tracegist-mcp-bridge 0.2.4 → 0.2.6

package/bin/tracegist-mcp-bridge.mjs

@@ -848,9 +848,14 @@ server.registerTool(
 const LIVE_SHADOW_PORT = 19384;
 const LIVE_SHADOW_PORT_RANGE = 5;
 const LIVE_SHADOW_EVENT_BUFFER_SIZE = 2000;
+const PENDING_QUESTIONS_MAX = 50;
+const OPENROUTER_API_URL = "https://openrouter.ai/api/v1/chat/completions";
+const OPENROUTER_API_KEY = process.env.OPENROUTER_API_KEY || "";
+const LIVE_TRANSCRIPTION_MODEL = "openai/gpt-audio-mini";
 
 /** @type {Array<{seq: number, ts: number, eventType: string, data: object}>} */
 const liveEvents = [];
+let liveEventSeq = 0;
 let liveSessionActive = false;
 /** @type {{sessionId: string, url: string, title: string, startedAt: number} | null} */
 let liveSessionMeta = null;
@@ -864,8 +869,97 @@ const questionResponses = new Map();
 /** @type {Array<{resolve: Function, reject: Function, timeout: ReturnType<typeof setTimeout>}>} */
 const screenshotWaiters = [];
 
+/**
+ * Transcribe a voice data URL using OpenRouter API or local Whisper fallback.
+ * @param {string} voiceBlobDataUrl - data:audio/...;base64,... URL
+ * @returns {Promise<{transcription?: string, error?: string}>}
+ */
+async function transcribeLiveVoice(voiceBlobDataUrl) {
+  // Try OpenRouter API first
+  if (OPENROUTER_API_KEY) {
+    try {
+      const match = voiceBlobDataUrl.match(/^data:([^;]+);base64,(.+)$/);
+      if (!match) return { error: "Invalid voice data URL format" };
+
+      const mimeType = match[1];
+      const base64Data = match[2];
+      const format = mimeType.includes("wav") ? "wav" : mimeType.includes("mp3") ? "mp3" : "webm";
+
+      const response = await fetch(OPENROUTER_API_URL, {
+        method: "POST",
+        headers: {
+          Authorization: `Bearer ${OPENROUTER_API_KEY}`,
+          "Content-Type": "application/json",
+          "HTTP-Referer": "npm:tracegist-mcp-bridge",
+          "X-Title": "TraceGist MCP Bridge",
+        },
+        body: JSON.stringify({
+          model: LIVE_TRANSCRIPTION_MODEL,
+          messages: [
+            {
+              role: "user",
+              content: [
+                {
+                  type: "text",
+                  text: "Transcribe this voice note verbatim. Return only the transcription text, no commentary.",
+                },
+                { type: "input_audio", input_audio: { data: base64Data, format } },
+              ],
+            },
+          ],
+          temperature: 0,
+        }),
+      });
+
+      if (!response.ok) {
+        const errorText = await response.text();
+        return { error: `OpenRouter API error (${response.status}): ${errorText}` };
+      }
+
+      const result = await response.json();
+      const text = result.choices?.[0]?.message?.content?.trim();
+      return text ? { transcription: text } : { error: "Empty transcription response" };
+    } catch (err) {
+      return { error: `OpenRouter transcription failed: ${err instanceof Error ? err.message : String(err)}` };
+    }
+  }
+
+  // Fallback to local Whisper
+  if (whisperDependencyWarnings.length === 0) {
+    try {
+      const match = voiceBlobDataUrl.match(/^data:([^;]+);base64,(.+)$/);
+      if (!match) return { error: "Invalid voice data URL format" };
+
+      const buffer = Buffer.from(match[2], "base64");
+      const ext = match[1].includes("wav") ? "wav" : match[1].includes("mp3") ? "mp3" : "webm";
+      const tmpDir = await fs.mkdtemp(path.join(os.tmpdir(), "tracegist-live-whisper-"));
+      const tmpPath = path.join(tmpDir, `voice.${ext}`);
+      await fs.writeFile(tmpPath, buffer);
+      try {
+        const text = await transcribeWithLocalWhisper(tmpPath, "base", undefined);
+        return { transcription: text };
+      } finally {
+        await fs.rm(tmpDir, { recursive: true, force: true });
+      }
+    } catch (err) {
+      return { error: `Whisper transcription failed: ${err instanceof Error ? err.message : String(err)}` };
+    }
+  }
+
+  return { error: "No transcription available (set OPENROUTER_API_KEY or install Whisper)" };
+}
+
+/** Push a live event into the ring buffer with auto-incrementing seq. */
+function pushLiveEvent(eventType, data) {
+  liveEventSeq++;
+  liveEvents.push({ seq: liveEventSeq, ts: Date.now(), eventType, data });
+  if (liveEvents.length > LIVE_SHADOW_EVENT_BUFFER_SIZE) {
+    const evictCount = Math.floor(LIVE_SHADOW_EVENT_BUFFER_SIZE * 0.25);
+    liveEvents.splice(0, evictCount);
+  }
+}
+
 function startLiveShadowServer() {
-  let port = LIVE_SHADOW_PORT;
   let attempts = 0;
 
   function tryPort(p) {
@@ -899,6 +993,11 @@ function startLiveShadowServer() {
           if (liveSessionActive) {
             liveSessionActive = false;
             console.error(`[${BRIDGE_NAME}] Live shadow: session ended (extension disconnected)`);
+            try {
+              server.sendResourceListChanged();
+            } catch {
+              // Not all transports support notifications
+            }
           }
         }
       });
@@ -918,11 +1017,13 @@ function startLiveShadowServer() {
     });
 
     httpServer.listen(p, "127.0.0.1", () => {
-      console.error(`[${BRIDGE_NAME}] Live shadow WebSocket server listening on ws://127.0.0.1:${p}`);
+      console.error(
+        `[${BRIDGE_NAME}] Live shadow WebSocket server listening on ws://127.0.0.1:${p}`,
+      );
     });
   }
 
