from fastapi import FastAPI
from pydantic import BaseModel
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

app = FastAPI()

# نموذج مفتوح المصدر
MODEL_NAME = "TheBloke/vicuna-7B-1.1-HF"
tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
model = AutoModelForCausalLM.from_pretrained(MODEL_NAME, device_map="auto", torch_dtype=torch.float16)

# تخزين تاريخ المحادثة
chat_histories = {}

class ChatRequest(BaseModel):
    user_id: str
    message: str

@app.post("/chat")
def chat_endpoint(request: ChatRequest):
    user_id = request.user_id
    message = request.message

    if user_id not in chat_histories:
        chat_histories[user_id] = []

    conversation = "NOVA AI: أنا كوميدي ومغربي. نفهم أي حاجة.\n"
    for q, a in chat_histories[user_id]:
        conversation += f"User: {q}\nNOVA AI: {a}\n"
    conversation += f"User: {message}\nNOVA AI:"

    inputs = tokenizer(conversation, return_tensors="pt").to(model.device)
    outputs = model.generate(**inputs, max_new_tokens=200)
    response = tokenizer.decode(outputs[0], skip_special_tokens=True).split("NOVA AI:")[-1].strip()

    chat_histories[user_id].append((message, response))
    if len(chat_histories[user_id]) > 10:
        chat_histories[user_id] = chat_histories[user_id][-10:]

    return {"response": response}