app.py

import torch
from fastapi import FastAPI, Request, Form
from fastapi.responses import HTMLResponse, JSONResponse
from fastapi.staticfiles import StaticFiles
from fastapi.templating import Jinja2Templates
from transformers import AutoTokenizer, AutoModelForCausalLM

# ---------------------------
# Models
# ---------------------------
MODEL_OPTIONS = {
    "DeepSeek Coder 1.3B": "deepseek-ai/deepseek-coder-1.3b-instruct",
    "StarCoder 1B": "bigcode/starcoderbase-1b",
    "CodeLLaMA 7B": "codellama/CodeLlama-7b-Instruct-hf"
}
loaded_models = {}

def get_model(model_key):
    if model_key not in loaded_models:
        model_name = MODEL_OPTIONS[model_key]
        tokenizer = AutoTokenizer.from_pretrained(model_name)
        model = AutoModelForCausalLM.from_pretrained(
            model_name,
            torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32,
            device_map="auto"
        )
        loaded_models[model_key] = (tokenizer, model)
    return loaded_models[model_key]

# ---------------------------
# FastAPI setup
# ---------------------------
app = FastAPI()
app.mount("/static", StaticFiles(directory="static"), name="static")
templates = Jinja2Templates(directory=".")

# ---------------------------
# Routes
# ---------------------------
@app.get("/", response_class=HTMLResponse)
async def home(request: Request):
    return templates.TemplateResponse("index.html", {"request": request, "models": list(MODEL_OPTIONS.keys())})

@app.post("/chat")
async def chat(user_input: str = Form(...), model_choice: str = Form(...), history: str = Form("[]")):
    import json
    history = json.loads(history)

    tokenizer, model = get_model(model_choice)

    # Build messages
    messages = []
    for role, content in history:
        messages.append({"role": "user" if role == "user" else "assistant", "content": content})
    messages.append({"role": "user", "content": user_input})

    # Tokenize
    inputs = tokenizer.apply_chat_template(
        messages,
        add_generation_prompt=True,
        tokenize=True,
        return_dict=True,
        return_tensors="pt"
    ).to(model.device)

    outputs = model.generate(**inputs, max_new_tokens=512, temperature=0.7, top_p=0.9)
    response = tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:], skip_special_tokens=True)

    # Update history
    history.append(("user", user_input))
    history.append(("assistant", response))

    return JSONResponse({"response": response, "history": history})