer.getSessionId(),
        messages: assistantMessages,
-        tokens: tokenSummary,
        projectPath: finalProjectPath
      };

--- a/server/routes/commands.js
+++ b/server/routes/commands.js
@@ -97,12 +97,6 @@ const builtInCommands = [
    namespace: 'builtin',
    metadata: { type: 'builtin' }
  },
-  {
-    name: '/cost',
-    description: 'Display token usage and cost information',
-    namespace: 'builtin',
-    metadata: { type: 'builtin' }
-  },
  {
    name: '/memory',
    description: 'Open CLAUDE.md memory file for editing',
@@ -209,86 +203,6 @@ Custom commands can be created in:
    };
  },

-  '/cost': async (args, context) => {
-    const tokenUsage = context?.tokenUsage || {};
-    const provider = context?.provider || 'claude';
-    const model =
-      context?.model ||
-      (provider === 'cursor'
-        ? CURSOR_MODELS.DEFAULT
-        : provider === 'codex'
-          ? CODEX_MODELS.DEFAULT
-          : CLAUDE_MODELS.DEFAULT);
-
-    const used = Number(tokenUsage.used ?? tokenUsage.totalUsed ?? tokenUsage.total_tokens ?? 0) || 0;
-    const total =
-      Number(
-        tokenUsage.total ??
-          tokenUsage.contextWindow ??
-          parseInt(process.env.CONTEXT_WINDOW || '160000', 10),
-      ) || 160000;
-    const percentage = total > 0 ? Number(((used / total) * 100).toFixed(1)) : 0;
-
-    const inputTokensRaw =
-      Number(
-        tokenUsage.inputTokens ??
-          tokenUsage.input ??
-          tokenUsage.cumulativeInputTokens ??
-          tokenUsage.promptTokens ??
-          0,
-      ) || 0;
-    const outputTokens =
-      Number(
-        tokenUsage.outputTokens ??
-          tokenUsage.output ??
-          tokenUsage.cumulativeOutputTokens ??
-          tokenUsage.completionTokens ??
-          0,
-      ) || 0;
-    const cacheTokens =
-      Number(
-        tokenUsage.cacheReadTokens ??
-          tokenUsage.cacheCreationTokens ??
-          tokenUsage.cacheTokens ??
-          tokenUsage.cachedTokens ??
-          0,
-      ) || 0;
-
-    // If we only have total used tokens, treat them as input for display/estimation.
-    const inputTokens =
-      inputTokensRaw > 0 || outputTokens > 0 || cacheTokens > 0 ? inputTokensRaw + cacheTokens : used;
-
-    // Rough default rates by provider (USD / 1M tokens).
-    const pricingByProvider = {
-      claude: { input: 3, output: 15 },
-      cursor: { input: 3, output: 15 },
-      codex: { input: 1.5, output: 6 },
-    };
-    const rates = pricingByProvider[provider] || pricingByProvider.claude;
-
-    const inputCost = (inputTokens / 1_000_000) * rates.input;
-    const outputCost = (outputTokens / 1_000_000) * rates.output;
-    const totalCost = inputCost + outputCost;
-
-    return {
-      type: 'builtin',
-      action: 'cost',
-      data: {
-        tokenUsage: {
-          used,
-          total,
-          percentage,
-        },
-        cost: {
-          input: inputCost.toFixed(4),
-          output: outputCost.toFixed(4),
-          total: totalCost.toFixed(4),
-        },
-        model,
-      },
-    };
-  },
-
  '/status': async (args, context) => {
    // Read version from package.json
    const packageJsonPath = path.join(path.dirname(__dirname), '..', 'package.json');
--- a/src/components/chat/hooks/useChatComposerState.ts
+++ b/src/components/chat/hooks/useChatComposerState.ts
@@ -42,7 +42,6 @@ interface UseChatComposerStateArgs {
  geminiModel: string;
  isLoading: boolean;
  canAbortSession: boolean;
-  tokenBudget: Record<string, unknown> | null;
  sendMessage: (message: unknown) => void;
  sendByCtrlEnter?: boolean;
  onSessionActive?: (sessionId?: string | null) => void;
@@ -57,7 +56,7 @@ interface UseChatComposerStateArgs {
  rewindMessages: (count: number) => void;
  setIsLoading: (loading: boolean) => void;
  setCanAbortSession: (canAbort: boolean) => void;
-  setClaudeStatus: (status: { text: string; tokens: number; can_interrupt: boolean } | null) => void;
+  setClaudeStatus: (status: { text: string; can_interrupt: boolean } | null) => void;
  setIsUserScrolledUp: (isScrolledUp: boolean) => void;
  setPendingPermissionRequests: Dispatch<SetStateAction<PendingPermissionRequest[]>>;
 }
@@ -114,7 +113,6 @@ export function useChatComposerState({
  geminiModel,
  isLoading,
  canAbortSession,
-  tokenBudget,
  sendMessage,
  sendByCtrlEnter,
  onSessionActive,
@@ -176,12 +174,6 @@ export function useChatComposerState({
          });
          break;

-        case 'cost': {
-          const costMessage = `**Token Usage**: ${data.tokenUsage.used.toLocaleString()} / ${data.tokenUsage.total.toLocaleString()} (${data.tokenUsage.percentage}%)\n\n**Estimated Cost**:\n- Input: $${data.cost.input}\n- Output: $${data.cost.output}\n- **Total**: $${data.cost.total}\n\n**Model**: ${data.model}`;
-          addMessage({ type: 'assistant', content: costMessage, timestamp: Date.now() });
-          break;
-        }
-
        case 'status': {
          const statusMessage = `**System Status**\n\n- Version: ${data.version}\n- Uptime: ${data.uptime}\n- Model: ${data.model}\n- Provider: ${data.provider}\n- Node.js: ${data.nodeVersion}\n- Platform: ${data.platform}`;
          addMessage({ type: 'assistant', content: statusMessage, timestamp: Date.now() });
@@ -282,7 +274,6 @@ export function useChatComposerState({
          sessionId: currentSessionId,
          provider,
          model: provider === 'cursor' ? cursorModel : provider === 'codex' ? codexModel : provider === 'gemini' ? geminiModel : claudeModel,
-          tokenUsage: tokenBudget,
        };

        const response = await authenticatedFetch('/api/commands/execute', {
@@ -339,7 +330,6 @@ export function useChatComposerState({
      provider,
      selectedProject,
      addMessage,
-      tokenBudget,
    ],
  );

@@ -543,7 +533,6 @@ export function useChatComposerState({
      setCanAbortSession(true);
      setClaudeStatus({
        text: 'Processing',
-        tokens: 0,
        can_interrupt: true,
      });

@@ -878,6 +867,30 @@ export function useChatComposerState({
    });
  }, [canAbortSession, currentSessionId, pendingViewSessionRef, provider, selectedSession?.id, sendMessage]);

