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

app = FastAPI()

# Laad jouw eigen getrainde model in
model_path = "./SpaceStar-0.01-Final"
tokenizer = AutoTokenizer.from_pretrained(model_path)
model = AutoModelForCausalLM.from_pretrained(model_path).to("cpu")

class ChatRequest(BaseModel):
    message: str
    system_prompt: str = "Je bent SpaceStar 0.01, een geavanceerde AI-assistent die gespecialiseerd is in programmeren, wiskunde en communicatie in het Nederlands en Engels."

@app.post("/chat")
async def chat(request: ChatRequest):
    # Formatteer het gesprek in de Qwen-stijl
    messages = [
        {"role": "system", "content": request.system_prompt},
        {"role": "user", "content": request.message}
    ]
    
    text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
    model_inputs = tokenizer([text], return_tensors="pt").to("cpu")
    
    # Genereer het antwoord
    generated_ids = model.generate(
        **model_inputs,
        max_new_tokens=256,
        temperature=0.7,
        top_p=0.9
    )
    
    # Filter de input-tokens weg zodat we alleen het antwoord overhouden
    generated_ids = [
        output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
    ]
    
    response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
    return {"response": response}

@app.get("/")
def home():
    return {"status": "SpaceStar 0.01 is online and ready!"}