init

app/audio.py

@@ -0,0 +1,104 @@
+import shutil
+import struct
+import subprocess
+from io import BytesIO
+
+
+def _find_ffmpeg(configured: str = "ffmpeg") -> str:
+    if shutil.which(configured):
+        return configured
+    try:
+        from imageio_ffmpeg import get_ffmpeg_exe
+        return get_ffmpeg_exe()
+    except Exception:
+        raise RuntimeError(
+            "ffmpeg not found: install it on PATH, set runtime.ffmpeg_path "
+            "in config.json, or `pip install imageio-ffmpeg`"
+        ) from None
+
+
+_SUPPORTED_FORMATS = {"mp3", "wav", "pcm", "opus", "aac", "flac"}
+
+_CONTENT_TYPES = {
+    "mp3": "audio/mpeg",
+    "wav": "audio/wav",
+    "pcm": "audio/pcm",
+    "opus": "audio/opus",
+    "aac": "audio/aac",
+    "flac": "audio/flac",
+}
+
+_FFMPEG_CODECS = {
+    "mp3": "libmp3lame",
+    "opus": "libopus",
+    "aac": "aac",
+    "flac": "flac",
+}
+
+
+def pcm_to_wav(pcm: bytes, sample_rate: int = 24000) -> bytes:
+    data_size = len(pcm)
+    byte_rate = sample_rate * 2
+    header = struct.pack(
+        "<4sI4s4sI4sI",
+        b"RIFF",
+        36 + data_size,
+        b"WAVE",
+        b"fmt ",
+        16,
+        b"data",
+        data_size,
+    )
+    fmt = struct.pack("<HHIIHH", 1, 1, sample_rate, byte_rate, 2, 16)
+    return header[:20] + fmt + header[20:] + pcm
+
+
+def pcm_to_raw(pcm: bytes) -> bytes:
+    return pcm
+
+
+def pcm_to_ffmpeg(
+    pcm: bytes, fmt: str, sample_rate: int = 24000, ffmpeg_path: str = "ffmpeg"
+) -> bytes:
+    codec = _FFMPEG_CODECS.get(fmt)
+    if codec is None:
+        raise ValueError(f"Unsupported ffmpeg format: {fmt}")
+
+    cmd = [
+        ffmpeg_path, "-y",
+        "-f", "s16le",
+        "-ar", str(sample_rate),
+        "-ac", "1",
+        "-i", "pipe:0",
+        "-f", _ffprobe_fmt(fmt),
+        "-codec:a", codec,
+        "pipe:1",
+    ]
+
+    proc = subprocess.run(
+        cmd,
+        input=pcm,
+        stdout=subprocess.PIPE,
+        stderr=subprocess.PIPE,
+    )
+    if proc.returncode != 0:
+        raise RuntimeError(f"ffmpeg failed: {proc.stderr.decode(errors='replace')}")
+    return proc.stdout
+
+
+def _ffprobe_fmt(fmt: str) -> str:
+    return {"mp3": "mp3", "opus": "opus", "aac": "adts", "flac": "flac"}.get(fmt, fmt)
+
+
+def encode(pcm: bytes, fmt: str, sample_rate: int = 24000, ffmpeg_path: str = "ffmpeg") -> tuple[bytes, str]:
+    if fmt not in _SUPPORTED_FORMATS:
+        raise ValueError(f"Unsupported response_format: {fmt}")
+
+    if fmt == "wav":
+        data = pcm_to_wav(pcm, sample_rate)
+    elif fmt == "pcm":
+        data = pcm
+    else:
+        data = pcm_to_ffmpeg(pcm, fmt, sample_rate, _find_ffmpeg(ffmpeg_path))
+
+    return data, _CONTENT_TYPES[fmt]