+  const handleTranscript = useCallback((text: string) => {
+    if (!text.trim()) {
+      return;
+    }
+
+    setInput((previousInput) => {
+      const newInput = previousInput.trim() ? `${previousInput} ${text}` : text;
+      inputValueRef.current = newInput;
+
+      setTimeout(() => {
+        if (!textareaRef.current) {
+          return;
+        }
+
+        textareaRef.current.style.height = 'auto';
+        textareaRef.current.style.height = `${textareaRef.current.scrollHeight}px`;
+        const lineHeight = parseInt(window.getComputedStyle(textareaRef.current).lineHeight);
+        setIsTextareaExpanded(textareaRef.current.scrollHeight > lineHeight * 2);
+      }, 0);
+
+      return newInput;
+    });
+  }, []);
+
  const handleGrantToolPermission = useCallback(
    (suggestion: { entry: string; toolName: string }) => {
      if (!suggestion || provider !== 'claude') {
@@ -970,6 +983,7 @@ export function useChatComposerState({
    syncInputOverlayScroll,
    handleClearInput,
    handleAbortSession,
+    handleTranscript,
    handlePermissionDecision,
    handleGrantToolPermission,
    handleInputFocusChange,
--- a/src/components/chat/hooks/useChatRealtimeHandlers.ts
+++ b/src/components/chat/hooks/useChatRealtimeHandlers.ts
@@ -38,9 +38,7 @@ type LatestChatMessage = {
  provider?: string;
  content?: string;
  text?: string;
-  tokens?: number;
  canInterrupt?: boolean;
-  tokenBudget?: unknown;
  newSessionId?: string;
  aborted?: boolean;
  [key: string]: any;
@@ -55,8 +53,7 @@ interface UseChatRealtimeHandlersArgs {
  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;
+  setClaudeStatus: (status: { text: string; can_interrupt: boolean } | null) => void;
  setPendingPermissionRequests: Dispatch<SetStateAction<PendingPermissionRequest[]>>;
  pendingViewSessionRef: MutableRefObject<PendingViewSession | null>;
  streamBufferRef: MutableRefObject<string>;
@@ -85,7 +82,6 @@ export function useChatRealtimeHandlers({
  setIsLoading,
  setCanAbortSession,
  setClaudeStatus,
-  setTokenBudget,
  setPendingPermissionRequests,
  pendingViewSessionRef,
  streamBufferRef,
@@ -140,7 +136,6 @@ export function useChatRealtimeHandlers({
          if (status) {
            const statusInfo = {
              text: status.text || 'Working...',
-              tokens: status.tokens || 0,
              can_interrupt: status.can_interrupt !== undefined ? status.can_interrupt : true,
            };
            setClaudeStatus(statusInfo);
@@ -311,7 +306,7 @@ export function useChatRealtimeHandlers({
        });
        setIsLoading(true);
        setCanAbortSession(true);
-        setClaudeStatus({ text: 'Waiting for permission', tokens: 0, can_interrupt: true });
+        setClaudeStatus({ text: 'Waiting for permission', can_interrupt: true });
        break;
      }

@@ -323,12 +318,9 @@ export function useChatRealtimeHandlers({
      }

      case 'status': {
-        if (msg.text === 'token_budget' && msg.tokenBudget) {
-          setTokenBudget(msg.tokenBudget as Record<string, unknown>);
-        } else if (msg.text) {
+        if (msg.text) {
          setClaudeStatus({
            text: msg.text,
-            tokens: msg.tokens || 0,
            can_interrupt: msg.canInterrupt !== undefined ? msg.canInterrupt : true,
          });
          setIsLoading(true);
@@ -352,7 +344,6 @@ export function useChatRealtimeHandlers({
    setIsLoading,
    setCanAbortSession,
    setClaudeStatus,
-    setTokenBudget,
    setPendingPermissionRequests,
    pendingViewSessionRef,
    streamBufferRef,
--- a/src/components/chat/hooks/useChatSessionState.ts
+++ b/src/components/chat/hooks/useChatSessionState.ts
@@ -1,6 +1,5 @@
 import { useCallback, useEffect, useLayoutEffect, useMemo, useRef, useState } from 'react';
 import type { MutableRefObject } from 'react';
-import { authenticatedFetch } from '../../../utils/api';
 import type { ChatMessage, Provider } from '../types/types';
 import type { Project, ProjectSession, SessionProvider } from '../../../types/app';
 import { createCachedDiffCalculator, type DiffCalculator } from '../utils/messageTransforms';
@@ -108,9 +107,8 @@ export function useChatSessionState({
  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 [claudeStatus, setClaudeStatus] = useState<{ text: string; can_interrupt: boolean } | null>(null);
  const [allMessagesLoaded, setAllMessagesLoaded] = useState(false);
  const [isLoadingAllMessages, setIsLoadingAllMessages] = useState(false);
  const [loadAllJustFinished, setLoadAllJustFinished] = useState(false);
@@ -319,7 +317,6 @@ export function useChatSessionState({
      messagesOffsetRef.current = 0;
      setHasMoreMessages(false);
      setTotalMessages(0);
-      setTokenBudget(null);
      lastLoadedSessionKeyRef.current = null;
      return;
    }
@@ -355,7 +352,6 @@ export function useChatSessionState({
    if (loadAllFinishedTimerRef.current) clearTimeout(loadAllFinishedTimerRef.current);

    if (sessionChanged) {
-      setTokenBudget(null);
      setIsLoading(false);
    }

@@ -383,7 +379,6 @@ export function useChatSessionState({
      if (slot) {
        setHasMoreMessages(slot.hasMore);
        setTotalMessages(slot.total);
-        if (slot.tokenUsage) setTokenBudget(slot.tokenUsage as Record<string, unknown>);
      }
      setIsLoadingSessionMessages(false);
    }).catch(() => {
@@ -539,31 +534,6 @@ export function useChatSessionState({
  // eslint-disable-next-line react-hooks/exhaustive-deps
  }, [chatMessages.length, isLoadingSessionMessages, searchTarget]);

