app.py

from fastapi import FastAPI, Request
from llama_cpp import Llama
import uvicorn

app = FastAPI()

# Model එක Load කිරීම (මෙයට විනාඩි කිහිපයක් ගතවේ)
llm = Llama(
    model_path="model.gguf",
    n_ctx=2048,
    n_threads=4  # CPU cores ගණන
)

@app.post("/v1/chat/completions")
async def chat(request: Request):
    data = await request.json()
    messages = data.get("messages", [])
    
    # Prompt එක සකසා ගැනීම
    full_prompt = ""
    for m in messages:
        full_prompt += f"{m['role']}: {m['content']}\n"
    full_prompt += "assistant: "

    # Model එකෙන් පිළිතුර ලබා ගැනීම
    output = llm(
        full_prompt,
        max_tokens=512,
        stop=["user:", "\n"],
        echo=False
    )
    
    response_text = output["choices"][0]["text"]
    
    return {
        "choices": [{
            "message": {
                "role": "assistant",
                "content": response_text.strip()
            }
        }]
    }

@app.get("/")
def home():
    return {"status": "MINZO-CORE v1.0 is Live"}