audio and transcript

cht/agent/runner.py

@@ -15,7 +15,7 @@ from pathlib import Path
 from threading import Thread
 from typing import Callable
 
-from cht.agent.base import AgentProvider, FrameRef, SessionContext
+from cht.agent.base import AgentProvider, FrameRef, TranscriptRef, SessionContext
 
 log = logging.getLogger(__name__)
 
@@ -98,6 +98,33 @@ def _load_frames(frames_dir: Path) -> list[FrameRef]:
         return []
 
 
+def _load_transcript(transcript_dir: Path) -> list[TranscriptRef]:
+    index_path = transcript_dir / "index.json"
+    if not index_path.exists():
+        return []
+    try:
+        entries = json.loads(index_path.read_text())
+        return [TranscriptRef(**e) for e in entries]
+    except Exception as e:
+        log.warning("Could not load transcript index: %s", e)
+        return []
+
+
+def _parse_transcript_mentions(message: str, segments: list[TranscriptRef]) -> list[TranscriptRef]:
+    """Extract @T references from message. Accepts @T0001, @t1, @T1."""
+    mentioned = []
+    seen = set()
+    for match in re.finditer(r"@[Tt](\d+)", message):
+        num = int(match.group(1))
+        tid = f"T{num:04d}"
+        if tid not in seen:
+            seg = next((s for s in segments if s.id == tid), None)
+            if seg:
+                mentioned.append(seg)
+                seen.add(tid)
+    return mentioned
+
+
 class AgentRunner:
     """Runs agent queries in a background thread, streams chunks to a callback."""
 
@@ -152,12 +179,16 @@ class AgentRunner:
             try:
                 provider = self._get_provider()
                 frames = _load_frames(stream_mgr.frames_dir)
-                mentioned = _parse_mentions(message, frames)
+                mentioned_frames = _parse_mentions(message, frames)
+                transcript = _load_transcript(stream_mgr.transcript_dir)
+                mentioned_transcripts = _parse_transcript_mentions(message, transcript)
                 context = SessionContext(
                     session_dir=stream_mgr.session_dir,
                     frames=frames,
                     duration=tracker.duration if tracker else 0.0,
-                    mentioned_frames=mentioned,
+                    mentioned_frames=mentioned_frames,
+                    transcript_segments=transcript,
+                    mentioned_transcripts=mentioned_transcripts,
                 )
                 for chunk in provider.stream(message, context):
                     on_chunk(chunk)