-  // Token usage fetch for Claude
-  useEffect(() => {
-    if (!selectedProject || !selectedSession?.id || selectedSession.id.startsWith('new-session-')) {
-      setTokenBudget(null);
-      return;
-    }
-    const sessionProvider = selectedSession.__provider || 'claude';
-    if (sessionProvider !== 'claude') return;
-
-    const fetchInitialTokenUsage = async () => {
-      try {
-        const url = `/api/projects/${selectedProject.name}/sessions/${selectedSession.id}/token-usage`;
-        const response = await authenticatedFetch(url);
-        if (response.ok) {
-          setTokenBudget(await response.json());
-        } else {
-          setTokenBudget(null);
-        }
-      } catch (error) {
-        console.error('Failed to fetch initial token usage:', error);
-      }
-    };
-    fetchInitialTokenUsage();
-  }, [selectedProject, selectedSession?.id, selectedSession?.__provider]);
-
  const visibleMessages = useMemo(() => {
    if (chatMessages.length <= visibleMessageCount) return chatMessages;
    return chatMessages.slice(-visibleMessageCount);
@@ -713,8 +683,6 @@ export function useChatSessionState({
    setCanAbortSession,
    isUserScrolledUp,
    setIsUserScrolledUp,
-    tokenBudget,
-    setTokenBudget,
    visibleMessageCount,
    visibleMessages,
    loadEarlierMessages,
--- a/src/components/chat/view/ChatInterface.tsx
+++ b/src/components/chat/view/ChatInterface.tsx
@@ -96,8 +96,6 @@ function ChatInterface({
    setCanAbortSession,
    isUserScrolledUp,
    setIsUserScrolledUp,
-    tokenBudget,
-    setTokenBudget,
    visibleMessageCount,
    visibleMessages,
    loadEarlierMessages,
@@ -165,6 +163,7 @@ function ChatInterface({
    syncInputOverlayScroll,
    handleClearInput,
    handleAbortSession,
+    handleTranscript,
    handlePermissionDecision,
    handleGrantToolPermission,
    handleInputFocusChange,
@@ -182,7 +181,6 @@ function ChatInterface({
    geminiModel,
    isLoading,
    canAbortSession,
-    tokenBudget,
    sendMessage,
    sendByCtrlEnter,
    onSessionActive,
@@ -226,7 +224,6 @@ function ChatInterface({
    setIsLoading,
    setCanAbortSession,
    setClaudeStatus,
-    setTokenBudget,
    setPendingPermissionRequests,
    pendingViewSessionRef,
    streamBufferRef,
@@ -351,7 +348,6 @@ function ChatInterface({
          onModeSwitch={cyclePermissionMode}
          thinkingMode={thinkingMode}
          setThinkingMode={setThinkingMode}
-          tokenBudget={tokenBudget}
          slashCommandsCount={slashCommandsCount}
          onToggleCommandMenu={handleToggleCommandMenu}
          hasInput={Boolean(input.trim())}
@@ -406,6 +402,7 @@ function ChatInterface({
          })}
          isTextareaExpanded={isTextareaExpanded}
          sendByCtrlEnter={sendByCtrlEnter}
+          onTranscript={handleTranscript}
        />
      </div>

--- a/src/components/chat/view/subcomponents/ChatComposer.tsx
+++ b/src/components/chat/view/subcomponents/ChatComposer.tsx
@@ -11,6 +11,7 @@ import type {
  SetStateAction,
  TouchEvent,
 } from 'react';
+import MicButton from '../../../mic-button/view/MicButton';
 import type { PendingPermissionRequest, PermissionMode, Provider } from '../../types/types';
 import CommandMenu from './CommandMenu';
 import ClaudeStatus from './ClaudeStatus';
@@ -40,7 +41,7 @@ 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;
+  claudeStatus: { text: string; can_interrupt: boolean } | null;
  isLoading: boolean;
  onAbortSession: () => void;
  provider: Provider | string;
@@ -48,7 +49,6 @@ interface ChatComposerProps {
  onModeSwitch: () => void;
  thinkingMode: string;
  setThinkingMode: Dispatch<SetStateAction<string>>;
-  tokenBudget: { used?: number; total?: number } | null;
  slashCommandsCount: number;
  onToggleCommandMenu: () => void;
  hasInput: boolean;
@@ -90,6 +90,7 @@ interface ChatComposerProps {
  placeholder: string;
  isTextareaExpanded: boolean;
  sendByCtrlEnter?: boolean;
+  onTranscript: (text: string) => void;
 }

 export default function ChatComposer({
@@ -104,7 +105,6 @@ export default function ChatComposer({
  onModeSwitch,
  thinkingMode,
  setThinkingMode,
-  tokenBudget,
  slashCommandsCount,
  onToggleCommandMenu,
  hasInput,
@@ -146,6 +146,7 @@ export default function ChatComposer({
  placeholder,
  isTextareaExpanded,
  sendByCtrlEnter,
+  onTranscript,
 }: ChatComposerProps) {
  const { t } = useTranslation('chat');
  const textareaRect = textareaRef.current?.getBoundingClientRect();
@@ -191,7 +192,6 @@ export default function ChatComposer({
          provider={provider}
          thinkingMode={thinkingMode}
          setThinkingMode={setThinkingMode}
-          tokenBudget={tokenBudget}
          slashCommandsCount={slashCommandsCount}
          onToggleCommandMenu={onToggleCommandMenu}
          hasInput={hasInput}
@@ -318,6 +318,10 @@ export default function ChatComposer({
              </svg>
            </button>

+            <div className="absolute right-16 top-1/2 -translate-y-1/2 transform sm:right-16" style={{ display: 'none' }}>
+              <MicButton onTranscript={onTranscript} className="h-10 w-10 sm:h-10 sm:w-10" />
+            </div>
+
            <button
              type="submit"
              disabled={!input.trim() || isLoading}
--- a/src/components/chat/view/subcomponents/ChatInputControls.tsx
+++ b/src/components/chat/view/subcomponents/ChatInputControls.tsx
@@ -2,7 +2,6 @@ import React from 'react';
 import { useTranslation } from 'react-i18next';
 import type { PermissionMode, Provider } from '../../types/types';
 import ThinkingModeSelector from './ThinkingModeSelector';
-import TokenUsagePie from './TokenUsagePie';

 interface ChatInputControlsProps {
  permissionMode: PermissionMode | string;
@@ -10,7 +9,6 @@ interface ChatInputControlsProps {
  provider: Provider | string;
  thinkingMode: string;
  setThinkingMode: React.Dispatch<React.SetStateAction<string>>;
-  tokenBudget: { used?: number; total?: number } | null;
  slashCommandsCount: number;
  onToggleCommandMenu: () => void;
  hasInput: boolean;
@@ -26,7 +24,6 @@ export default function ChatInputControls({
  provider,
  thinkingMode,
  setThinkingMode,
-  tokenBudget,
  slashCommandsCount,
  onToggleCommandMenu,
  hasInput,
@@ -78,8 +75,6 @@ export default function ChatInputControls({
        <ThinkingModeSelector selectedMode={thinkingMode} onModeChange={setThinkingMode} onClose={() => {}} className="" />
      )}

