app.py

from fastapi import FastAPI, HTTPException
from pydantic import BaseModel
from ctransformers import AutoModelForCausalLM
import os
import uvicorn
from typing import Optional, List
import logging

# Set up loggings
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)

app = FastAPI(title="llm-apiku", version="1.0.0")

# Request model - fleksibel untuk menerima semua parameter
class TextRequest(BaseModel):
    inputs: str
    system_prompt: Optional[str] = None
    max_tokens: Optional[int] = 512
    temperature: Optional[float] = 0.7
    top_k: Optional[int] = 50
    top_p: Optional[float] = 0.9
    repeat_penalty: Optional[float] = 1.1
    stop: Optional[List[str]] = None

# Response model
class TextResponse(BaseModel):
    generated_text: str

# Global model variable
model = None

@app.on_event("startup")
async def load_model():
    global model
    try:
        logger.info("Loading model...")
        model = AutoModelForCausalLM.from_pretrained(
            "Dnfs/gema-4b-indra10k-model1-Q4_K_M-GGUF",
            model_file="gema-4b-indra10k-model1-q4_k_m.gguf",
            model_type="llama",
            gpu_layers=0,  # Set to appropriate number if using GPU
            context_length=2048,
            threads=os.cpu_count()
        )
        logger.info("Model loaded successfully!")
    except Exception as e:
        logger.error(f"Failed to load model: {e}")
        raise e

@app.post("/generate", response_model=TextResponse)
async def generate_text(request: TextRequest):
    if model is None:
        raise HTTPException(status_code=500, detail="Model not loaded")
    
    try:
        # Buat prompt - gunakan system_prompt jika ada, atau langsung input user
        if request.system_prompt:
            full_prompt = f"{request.system_prompt}\n\nUser: {request.inputs}\nAssistant:"
        else:
            full_prompt = request.inputs
        
        # Generate text dengan parameter dari request
        generated_text = model(
            full_prompt,
            max_new_tokens=request.max_tokens,
            temperature=request.temperature,
            top_p=request.top_p,
            top_k=request.top_k,
            repetition_penalty=request.repeat_penalty,
            stop=request.stop or []
        )
        
        # Bersihkan response dari system prompt jika ada
        if "Assistant:" in generated_text:
            generated_text = generated_text.split("Assistant:")[-1].strip()
        
        return TextResponse(generated_text=generated_text)
    
    except Exception as e:
        logger.error(f"Generation error: {e}")
        raise HTTPException(status_code=500, detail=f"Generation failed: {str(e)}")

@app.get("/health")
async def health_check():
    return {"status": "healthy", "model_loaded": model is not None}

@app.get("/")
async def root():
    return {"message": "Gema 4B Model API", "docs": "/docs"}

if __name__ == "__main__":
    uvicorn.run(app, host="0.0.0.0", port=8000, log_level="info")