api.py

from __future__ import annotations

import os
import tempfile

from fastapi import FastAPI, File, Form, HTTPException, UploadFile

try:
    from .config import VoiceRuntimeConfig
    from .service import process_diarization_only, process_voice
except ImportError:  # HF flat-root execution fallback
    from config import VoiceRuntimeConfig
    from service import process_diarization_only, process_voice

app = FastAPI(title="Voice Intelligence Module", version="1.0.0")

_ALLOWED = {".wav", ".mp3", ".m4a", ".ogg", ".flac", ".aac", ".mp4", ".mov", ".mkv", ".webm"}


def _save_upload_temp(upload: UploadFile) -> str:
    filename = upload.filename or "input_audio"
    ext = os.path.splitext(filename)[1].lower()
    if ext not in _ALLOWED:
        allowed = ", ".join(sorted(_ALLOWED))
        raise HTTPException(status_code=400, detail=f"Unsupported input extension: {ext or 'unknown'}. Allowed: {allowed}")

    fd, tmp_path = tempfile.mkstemp(prefix="voice-intel-upload-", suffix=ext)
    os.close(fd)

    with open(tmp_path, "wb") as f:
        while True:
            chunk = upload.file.read(1024 * 1024)
            if not chunk:
                break
            f.write(chunk)

    return tmp_path


@app.get("/health")
def health() -> dict:
    return {"ok": True, "module": "voice-intelligence"}


@app.post("/v1/voice/align/trimmed")
def align_trimmed(
    file: UploadFile = File(...),
    language_hint: str = Form("auto"),
    silence_threshold_db: float = Form(-40.0),
    min_silence_sec: float = Form(0.30),
    keep_padding_sec: float = Form(0.05),
    analysis_window_ms: int = Form(10),
    diarization_enabled: str = Form("true"),
    diarization_min_speakers: int = Form(0),
    diarization_max_speakers: int = Form(0),
) -> dict:
    tmp_path = _save_upload_temp(file)
    config = VoiceRuntimeConfig.from_env()
    config.silence_threshold_db = silence_threshold_db
    config.min_silence_sec = min_silence_sec
    config.keep_padding_sec = keep_padding_sec
    config.analysis_window_ms = analysis_window_ms
    config.diarization_enabled = diarization_enabled.strip().lower() in ("1", "true", "yes", "on")
    config.diarization_min_speakers = diarization_min_speakers
    config.diarization_max_speakers = diarization_max_speakers

    try:
        response = process_voice(
            input_audio_path=tmp_path,
            config=config,
            language_hint=language_hint,
            trim_silence_enabled=True,
            include_audio_payload=True,
            minimal_output=False,
        )
        response["endpoint"] = "/v1/voice/align/trimmed"
        response["description"] = "Returns word-level data and silence-removed WAV payload."
        return response
    except HTTPException:
        raise
    except Exception as exc:
        raise HTTPException(status_code=500, detail=f"trimmed endpoint failed: {exc}") from exc
    finally:
        try:
            os.remove(tmp_path)
        except OSError:
            pass


@app.post("/v1/voice/align/raw")
def align_raw(
    file: UploadFile = File(...),
    language_hint: str = Form("auto"),
    diarization_enabled: str = Form("true"),
    diarization_min_speakers: int = Form(0),
    diarization_max_speakers: int = Form(0),
) -> dict:
    tmp_path = _save_upload_temp(file)
    config = VoiceRuntimeConfig.from_env()
    config.diarization_enabled = diarization_enabled.strip().lower() in ("1", "true", "yes", "on")
    config.diarization_min_speakers = diarization_min_speakers
    config.diarization_max_speakers = diarization_max_speakers

    try:
        response = process_voice(
            input_audio_path=tmp_path,
            config=config,
            language_hint=language_hint,
            trim_silence_enabled=False,
            include_audio_payload=False,
            minimal_output=True,
        )
        response["endpoint"] = "/v1/voice/align/raw"
        response["description"] = "Returns only required alignment data; input audio remains unchanged."
        return response
    except HTTPException:
        raise
    except Exception as exc:
        raise HTTPException(status_code=500, detail=f"raw endpoint failed: {exc}") from exc
    finally:
        try:
            os.remove(tmp_path)
        except OSError:
            pass


@app.post("/v1/voice/diarization")
def diarization_only(
    file: UploadFile = File(...),
    diarization_enabled: str = Form("true"),
    diarization_min_speakers: int = Form(0),
    diarization_max_speakers: int = Form(0),
) -> dict:
    tmp_path = _save_upload_temp(file)
    config = VoiceRuntimeConfig.from_env()
    config.diarization_enabled = diarization_enabled.strip().lower() in ("1", "true", "yes", "on")
    config.diarization_min_speakers = diarization_min_speakers
    config.diarization_max_speakers = diarization_max_speakers

    try:
        response = process_diarization_only(
            input_audio_path=tmp_path,
            config=config,
        )
        response["endpoint"] = "/v1/voice/diarization"
        response["description"] = "Returns diarization only (speaker segments + summary), no transcription payload."
        return response
    except HTTPException:
        raise
    except Exception as exc:
        raise HTTPException(status_code=500, detail=f"diarization endpoint failed: {exc}") from exc
    finally:
        try:
            os.remove(tmp_path)
        except OSError:
            pass


if __name__ == "__main__":
    import uvicorn

    port = int(os.environ.get("PORT", "8091"))
    uvicorn.run("core.voice_intelligence.api:app", host="0.0.0.0", port=port, reload=False)