-      <TokenUsagePie used={tokenBudget?.used || 0} total={tokenBudget?.total || parseInt(import.meta.env.VITE_CONTEXT_WINDOW) || 160000} />
-
      <button
        type="button"
        onClick={onToggleCommandMenu}
--- a/src/components/chat/view/subcomponents/ClaudeStatus.tsx
+++ b/src/components/chat/view/subcomponents/ClaudeStatus.tsx
@@ -6,7 +6,6 @@ import SessionProviderLogo from '../../../llm-logo-provider/SessionProviderLogo'
 type ClaudeStatusProps = {
  status: {
    text?: string;
-    tokens?: number;
    can_interrupt?: boolean;
  } | null;
  onAbort?: () => void;
@@ -126,4 +125,4 @@ export default function ClaudeStatus({
      </div>
    </div>
  );
-}
+}
--- a/src/components/chat/view/subcomponents/TokenUsagePie.tsx
+++ b/src/components/chat/view/subcomponents/TokenUsagePie.tsx
@@ -1,54 +0,0 @@
-type TokenUsagePieProps = {
-  used: number;
-  total: number;
-};
-
-export default function TokenUsagePie({ used, total }: TokenUsagePieProps) {
-  // Token usage visualization component
-  // Only bail out on missing values or non‐positive totals; allow used===0 to render 0%
-  if (used == null || total == null || total <= 0) return null;
-
-  const percentage = Math.min(100, (used / total) * 100);
-  const radius = 10;
-  const circumference = 2 * Math.PI * radius;
-  const offset = circumference - (percentage / 100) * circumference;
-
-  // Color based on usage level
-  const getColor = () => {
-    if (percentage < 50) return '#3b82f6'; // blue
-    if (percentage < 75) return '#f59e0b'; // orange
-    return '#ef4444'; // red
-  };
-
-  return (
-    <div className="flex items-center gap-2 text-xs text-gray-600 dark:text-gray-400">
-      <svg width="24" height="24" viewBox="0 0 24 24" className="-rotate-90 transform">
-        {/* Background circle */}
-        <circle
-          cx="12"
-          cy="12"
-          r={radius}
-          fill="none"
-          stroke="currentColor"
-          strokeWidth="2"
-          className="text-gray-300 dark:text-gray-600"
-        />
-        {/* Progress circle */}
-        <circle
-          cx="12"
-          cy="12"
-          r={radius}
-          fill="none"
-          stroke={getColor()}
-          strokeWidth="2"
-          strokeDasharray={circumference}
-          strokeDashoffset={offset}
-          strokeLinecap="round"
-        />
-      </svg>
-      <span title={`${used.toLocaleString()} / ${total.toLocaleString()} tokens`}>
-        {percentage.toFixed(1)}%
-      </span>
-    </div>
-  );
-}
--- a/src/components/git-panel/view/changes/CommitComposer.tsx
+++ b/src/components/git-panel/view/changes/CommitComposer.tsx
@@ -1,5 +1,6 @@
 import { Check, ChevronDown, GitCommit, RefreshCw, Sparkles } from 'lucide-react';
 import { useState } from 'react';
+import MicButton from '../../../mic-button/view/MicButton';
 import type { ConfirmationRequest } from '../../types/types';

 // Persists commit messages across unmount/remount, keyed by project path
@@ -146,6 +147,13 @@ export default function CommitComposer({
                  <Sparkles className="h-4 w-4" />
                )}
              </button>
+              <div style={{ display: 'none' }}>
+                <MicButton
+                  onTranscript={(transcript) => setCommitMessage(transcript)}
+                  mode="default"
+                  className="p-1.5"
+                />
+              </div>
            </div>
          </div>

