almost back to working state with rust transport

cht/stream/ffmpeg.py

@@ -178,15 +178,45 @@ def extract_scene_frames(input_path, output_dir, scene_threshold=0.10,
     return stdout.decode("utf-8", errors="replace"), stderr.decode("utf-8", errors="replace")
 
 
+def detect_scenes_from_pipe(scene_threshold=0.10, flush_frames=2, fps=30):
+    """Scene-detect from piped raw H.264 on stdin. Returns a node for run_async.
+
+    Used when Rust server provides a live H.264 stream via Unix socket.
+    Caller bridges the socket to ffmpeg's stdin and reads stdout/stderr:
+      - stdin:  raw H.264 from the socket
+      - stdout: MJPEG pipe (JPEG frames on scene change)
+      - stderr: showinfo lines with pts_time timestamps
+    """
+    stream = ffmpeg.input("pipe:0", f="h264", framerate=fps, hwaccel="cuda")
+    scene_expr = f"gt(scene,{scene_threshold})"
+    if flush_frames > 0:
+        mod_val = 1 + flush_frames
+        flush_expr = f"eq(n,prev_selected_n+1)*mod(selected_n,{mod_val})"
+        select_expr = f"{scene_expr}+{flush_expr}"
+    else:
+        select_expr = scene_expr
+    scene_stream = stream.filter("select", select_expr).filter("showinfo")
+    return ffmpeg.output(
+        scene_stream, "pipe:1",
+        f="image2pipe", vcodec="mjpeg",
+        flush_packets=1, strict="unofficial",
+        **{"q:v": "2", "fps_mode": "passthrough"},
+    ).global_args(*GLOBAL_ARGS)
+
+
 def extract_audio_chunk(input_path, output_path, start_time=0.0, duration=None):
     """Extract audio from recording as 16kHz mono WAV (optimal for Whisper).
 
     Uses input-level seeking (-ss before -i) for fast keyframe-based seek.
+    Supports fMP4 (auto-detect) and raw AAC files (explicit format hint).
     Returns (stdout, stderr) as decoded strings.
     """
     kwargs = {"ss": start_time}
     if duration is not None:
         kwargs["t"] = duration
+    # Raw AAC files need explicit format hint
+    if str(input_path).endswith(".aac"):
+        kwargs["f"] = "aac"
     stream = ffmpeg.input(str(input_path), **kwargs)
     output = (
         ffmpeg.output(
@@ -219,10 +249,11 @@ def extract_frame_at(input_path, output_path, timestamp):
     output.run(capture_stdout=True, capture_stderr=True)
 
 
-def run_async(output_node, pipe_stdout=False, pipe_stderr=False):
+def run_async(output_node, pipe_stdin=False, pipe_stdout=False, pipe_stderr=False):
     """Start an ffmpeg pipeline asynchronously via ffmpeg-python's run_async."""
     log.info("run_async: %s", " ".join(output_node.compile()))
     return output_node.run_async(
+        pipe_stdin=pipe_stdin,
         pipe_stdout=pipe_stdout,
         pipe_stderr=pipe_stderr,
     )