app.py

import os
from fastapi import FastAPI, Request
from fastapi.responses import JSONResponse, HTMLResponse
from fastapi.staticfiles import StaticFiles
from pathlib import Path
from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline

# Model: Falcon-rw-1b (decoder-only, not instruction-tuned)
model_id = "TheBloke/Mistral-7B-Instruct-v0.2-GGUF"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id)

pipe = pipeline("text-generation", model=model, tokenizer=tokenizer)

# Init FastAPI app
app = FastAPI()

# Mount static directory
app.mount("/static", StaticFiles(directory="."), name="static")

# Serve HTML UI
@app.get("/", response_class=HTMLResponse)
async def serve_page():
    return HTMLResponse(Path("index.html").read_text())

# Chat API
@app.post("/api")
async def ask_ai(request: Request):
    data = await request.json()
    question = data.get("question", "").strip()

    if not question:
        return JSONResponse(content={"answer": "❗ Please enter a valid question."})

    try:
        # Falcon works with plain prompts
        output = pipe(question, max_new_tokens=256, return_full_text=False)[0]["generated_text"]
        return JSONResponse(content={"answer": output.strip()})
    except Exception as e:
        return JSONResponse(content={"answer": f"⚠️ Error: {str(e)}"})