import os
from fastapi import FastAPI
from pydantic import BaseModel
from huggingface_hub import InferenceClient
import uvicorn

app = FastAPI()

# ✅ CHANGE: Get API token from Hugging Face Spaces secret instead of .env
HF_TOKEN = os.getenv("HF_TOKEN")

if not HF_TOKEN:
    raise ValueError("Hugging Face API token (HF_TOKEN) is missing!")

# ✅ CHANGE: Pass the token to InferenceClient
client = InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1", token=HF_TOKEN)

class Item(BaseModel):
    prompt: str
    history: list
    system_prompt: str
    temperature: float = 0.0
    max_new_tokens: int = 1048
    top_p: float = 0.15
    repetition_penalty: float = 1.0

def format_prompt(message, history):
    prompt = "<s>"
    for user_prompt, bot_response in history:
        prompt += f"[INST] {user_prompt} [/INST] {bot_response}</s> "
    prompt += f"[INST] {message} [/INST]"
    return prompt

def generate(item: Item):
    temperature = max(float(item.temperature), 1e-2)
    top_p = float(item.top_p)

    generate_kwargs = {
        "temperature": temperature,
        "max_new_tokens": item.max_new_tokens,
        "top_p": top_p,
        "repetition_penalty": item.repetition_penalty,
        "do_sample": True,
        "seed": 42,
    }

    formatted_prompt = format_prompt(f"{item.system_prompt}, {item.prompt}", item.history)
    
    # ✅ CHANGE: Use token authentication for API call
    stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)
    
    output = "".join(response.token.text for response in stream)
    return output

@app.post("/generate/")
async def generate_text(item: Item):
    return {"response": generate(item)}

# ✅ ADDING TEST API ENDPOINT
@app.get("/test/")
async def test_api():
    return {"message": "API is working!"}

# ✅ ADDING MAIN CHECK FOR LOCAL TESTING
if __name__ == "__main__":
    uvicorn.run(app, host="0.0.0.0", port=7860)