Add speaker-attributed meeting transcription

bot.py

@@ -1,6 +1,7 @@
 from __future__ import annotations
 
 import asyncio
+import json
 import logging
 import os
 import uuid
@@ -14,7 +15,7 @@ from dotenv import load_dotenv
 
 import config
 from helpers import chunk_message, command_channel_error
-from openrouter_client import summarize, transcribe
+from openrouter_client import summarize, transcribe, transcribe_tracks
 from voice import MeetingRecorder
 
 logging.basicConfig(
@@ -187,6 +188,74 @@ async def _save_retry_state(guild_id: int, file_path: str, source_channel_id: in
     )
 
 
+def _load_speaker_tracks(file_path: str) -> list[dict]:
+    """Return per-speaker track entries for a recording, or [] if none.
+
+    Looks for a ``tracks/tracks.json`` manifest next to the mixed archive WAV.
+    Track paths in the manifest are stored relative to the app working
+    directory; they are resolved against the manifest location when needed so
+    the lookup works regardless of absolute vs relative archive paths.
+    """
+    manifest_path = Path(file_path).parent / "tracks" / "tracks.json"
+    if not manifest_path.exists():
+        return []
+
+    try:
+        manifest = json.loads(manifest_path.read_text())
+    except (json.JSONDecodeError, OSError) as exc:
+        logger.warning("Failed reading speaker-track manifest path=%s error=%s", manifest_path, exc)
+        return []
+
+    tracks = manifest.get("tracks")
+    if not isinstance(tracks, list):
+        return []
+
+    resolved: list[dict] = []
+    for track in tracks:
+        if not isinstance(track, dict):
+            continue
+        path = track.get("path")
+        display_name = track.get("display_name") or "Unknown User"
+        if not path:
+            continue
+        candidate = Path(path)
+        if not candidate.exists():
+            # Manifest paths may be relative to the app cwd; also try resolving
+            # the file name against the manifest's own directory.
+            alt = manifest_path.parent / candidate.name
+            if alt.exists():
+                candidate = alt
+            else:
+                logger.warning("Speaker track file missing: display_name=%s path=%s", display_name, path)
+                continue
+        resolved.append({"display_name": display_name, "path": str(candidate)})
+
+    return resolved
+
+
+async def _transcribe_recording(file_path: str, guild_id: int) -> str:
+    """Transcribe using per-speaker tracks when available, else the mixed file."""
+    tracks = _load_speaker_tracks(file_path)
+    if tracks:
+        logger.info(
+            "Using speaker-attributed transcription: guild_id=%s speakers=%s",
+            guild_id,
+            len(tracks),
+        )
+        try:
+            return await transcribe_tracks(tracks)
+        except Exception as exc:  # noqa: BLE001 - fall back to mixed file
+            logger.warning(
+                "Speaker-track transcription failed; falling back to mixed archive: guild_id=%s error=%s",
+                guild_id,
+                exc,
+            )
+
+    logger.info("Using single-file transcription: guild_id=%s path=%s", guild_id, file_path)
+    return await transcribe(file_path)
+
+
+
 async def process_recording(
     file_path: str,
     guild_id: int,
@@ -237,7 +306,7 @@ async def process_recording(
                 os.path.getsize(file_path) / 1024 / 1024,
             )
 
-        transcript = await transcribe(file_path)
+        transcript = await _transcribe_recording(file_path, guild_id)
         logger.info("Transcription complete for guild_id=%s transcript_chars=%s", guild_id, len(transcript))
         summary = await summarize(transcript)
         logger.info("Summarization complete for guild_id=%s summary_chars=%s", guild_id, len(summary))