feat: singleton audio, transcript reflection, voice persistence

Matthias Nott

2026-03-07 0e888d62af1434fef231e11a5c307a5b48a8deb1

 contexts/ChatContext.tsx
..
..
@@ -26,7 +26,9 @@
26
26
 
27
27
 const MESSAGES_DIR = "pailot-messages";
28
28
 
29
-/** Strip heavy fields (base64 images, audio URIs) before persisting. */
29
+/** Strip heavy fields (base64 images, audio URIs) before persisting.
30
+ *  Voice messages keep their content (transcript) but lose audioUri
31
+ *  since cache files won't survive app restarts. */
30
32
 function lightMessage(m: Message): Message {
31
33
   const light = { ...m };
32
34
   if (light.imageBase64) light.imageBase64 = undefined;
..
..
@@ -63,7 +65,16 @@
63
65
       if (!file.endsWith(".json")) continue;
64
66
       const sessionId = file.replace(".json", "");
65
67
       const content = await fs.readAsStringAsync(`${dir}${file}`);
66
-      result[sessionId] = JSON.parse(content) as Message[];
68
+      result[sessionId] = (JSON.parse(content) as Message[])
69
+        // Drop voice messages with no audio and no content (empty chunks)
70
+        .filter((m) => !(m.type === "voice" && !m.audioUri && !m.content))
71
+        .map((m) => {
72
+          // Voice messages without audio but with transcript → show as text
73
+          if (m.type === "voice" && !m.audioUri && m.content) {
74
+            return { ...m, type: "text" };
75
+          }
76
+          return m;
77
+        });
67
78
     }
68
79
     return result;
69
80
   } catch {
..
..
@@ -179,12 +190,15 @@
179
190
     }
180
191
   }, [messages]);
181
192
 
182
-  // On connect: ask gateway to detect the focused iTerm2 session and sync
193
+  // On connect: ask gateway to sync sessions. If we already had a session
194
+  // selected, tell the gateway so it preserves our selection instead of
195
+  // jumping to whatever iTerm has focused on the Mac.
183
196
   useEffect(() => {
184
197
     if (status === "connected") {
185
198
       needsSync.current = true;
186
-      sendCommand("sync");
199
+      sendCommand("sync", activeSessionId ? { activeSessionId } : undefined);
187
200
     }
201
+    // eslint-disable-next-line react-hooks/exhaustive-deps — only fire on status change
188
202
   }, [status, sendCommand]);
189
203
 
190
204
   // Helper: add a message to the active session
..
..
@@ -233,6 +247,23 @@
233
247
     },
234
248
     []
235
249
   );
250
+
251
+  // Update a message's content (e.g., voice transcript reflection)
252
+  const updateMessageContent = useCallback((id: string, content: string) => {
253
+    setMessages((prev) => {
254
+      const next = prev.map((m) =>
255
+        m.id === id ? { ...m, content } : m
256
+      );
257
+      setActiveSessionId((sessId) => {
258
+        if (sessId) {
259
+          messagesMapRef.current[sessId] = next;
260
+          debouncedSave(messagesMapRef.current);
261
+        }
262
+        return sessId;
263
+      });
264
+      return next;
265
+    });
266
+  }, []);
236
267
 
237
268
   // Handle incoming WebSocket messages
238
269
   useEffect(() => {
..
..
@@ -322,6 +353,11 @@
322
353
           sendCommand("sessions");
323
354
           break;
324
355
         }
356
+        case "transcript": {
357
+          // Voice → text reflection: replace voice bubble with transcribed text
358
+          updateMessageContent(data.messageId, data.content);
359
+          break;
360
+        }
325
361
         case "error": {
326
362
           const msg: Message = {
327
363
             id: generateId(),
..
..
@@ -339,7 +375,7 @@
339
375
     return () => {
340
376
       onMessageReceived.current = null;
341
377
     };
342
-  }, [onMessageReceived, sendCommand, addMessageToActive, syncActiveFromSessions]);
378
+  }, [onMessageReceived, sendCommand, addMessageToActive, updateMessageContent, syncActiveFromSessions]);
343
379
 
344
380
   const sendTextMessage = useCallback(
345
381
     (text: string) => {
..
..
@@ -375,7 +411,7 @@
375
411
       addMessageToActive(msg);
376
412
       try {
377
413
         const base64 = await encodeAudioToBase64(audioUri);
378
-        const sent = wsVoice(base64);
414
+        const sent = wsVoice(base64, "", id);
379
415
         updateMessageStatus(id, sent ? "sent" : "error");
380
416
       } catch (err) {
381
417
         console.error("Failed to encode audio:", err);