add root readme

cht/summary/pipeline.py

@@ -0,0 +1,72 @@
+"""End-to-end orchestrator for the summarization export.
+
+Two operations:
+  diarize(...)   — heavy: assembles audio, runs whisperx, caches diarized.json.
+  export(...)    — cheap: merges cached diarization with selected frames and
+                   writes <session>_enhanced.txt. Re-run any time the user
+                   tweaks frame selection or speaker names.
+"""
+
+import json
+import logging
+from pathlib import Path
+
+from cht.session import load_frame_index
+from cht.summary import audio, diarize, merger, output
+
+log = logging.getLogger(__name__)
+
+
+def diarized_path(session_dir: Path) -> Path:
+    return session_dir / "summary" / "diarized.json"
+
+
+def has_diarization(session_dir: Path) -> bool:
+    return diarized_path(session_dir).exists()
+
+
+def load_diarization(session_dir: Path) -> dict:
+    return json.loads(diarized_path(session_dir).read_text())
+
+
+def run_diarization(session_dir: Path, *, num_speakers: int, on_progress=None) -> dict:
+    """Assemble audio, run whisperx, cache and return the JSON."""
+    if on_progress:
+        on_progress("assembling audio", 0.05)
+    wav = audio.assemble_session_wav(session_dir)
+
+    if on_progress:
+        on_progress("running whisperx", 0.15)
+    summary_dir = session_dir / "summary"
+    diarized = diarize.run_whisperx(
+        wav, summary_dir,
+        num_speakers=num_speakers,
+        on_progress=lambda line, _frac: on_progress(line, None) if on_progress else None,
+    )
+
+    diarized_path(session_dir).write_text(json.dumps(diarized, indent=2))
+    if on_progress:
+        on_progress("diarization done", 1.0)
+    return diarized
+
+
+def export(session_dir: Path,
+           *,
+           selected_frame_ids: set[str] | None = None,
+           name_map: dict[str, str] | None = None) -> Path:
+    """Merge cached diarization + selected frames and write enhanced.txt."""
+    if not has_diarization(session_dir):
+        raise RuntimeError("No diarization available — run diarization first.")
+    diarized = load_diarization(session_dir)
+    audio_segs = merger.whisperx_to_audio_segments(diarized)
+
+    frames = load_frame_index(session_dir / "frames")
+    if selected_frame_ids is not None:
+        frames = [f for f in frames if f["id"] in selected_frame_ids]
+    frame_segs = [
+        {"timestamp": f["timestamp"], "frame_path": str(f["path"])}
+        for f in frames
+    ]
+
+    merged = merger.merge(audio_segs, frame_segs, name_map=name_map)
+    return output.write_outputs(session_dir, merged)