--- a/src/components/mic-button/constants/constants.ts
+++ b/src/components/mic-button/constants/constants.ts
@@ -0,0 +1,45 @@
+import type { MicButtonState } from '../types/types';
+
+export const MIC_BUTTON_STATES = {
+  IDLE: 'idle',
+  RECORDING: 'recording',
+  TRANSCRIBING: 'transcribing',
+  PROCESSING: 'processing',
+} as const;
+
+export const MIC_TAP_DEBOUNCE_MS = 300;
+export const PROCESSING_STATE_DELAY_MS = 2000;
+
+export const DEFAULT_WHISPER_MODE = 'default';
+
+// Modes that use post-transcription enhancement on the backend.
+export const ENHANCEMENT_WHISPER_MODES = new Set([
+  'prompt',
+  'vibe',
+  'instructions',
+  'architect',
+]);
+
+export const BUTTON_BACKGROUND_BY_STATE: Record<MicButtonState, string> = {
+  idle: '#374151',
+  recording: '#ef4444',
+  transcribing: '#3b82f6',
+  processing: '#a855f7',
+};
+
+export const MIC_ERROR_BY_NAME = {
+  NotAllowedError: 'Microphone access denied. Please allow microphone permissions.',
+  NotFoundError: 'No microphone found. Please check your audio devices.',
+  NotSupportedError: 'Microphone not supported by this browser.',
+  NotReadableError: 'Microphone is being used by another application.',
+} as const;
+
+export const MIC_NOT_AVAILABLE_ERROR =
+  'Microphone access not available. Please use HTTPS or a supported browser.';
+
+export const MIC_NOT_SUPPORTED_ERROR =
+  'Microphone not supported. Please use HTTPS or a modern browser.';
+
+export const MIC_SECURE_CONTEXT_ERROR =
+  'Microphone requires HTTPS. Please use a secure connection.';
+
--- a/src/components/mic-button/data/whisper.ts
+++ b/src/components/mic-button/data/whisper.ts
@@ -0,0 +1,52 @@
+import { api } from '../../../utils/api';
+
+type WhisperStatus = 'transcribing';
+
+type WhisperResponse = {
+  text?: string;
+  error?: string;
+};
+
+export async function transcribeWithWhisper(
+  audioBlob: Blob,
+  onStatusChange?: (status: WhisperStatus) => void,
+): Promise<string> {
+  const formData = new FormData();
+  const fileName = `recording_${Date.now()}.webm`;
+  const file = new File([audioBlob], fileName, { type: audioBlob.type });
+
+  formData.append('audio', file);
+
+  const whisperMode = window.localStorage.getItem('whisperMode') || 'default';
+  formData.append('mode', whisperMode);
+
+  try {
+    // Keep existing status callback behavior.
+    if (onStatusChange) {
+      onStatusChange('transcribing');
+    }
+
+    const response = (await api.transcribe(formData)) as Response;
+
+    if (!response.ok) {
+      const errorData = (await response.json().catch(() => ({}))) as WhisperResponse;
+      throw new Error(
+        errorData.error ||
+          `Transcription error: ${response.status} ${response.statusText}`,
+      );
+    }
+
+    const data = (await response.json()) as WhisperResponse;
+    return data.text || '';
+  } catch (error) {
+    if (
+      error instanceof Error
+      && error.name === 'TypeError'
+      && error.message.includes('fetch')
+    ) {
+      throw new Error('Cannot connect to server. Please ensure the backend is running.');
+    }
+    throw error;
+  }
+}
+
--- a/src/components/mic-button/hooks/useMicButtonController.ts
+++ b/src/components/mic-button/hooks/useMicButtonController.ts
@@ -0,0 +1,204 @@
+import { useEffect, useRef, useState } from 'react';
+import type { MouseEvent } from 'react';
+import { transcribeWithWhisper } from '../data/whisper';
+import {
+  DEFAULT_WHISPER_MODE,
+  ENHANCEMENT_WHISPER_MODES,
+  MIC_BUTTON_STATES,
+  MIC_ERROR_BY_NAME,
+  MIC_NOT_AVAILABLE_ERROR,
+  MIC_NOT_SUPPORTED_ERROR,
+  MIC_SECURE_CONTEXT_ERROR,
+  MIC_TAP_DEBOUNCE_MS,
+  PROCESSING_STATE_DELAY_MS,
+} from '../constants/constants';
+import type { MicButtonState } from '../types/types';
+
+type UseMicButtonControllerArgs = {
+  onTranscript?: (transcript: string) => void;
+};
+
+type UseMicButtonControllerResult = {
+  state: MicButtonState;
+  error: string | null;
+  isSupported: boolean;
+  handleButtonClick: (event?: MouseEvent<HTMLButtonElement>) => void;
+};
+
+const getRecordingErrorMessage = (error: unknown): string => {
+  if (error instanceof Error && error.message.includes('HTTPS')) {
+    return error.message;
+  }
+
+  if (error instanceof DOMException) {
+    return MIC_ERROR_BY_NAME[error.name as keyof typeof MIC_ERROR_BY_NAME] || 'Microphone access failed';
+  }
+
+  return 'Microphone access failed';
+};
+
+const getRecorderMimeType = (): string => (
+  MediaRecorder.isTypeSupported('audio/webm') ? 'audio/webm' : 'audio/mp4'
+);
+
+export function useMicButtonController({
+  onTranscript,
+}: UseMicButtonControllerArgs): UseMicButtonControllerResult {
+  const [state, setState] = useState<MicButtonState>(MIC_BUTTON_STATES.IDLE);
+  const [error, setError] = useState<string | null>(null);
+  const [isSupported, setIsSupported] = useState(true);
+
+  const mediaRecorderRef = useRef<MediaRecorder | null>(null);
+  const streamRef = useRef<MediaStream | null>(null);
+  const chunksRef = useRef<BlobPart[]>([]);
+  const lastTapRef = useRef(0);
+  const processingTimerRef = useRef<number | null>(null);
+
+  const clearProcessingTimer = (): void => {
+    if (processingTimerRef.current !== null) {
+      window.clearTimeout(processingTimerRef.current);
+      processingTimerRef.current = null;
+    }
+  };
+
+  const stopStreamTracks = (): void => {
+    if (!streamRef.current) {
+      return;
+    }
+
+    streamRef.current.getTracks().forEach((track) => track.stop());
+    streamRef.current = null;
+  };
+
+  const handleStopRecording = async (mimeType: string): Promise<void> => {
+    const audioBlob = new Blob(chunksRef.current, { type: mimeType });
+
+    // Release the microphone immediately once recording ends.
+    stopStreamTracks();
+    setState(MIC_BUTTON_STATES.TRANSCRIBING);
+
+    const whisperMode = window.localStorage.getItem('whisperMode') || DEFAULT_WHISPER_MODE;
+    const shouldShowProcessingState = ENHANCEMENT_WHISPER_MODES.has(whisperMode);
+
+    if (shouldShowProcessingState) {
+      processingTimerRef.current = window.setTimeout(() => {
+        setState(MIC_BUTTON_STATES.PROCESSING);
+      }, PROCESSING_STATE_DELAY_MS);
+    }
+
+    try {
+      const transcript = await transcribeWithWhisper(audioBlob);
+      if (transcript && onTranscript) {
+        onTranscript(transcript);
+      }
+    } catch (transcriptionError) {
+      const message = transcriptionError instanceof Error ? transcriptionError.message : 'Transcription error';
+      setError(message);
+    } finally {
+      clearProcessingTimer();
+      setState(MIC_BUTTON_STATES.IDLE);
+    }
+  };
+
+  const startRecording = async (): Promise<void> => {
+    try {
+      setError(null);
+      chunksRef.current = [];
+
+      if (!navigator.mediaDevices || !navigator.mediaDevices.getUserMedia) {
