app.py

import os
os.environ["HF_HOME"] = "/tmp/hf_cache"
os.environ["TRANSFORMERS_CACHE"] = "/tmp/hf_cache"

from fastapi import FastAPI, File, UploadFile
from vision_model import describe_image
import shutil

app = FastAPI()

@app.post("/vision")
async def process_vision(image: UploadFile = File(...)):
    with open("frame.jpg", "wb") as f:
        shutil.copyfileobj(image.file, f)
    desc = describe_image("frame.jpg")
    return {"description": desc}