-  tryPort(port);
+  tryPort(LIVE_SHADOW_PORT);
 }
 
 function handleExtensionMessage(msg) {
@@ -955,11 +1056,13 @@ function handleExtensionMessage(msg) {
         data: msg.data,
       });
       if (liveEvents.length > LIVE_SHADOW_EVENT_BUFFER_SIZE) {
-        liveEvents.shift();
+        // Batch-evict oldest 25% to amortize the O(n) splice cost
+        const evictCount = Math.floor(LIVE_SHADOW_EVENT_BUFFER_SIZE * 0.25);
+        liveEvents.splice(0, evictCount);
       }
       break;
 
-    case "question-response":
+    case "question-response": {
       questionResponses.set(msg.questionId, {
         questionId: msg.questionId,
         timestamp: msg.timestamp,
@@ -967,11 +1070,17 @@ function handleExtensionMessage(msg) {
         screenshot: msg.screenshot || null,
         highlightCaptures: msg.highlightCaptures || [],
       });
+      const rIdx = pendingQuestions.findIndex((q) => q.questionId === msg.questionId);
+      if (rIdx >= 0) pendingQuestions.splice(rIdx, 1);
       break;
+    }
 
-    case "question-dismissed":
+    case "question-dismissed": {
       questionResponses.set(msg.questionId, { dismissed: true });
+      const dIdx = pendingQuestions.findIndex((q) => q.questionId === msg.questionId);
+      if (dIdx >= 0) pendingQuestions.splice(dIdx, 1);
       break;
+    }
 
     case "screenshot-response":
       // Resolve any pending screenshot waiters
@@ -981,9 +1090,35 @@ function handleExtensionMessage(msg) {
       }
       break;
 
+    case "marker-voice": {
+      const { markerId, voiceBlobDataUrl } = msg;
+      transcribeLiveVoice(voiceBlobDataUrl)
+        .then((result) => {
+          pushLiveEvent("voice-transcription", {
+            markerId,
+            transcription: result.transcription || null,
+            error: result.error || undefined,
+          });
+        })
+        .catch((err) => {
+          console.error(`[${BRIDGE_NAME}] Live voice transcription failed:`, err);
+          pushLiveEvent("voice-transcription", {
+            markerId,
+            transcription: null,
+            error: String(err),
+          });
+        });
+      break;
+    }
+
     case "session-end":
       liveSessionActive = false;
       console.error(`[${BRIDGE_NAME}] Live shadow: session ended`);
+      try {
+        server.sendResourceListChanged();
+      } catch {
+        // Not all transports support notifications
+      }
       break;
   }
 }
@@ -1048,6 +1183,22 @@ server.tool(
               events,
               latestSeq,
               eventCount: events.length,
+              ...(liveSessionActive
+                ? {
+                    availableActions: [
+                      {
+                        tool: "ask_tester_question",
+                        description:
+                          "Ask the tester a short question (max 200 chars). They respond with voice + highlights.",
+                      },
+                      {
+                        tool: "get_live_screenshot",
+                        description:
+                          "Capture a screenshot of the tester's current browser tab.",
+                      },
+                    ],
+                  }
+                : {}),
             },
             null,
             2,
@@ -1093,6 +1244,9 @@ server.tool(
     }
 
     pendingQuestions.push({ questionId, question, sentAt: Date.now() });
+    while (pendingQuestions.length > PENDING_QUESTIONS_MAX) {
+      pendingQuestions.shift();
+    }
 
     return {
       content: [
@@ -1191,10 +1345,18 @@ server.tool(
     }
 
     if (response.voiceBlobDataUrl) {
-      content.push({
-        type: "text",
-        text: "Voice note is attached. Use the transcribe_tracegist_package_voice_notes tool or listen locally.",
-      });
+      const voiceResult = await transcribeLiveVoice(response.voiceBlobDataUrl);
+      if (voiceResult.transcription) {
+        content.push({
+          type: "text",
+          text: `Voice transcription:\n${voiceResult.transcription}`,
+        });
+      } else {
+        content.push({
+          type: "text",
+          text: `Voice note is attached but could not be transcribed${voiceResult.error ? ` (${voiceResult.error})` : ""}. Set OPENROUTER_API_KEY or install local Whisper for transcription.`,
+        });
+      }
     }
 
     // Clean up after retrieval

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "tracegist-mcp-bridge",
-  "version": "0.2.4",
+  "version": "0.2.6",
   "description": "Local-first MCP bridge for reading and transcribing TraceGist package zips.",
   "type": "module",
   "bin": {