+        throw new Error(MIC_NOT_AVAILABLE_ERROR);
+      }
+
+      const stream = await navigator.mediaDevices.getUserMedia({ audio: true });
+      streamRef.current = stream;
+
+      const mimeType = getRecorderMimeType();
+      const recorder = new MediaRecorder(stream, { mimeType });
+      mediaRecorderRef.current = recorder;
+
+      recorder.ondataavailable = (event: BlobEvent) => {
+        if (event.data.size > 0) {
+          chunksRef.current.push(event.data);
+        }
+      };
+
+      recorder.onstop = () => {
+        void handleStopRecording(mimeType);
+      };
+
+      recorder.start();
+      setState(MIC_BUTTON_STATES.RECORDING);
+    } catch (recordingError) {
+      stopStreamTracks();
+      setError(getRecordingErrorMessage(recordingError));
+      setState(MIC_BUTTON_STATES.IDLE);
+    }
+  };
+
+  const stopRecording = (): void => {
+    if (mediaRecorderRef.current && mediaRecorderRef.current.state === 'recording') {
+      mediaRecorderRef.current.stop();
+      return;
+    }
+
+    stopStreamTracks();
+    setState(MIC_BUTTON_STATES.IDLE);
+  };
+
+  const handleButtonClick = (event?: MouseEvent<HTMLButtonElement>): void => {
+    if (event) {
+      event.preventDefault();
+      event.stopPropagation();
+    }
+
+    if (!isSupported) {
+      return;
+    }
+
+    // Mobile tap handling can trigger duplicate click events in quick succession.
+    const now = Date.now();
+    if (now - lastTapRef.current < MIC_TAP_DEBOUNCE_MS) {
+      return;
+    }
+    lastTapRef.current = now;
+
+    if (state === MIC_BUTTON_STATES.IDLE) {
+      void startRecording();
+      return;
+    }
+
+    if (state === MIC_BUTTON_STATES.RECORDING) {
+      stopRecording();
+    }
+  };
+
+  useEffect(() => {
+    // getUserMedia needs both browser support and a secure context.
+    if (!navigator.mediaDevices || !navigator.mediaDevices.getUserMedia) {
+      setIsSupported(false);
+      setError(MIC_NOT_SUPPORTED_ERROR);
+      return;
+    }
+
+    if (location.protocol !== 'https:' && location.hostname !== 'localhost') {
+      setIsSupported(false);
+      setError(MIC_SECURE_CONTEXT_ERROR);
+      return;
+    }
+
+    setIsSupported(true);
+    setError(null);
+  }, []);
+
+  useEffect(() => () => {
+    clearProcessingTimer();
+    stopStreamTracks();
+  }, []);
+
+  return {
+    state,
+    error,
+    isSupported,
+    handleButtonClick,
+  };
+}
--- a/src/components/mic-button/types/types.ts
+++ b/src/components/mic-button/types/types.ts
@@ -0,0 +1,2 @@
+export type MicButtonState = 'idle' | 'recording' | 'transcribing' | 'processing';
+
--- a/src/components/mic-button/view/MicButton.tsx
+++ b/src/components/mic-button/view/MicButton.tsx
@@ -0,0 +1,32 @@
+import { useMicButtonController } from '../hooks/useMicButtonController';
+import MicButtonView from './MicButtonView';
+
+type MicButtonProps = {
+  onTranscript?: (transcript: string) => void;
+  className?: string;
+  mode?: string;
+};
+
+export default function MicButton({
+  onTranscript,
+  className = '',
+  mode: _mode,
+}: MicButtonProps) {
+  const { state, error, isSupported, handleButtonClick } = useMicButtonController({
+    onTranscript,
+  });
+
+  // Keep `mode` in the public props for backwards compatibility.
+  void _mode;
+
+  return (
+    <MicButtonView
+      state={state}
+      error={error}
+      isSupported={isSupported}
+      className={className}
+      onButtonClick={handleButtonClick}
+    />
+  );
+}
+
--- a/src/components/mic-button/view/MicButtonView.tsx
+++ b/src/components/mic-button/view/MicButtonView.tsx
@@ -0,0 +1,86 @@
+import { Brain, Loader2, Mic } from 'lucide-react';
+import type { MouseEvent, ReactElement } from 'react';
+import { BUTTON_BACKGROUND_BY_STATE, MIC_BUTTON_STATES } from '../constants/constants';
+import type { MicButtonState } from '../types/types';
+
+type MicButtonViewProps = {
+  state: MicButtonState;
+  error: string | null;
+  isSupported: boolean;
+  className: string;
+  onButtonClick: (event?: MouseEvent<HTMLButtonElement>) => void;
+};
+
+const getButtonIcon = (state: MicButtonState, isSupported: boolean): ReactElement => {
+  if (!isSupported) {
+    return <Mic className="h-5 w-5" />;
+  }
+
+  if (state === MIC_BUTTON_STATES.TRANSCRIBING) {
+    return <Loader2 className="h-5 w-5 animate-spin" />;
+  }
+
+  if (state === MIC_BUTTON_STATES.PROCESSING) {
+    return <Brain className="h-5 w-5 animate-pulse" />;
+  }
+
+  if (state === MIC_BUTTON_STATES.RECORDING) {
+    return <Mic className="h-5 w-5 text-white" />;
+  }
+
+  return <Mic className="h-5 w-5" />;
+};
+
+export default function MicButtonView({
+  state,
+  error,
+  isSupported,
+  className,
+  onButtonClick,
+}: MicButtonViewProps) {
+  const isDisabled = !isSupported || state === MIC_BUTTON_STATES.TRANSCRIBING || state === MIC_BUTTON_STATES.PROCESSING;
+  const icon = getButtonIcon(state, isSupported);
+
+  return (
+    <div className="relative">
+      <button
+        type="button"
+        style={{ backgroundColor: BUTTON_BACKGROUND_BY_STATE[state] }}
+        className={`
+          touch-action-manipulation flex h-12
+          w-12 items-center justify-center
+          rounded-full text-white transition-all
+          duration-200 focus:outline-none focus:ring-2 focus:ring-blue-500
+          focus:ring-offset-2
+          dark:ring-offset-gray-800
+          ${isDisabled ? 'cursor-not-allowed opacity-75' : 'cursor-pointer'}
+          ${state === MIC_BUTTON_STATES.RECORDING ? 'animate-pulse' : ''}
+          hover:opacity-90
+          ${className}
+        `}
+        onClick={onButtonClick}
+        disabled={isDisabled}
+      >
+        {icon}
+      </button>
+
+      {error && (
+        <div
+          className="animate-fade-in absolute left-1/2 top-full z-10 mt-2
+                        -translate-x-1/2 transform whitespace-nowrap rounded bg-red-500 px-2 py-1 text-xs
+                        text-white"
+        >
+          {error}
+        </div>
+      )}
+
+      {state === MIC_BUTTON_STATES.RECORDING && (
+        <div className="pointer-events-none absolute -inset-1 animate-ping rounded-full border-2 border-red-500" />
+      )}
+
+      {state === MIC_BUTTON_STATES.PROCESSING && (
+        <div className="pointer-events-none absolute -inset-1 animate-ping rounded-full border-2 border-purple-500" />
+      )}
+    </div>
+  );
+}
--- a/src/components/quick-settings-panel/constants.ts
+++ b/src/components/quick-settings-panel/constants.ts
@@ -2,12 +2,21 @@ import {
  ArrowDown,
  Brain,
  Eye,
+  FileText,
  Languages,
  Maximize2,
+  Mic,
+  Sparkles,
 } from 'lucide-react';
-import type { PreferenceToggleItem } from './types';
+import type {
+  PreferenceToggleItem,
+  WhisperMode,
+  WhisperOption,
+} from './types';

