cueseg: topic chapters from a transcript, and a worked pipeline in the readme

parity/oracle/cueseg.py

@@ -0,0 +1,48 @@
+"""Segment a transcript the way the Python pipeline does.
+
+Reads a cue stream — one `{"start", "end", "text"}` object per line — and a
+duration in seconds as the first argument, and writes one chapter per line.
+
+    CATENA_PARITY_PYTHONPATH=/path/to/podcast-pipeline/src \
+        python3 oracle/cueseg.py 7245 < cues.ndjson
+"""
+
+import json
+import os
+import sys
+
+for path in os.environ.get("CATENA_PARITY_PYTHONPATH", "").split(os.pathsep):
+    if path:
+        sys.path.insert(0, path)
+
+from podcast_pipeline.topic_segments import Cue, segment_transcript  # noqa: E402
+
+
+def main() -> int:
+    duration = float(sys.argv[1])
+    cues = []
+    for line in sys.stdin.buffer.read().decode("utf-8").split("\n"):
+        if not line.strip():
+            continue
+        cue = json.loads(line)
+        cues.append(Cue(cue["start"], cue["end"], cue["text"]))
+
+    out = sys.stdout
+    for seg in segment_transcript(cues, duration):
+        out.write(
+            json.dumps(
+                {
+                    "start": seg.start_time,
+                    "end": seg.end_time,
+                    "title": seg.fallback_title,
+                    "excerpt": seg.excerpt,
+                },
+                ensure_ascii=False,
+            )
+            + "\n"
+        )
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())