app.py

import os

# Set cache directory to a writable folder inside the app directory
CACHE_DIR = os.path.join(os.getcwd(), ".cache")
os.environ["XDG_CACHE_HOME"] = CACHE_DIR
os.environ["TRANSFORMERS_CACHE"] = CACHE_DIR
os.environ["HF_HOME"] = CACHE_DIR

# Create the cache directory if it doesn't exist
os.makedirs(CACHE_DIR, exist_ok=True)

from fastapi import FastAPI
from fastapi.responses import JSONResponse
import whisper
from transformers import pipeline
import subprocess

# Check ffmpeg
def check_ffmpeg():
    try:
        subprocess.run(["ffmpeg", "-version"], check=True, stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)
    except Exception:
        raise RuntimeError("ffmpeg is not installed or not found in PATH. It is required for Whisper to process audio/video.")

check_ffmpeg()

app = FastAPI(title="Auto Audio/Video Listener API")

# Load models AFTER setting cache environment variables
asr_model = whisper.load_model("base")
summarizer = pipeline("summarization", model="facebook/bart-large-cnn")

AUDIO_FILE_PATH = "./uploads/input_media.mp4"

@app.get("/auto_summarize")
async def auto_summarize():
    try:
        if not os.path.exists(AUDIO_FILE_PATH):
            return JSONResponse(status_code=404, content={"error": "Media file not found."})

        result = asr_model.transcribe(AUDIO_FILE_PATH)
        transcription = result.get("text", "")

        summary = summarizer(
            transcription, max_length=150, min_length=30, do_sample=False
        )[0]['summary_text']

        return JSONResponse(content={"transcription": transcription, "summary": summary})

    except Exception as e:
        return JSONResponse(status_code=500, content={"error": str(e)})