 export const HANDLE_POSITION_STORAGE_KEY = 'quickSettingsHandlePosition';
+export const WHISPER_MODE_STORAGE_KEY = 'whisperMode';
+export const WHISPER_MODE_CHANGED_EVENT = 'whisperModeChanged';

 export const DEFAULT_HANDLE_POSITION = 50;
 export const HANDLE_POSITION_MIN = 10;
@@ -55,3 +64,30 @@ export const INPUT_SETTING_TOGGLES: PreferenceToggleItem[] = [
    icon: Languages,
  },
 ];
+
+export const WHISPER_OPTIONS: WhisperOption[] = [
+  {
+    value: 'default',
+    titleKey: 'quickSettings.whisper.modes.default',
+    descriptionKey: 'quickSettings.whisper.modes.defaultDescription',
+    icon: Mic,
+  },
+  {
+    value: 'prompt',
+    titleKey: 'quickSettings.whisper.modes.prompt',
+    descriptionKey: 'quickSettings.whisper.modes.promptDescription',
+    icon: Sparkles,
+  },
+  {
+    value: 'vibe',
+    titleKey: 'quickSettings.whisper.modes.vibe',
+    descriptionKey: 'quickSettings.whisper.modes.vibeDescription',
+    icon: FileText,
+  },
+];
+
+export const VIBE_MODE_ALIASES: WhisperMode[] = [
+  'vibe',
+  'instructions',
+  'architect',
+];
--- a/src/components/quick-settings-panel/hooks/useWhisperMode.ts
+++ b/src/components/quick-settings-panel/hooks/useWhisperMode.ts
@@ -0,0 +1,59 @@
+import { useCallback, useState } from 'react';
+import {
+  VIBE_MODE_ALIASES,
+  WHISPER_MODE_CHANGED_EVENT,
+  WHISPER_MODE_STORAGE_KEY,
+} from '../constants';
+import type { WhisperMode, WhisperOptionValue } from '../types';
+
+const ALL_VALID_MODES: WhisperMode[] = [
+  'default',
+  'prompt',
+  'vibe',
+  'instructions',
+  'architect',
+];
+
+const isWhisperMode = (value: string): value is WhisperMode => (
+  ALL_VALID_MODES.includes(value as WhisperMode)
+);
+
+const readStoredMode = (): WhisperMode => {
+  if (typeof window === 'undefined') {
+    return 'default';
+  }
+
+  const storedValue = localStorage.getItem(WHISPER_MODE_STORAGE_KEY);
+  if (!storedValue) {
+    return 'default';
+  }
+
+  return isWhisperMode(storedValue) ? storedValue : 'default';
+};
+
+export function useWhisperMode() {
+  const [whisperMode, setWhisperModeState] = useState<WhisperMode>(readStoredMode);
+
+  const setWhisperMode = useCallback((value: WhisperOptionValue) => {
+    setWhisperModeState(value);
+    localStorage.setItem(WHISPER_MODE_STORAGE_KEY, value);
+    window.dispatchEvent(new Event(WHISPER_MODE_CHANGED_EVENT));
+  }, []);
+
+  const isOptionSelected = useCallback(
+    (value: WhisperOptionValue) => {
+      if (value === 'vibe') {
+        return VIBE_MODE_ALIASES.includes(whisperMode);
+      }
+
+      return whisperMode === value;
+    },
+    [whisperMode],
+  );
+
+  return {
+    whisperMode,
+    setWhisperMode,
+    isOptionSelected,
+  };
+}
--- a/src/components/quick-settings-panel/types.ts
+++ b/src/components/quick-settings-panel/types.ts
@@ -16,4 +16,20 @@ export type PreferenceToggleItem = {
  icon: LucideIcon;
 };

+export type WhisperMode =
+  | 'default'
+  | 'prompt'
+  | 'vibe'
+  | 'instructions'
+  | 'architect';
+
+export type WhisperOptionValue = 'default' | 'prompt' | 'vibe';
+
+export type WhisperOption = {
+  value: WhisperOptionValue;
+  titleKey: string;
+  descriptionKey: string;
+  icon: LucideIcon;
+};
+
 export type QuickSettingsHandleStyle = CSSProperties;
--- a/src/components/quick-settings-panel/view/QuickSettingsContent.tsx
+++ b/src/components/quick-settings-panel/view/QuickSettingsContent.tsx
@@ -15,6 +15,7 @@ import type {
 } from '../types';
 import QuickSettingsSection from './QuickSettingsSection';
 import QuickSettingsToggleRow from './QuickSettingsToggleRow';
+import QuickSettingsWhisperSection from './QuickSettingsWhisperSection';

 type QuickSettingsContentProps = {
  isDarkMode: boolean;
@@ -72,6 +73,8 @@ export default function QuickSettingsContent({
          {t('quickSettings.sendByCtrlEnterDescription')}
        </p>
      </QuickSettingsSection>
+
+      <QuickSettingsWhisperSection />
    </div>
  );
 }
--- a/src/components/quick-settings-panel/view/QuickSettingsWhisperSection.tsx
+++ b/src/components/quick-settings-panel/view/QuickSettingsWhisperSection.tsx
@@ -0,0 +1,44 @@
+import { useTranslation } from 'react-i18next';
+import { TOGGLE_ROW_CLASS, WHISPER_OPTIONS } from '../constants';
+import { useWhisperMode } from '../hooks/useWhisperMode';
+import QuickSettingsSection from './QuickSettingsSection';
+
+export default function QuickSettingsWhisperSection() {
+  const { t } = useTranslation('settings');
+  const { setWhisperMode, isOptionSelected } = useWhisperMode();
+
+  return (
+    // This section stays hidden intentionally until dictation modes are reintroduced.
+    <QuickSettingsSection
+      title={t('quickSettings.sections.whisperDictation')}
+      className="hidden"
+    >
+      <div className="space-y-2">
+        {WHISPER_OPTIONS.map(({ value, icon: Icon, titleKey, descriptionKey }) => (
+          <label
+            key={value}
+            className={`${TOGGLE_ROW_CLASS} flex items-start`}
+          >
+            <input
+              type="radio"
+              name="whisperMode"
+              value={value}
+              checked={isOptionSelected(value)}
+              onChange={() => setWhisperMode(value)}
+              className="mt-0.5 h-4 w-4 border-gray-300 text-blue-600 focus:ring-blue-500 dark:border-gray-600 dark:bg-gray-800 dark:text-blue-500 dark:checked:bg-blue-600 dark:focus:ring-blue-400"
+            />
+            <div className="ml-3 flex-1">
+              <span className="flex items-center gap-2 text-sm font-medium text-gray-900 dark:text-white">
+                <Icon className="h-4 w-4 text-gray-600 dark:text-gray-400" />
+                {t(titleKey)}
+              </span>
+              <p className="mt-1 text-xs text-gray-500 dark:text-gray-400">
+                {t(descriptionKey)}
+              </p>
+            </div>
+          </label>
+        ))}
+      </div>
+    </QuickSettingsSection>
+  );
+}
--- a/src/i18n/locales/de/settings.json
+++ b/src/i18n/locales/de/settings.json
@@ -55,7 +55,8 @@
      "appearance": "Darstellung",
      "toolDisplay": "Werkzeuganzeige",
      "viewOptions": "Anzeigeoptionen",
-      "inputSettings": "Eingabeeinstellungen"
+      "inputSettings": "Eingabeeinstellungen",
+      "whisperDictation": "Whisper-Diktat"
    },
    "darkMode": "Darkmode",
    "autoExpandTools": "Werkzeuge automatisch erweitern",
@@ -70,6 +71,16 @@
      "openPanel": "Einstellungspanel öffnen",
      "draggingStatus": "Wird gezogen...",
      "toggleAndMove": "Klicken zum Umschalten, ziehen zum Verschieben"
+    },
+    "whisper": {
+      "modes": {
+        "default": "Standardmodus",
+        "defaultDescription": "Direkte Transkription deiner Sprache",
+        "prompt": "Prompt-Verbesserung",
+        "promptDescription": "Rohe Ideen in klare, detaillierte KI-Prompts umwandeln",
+        "vibe": "Vibe-Modus",
+        "vibeDescription": "Ideen als klare Agentenanweisungen mit Details formatieren"
+      }
    }
  },
  "terminalShortcuts": {
--- a/src/i18n/locales/en/settings.json
+++ b/src/i18n/locales/en/settings.json
@@ -55,7 +55,8 @@
      "appearance": "Appearance",
      "toolDisplay": "Tool Display",
      "viewOptions": "View Options",
-      "inputSettings": "Input Settings"
+      "inputSettings": "Input Settings",
+      "whisperDictation": "Whisper Dictation"
    },
    "darkMode": "Dark Mode",
    "autoExpandTools": "Auto-expand tools",
@@ -70,6 +71,16 @@
      "openPanel": "Open settings panel",
      "draggingStatus": "Dragging...",
      "toggleAndMove": "Click to toggle, drag to move"
+    },
+    "whisper": {
+      "modes": {
+        "default": "Default Mode",
+        "defaultDescription": "Direct transcription of your speech",
+        "prompt": "Prompt Enhancement",
+        "promptDescription": "Transform rough ideas into clear, detailed AI prompts",
+        "vibe": "Vibe Mode",
+        "vibeDescription": "Format ideas as clear agent instructions with details"
+      }
    }
  },
  "terminalShortcuts": {
@@ -487,4 +498,4 @@
    "tab": "tab",
    "runningStatus": "running"
  }
-}
+}
--- a/src/i18n/locales/ja/settings.json
+++ b/src/i18n/locales/ja/settings.json
@@ -55,7 +55,8 @@
      "appearance": "外観",
      "toolDisplay": "ツール表示",
      "viewOptions": "表示オプション",
-      "inputSettings": "入力設定"
+      "inputSettings": "入力設定",
+      "whisperDictation": "Whisper音声入力"
    },
    "darkMode": "ダークモード",
    "autoExpandTools": "ツールを自動展開",
@@ -70,6 +71,16 @@
      "openPanel": "設定パネルを開く",
      "draggingStatus": "ドラッグ中...",
      "toggleAndMove": "クリックで切替、ドラッグで移動"
+    },
+    "whisper": {
+      "modes": {
+        "default": "標準モード",
+        "defaultDescription": "音声をそのまま文字起こしします",
+        "prompt": "プロンプト強化",
+        "promptDescription": "ラフなアイデアを明確で詳細なAIプロンプトに変換します",
+        "vibe": "バイブモード",
+        "vibeDescription": "アイデアを明確なエージェント指示に整形します"
+      }
    }
  },
  "terminalShortcuts": {
@@ -481,4 +492,4 @@
    "tab": "タブ",
    "runningStatus": "実行中"
  }
-}
+}
--- a/src/i18n/locales/ko/settings.json
+++ b/src/i18n/locales/ko/settings.json
@@ -55,7 +55,8 @@
      "appearance": "외관",
      "toolDisplay": "도구 표시",
      "viewOptions": "보기 옵션",
-      "inputSettings": "입력 설정"
+      "inputSettings": "입력 설정",
+      "whisperDictation": "Whisper 음성 인식"
    },
    "darkMode": "다크 모드",
    "autoExpandTools": "도구 자동 펼치기",
@@ -70,6 +71,16 @@
      "openPanel": "설정 패널 열기",
      "draggingStatus": "드래그 중...",
      "toggleAndMove": "클릭하여 토글, 드래그하여 이동"
+    },
+    "whisper": {
+      "modes": {
+        "default": "기본 모드",
+        "defaultDescription": "음성을 그대로 텍스트로 변환",
+        "prompt": "프롬프트 향상",
+        "promptDescription": "거친 아이디어를 명확하고 상세한 AI 프롬프트로 변환",
+        "vibe": "Vibe 모드",
+        "vibeDescription": "아이디어를 상세한 에이전트 지침 형식으로 변환"
+      }
    }
  },
  "terminalShortcuts": {
@@ -481,4 +492,4 @@
    "tab": "탭",
    "runningStatus": "실행 중"
  }
-}
+}
--- a/src/i18n/locales/ru/settings.json
+++ b/src/i18n/locales/ru/settings.json
@@ -55,7 +55,8 @@
      "appearance": "Внешний вид",
      "toolDisplay": "Отображение инструментов",
      "viewOptions": "Параметры просмотра",
-      "inputSettings": "Настройки ввода"
+      "inputSettings": "Настройки ввода",
+      "whisperDictation": "Диктовка Whisper"
    },
    "darkMode": "Темная тема",
    "autoExpandTools": "Автоматически разворачивать инструменты",
@@ -70,6 +71,16 @@
      "openPanel": "Открыть панель настроек",
      "draggingStatus": "Перетаскивание...",
      "toggleAndMove": "Нажмите для переключения, перетащите для перемещения"
+    },
+    "whisper": {
+      "modes": {
+        "default": "Режим по умолчанию",
+        "defaultDescription": "Прямая транскрипция вашей речи",
+        "prompt": "Улучшение запроса",
+        "promptDescription": "Преобразование грубых идей в четкие, детальные AI-запросы",
+        "vibe": "Режим Vibe",
+        "vibeDescription": "Форматирование идей как четких инструкций агента с деталями"
+      }
    }
  },
  "terminalShortcuts": {
@@ -461,4 +472,4 @@
    "tab": "вкладка",
    "runningStatus": "запущен"
  }
-}
+}
--- a/src/i18n/locales/zh-CN/settings.json
+++ b/src/i18n/locales/zh-CN/settings.json
@@ -55,7 +55,8 @@
      "appearance": "外观",
      "toolDisplay": "工具显示",
      "viewOptions": "视图选项",
-      "inputSettings": "输入设置"
+      "inputSettings": "输入设置",
+      "whisperDictation": "Whisper 听写"
    },
    "darkMode": "深色模式",
    "autoExpandTools": "自动展开工具",
@@ -70,6 +71,16 @@
      "openPanel": "打开设置面板",
      "draggingStatus": "正在拖拽...",
      "toggleAndMove": "点击切换，拖拽移动"
+    },
+    "whisper": {
+      "modes": {
+        "default": "默认模式",
+        "defaultDescription": "直接转录您的语音",
+        "prompt": "提示词增强",
+        "promptDescription": "将粗略的想法转化为清晰、详细的 AI 提示词",
+        "vibe": "Vibe 模式",
+        "vibeDescription": "将想法格式化为带有详细说明的清晰智能体指令"
+      }
    }
  },
  "terminalShortcuts": {
@@ -481,4 +492,4 @@
    "tab": "标签",
    "runningStatus": "运行中"
  }
-}
+}
--- a/src/stores/useSessionStore.ts
+++ b/src/stores/useSessionStore.ts
@@ -46,9 +46,7 @@ export interface NormalizedMessage {
  toolResult?: { content: string; isError: boolean; toolUseResult?: unknown } | null;
  isError?: boolean;
  text?: string;
-  tokens?: number;
  canInterrupt?: boolean;
-  tokenBudget?: unknown;
  requestId?: string;
  input?: unknown;
  context?: unknown;
@@ -81,7 +79,6 @@ export interface SessionSlot {
  total: number;
  hasMore: boolean;
  offset: number;
-  tokenUsage: unknown;
 }

 const EMPTY: NormalizedMessage[] = [];
@@ -98,7 +95,6 @@ function createEmptySlot(): SessionSlot {
    total: 0,
    hasMore: false,
    offset: 0,
-    tokenUsage: null,
  };
 }

@@ -208,9 +204,6 @@ export function useSessionStore() {
      slot.fetchedAt = Date.now();
      slot.status = 'idle';
      recomputeMergedIfNeeded(slot);
-      if (data.tokenUsage) {
-        slot.tokenUsage = data.tokenUsage;
-      }

      notify(sessionId);
      return slot;
--- a/src/utils/api.js
+++ b/src/utils/api.js
@@ -147,6 +147,13 @@ export const api = {
      headers: {}, // Let browser set Content-Type for FormData
    }),

+  transcribe: (formData) =>
+    authenticatedFetch('/api/transcribe', {
+      method: 'POST',
+      body: formData,
+      headers: {}, // Let browser set Content-Type for FormData
+    }),
+
  // TaskMaster endpoints
  taskmaster: {
    // Initialize TaskMaster in a project
Author	SHA1	Message	Date
Simos Mikelatos	fce8ad0893	Merge branch 'main' into refactor/remove-token-usage-logic	2026-04-10 15:30:43 +02:00
Haileyesus	cc7f652044	refactor: removed the token calculator logic The information given was not accurate for the model providers	2026-04-10 16:21:01 +03:00
				`@@ -0,0 +1,2 @@`
				`export type MicButtonState = 'idle' \| 'recording' \| 'transcribing' \| 'processing';`