api.py

from fastapi import FastAPI, HTTPException, UploadFile, File, Form
from fastapi.middleware.cors import CORSMiddleware
from pydantic import BaseModel
from typing import Optional, Dict, Any
import json
import base64
from PIL import Image
from io import BytesIO
import uvicorn
from app import llm_client

# Create FastAPI application
api_app = FastAPI(
    title="LLM Structured Output API",
    description="API for generating structured responses from local GGUF models via llama-cpp-python",
    version="1.0.0"
)

# Setup CORS
api_app.add_middleware(
    CORSMiddleware,
    allow_origins=["*"],
    allow_credentials=True,
    allow_methods=["*"],
    allow_headers=["*"],
)

# Data models for API
class StructuredOutputRequest(BaseModel):
    prompt: str
    json_schema: Dict[str, Any]
    image_base64: Optional[str] = None
    use_grammar: bool = True

class StructuredOutputResponse(BaseModel):
    success: bool
    data: Optional[Dict[str, Any]] = None
    error: Optional[str] = None
    raw_response: Optional[str] = None

def decode_base64_image(base64_string: str) -> Image.Image:
    """Decode base64 string to PIL Image"""
    try:
        image_data = base64.b64decode(base64_string)
        image = Image.open(BytesIO(image_data))
        return image
    except Exception as e:
        raise HTTPException(status_code=400, detail=f"Image decoding error: {str(e)}")

@api_app.post("/generate", response_model=StructuredOutputResponse)
async def generate_structured_output(request: StructuredOutputRequest):
    """
    Main endpoint for generating structured response
    
    Args:
        request: Request containing prompt, JSON schema and optionally base64 image
    
    Returns:
        StructuredOutputResponse: Structured response or error
    """
    # Check model initialization
    if llm_client is None:
        raise HTTPException(
            status_code=503, 
            detail="LLM model not initialized. Check server configuration."
        )
    
    try:
        # Validate input data
        if not request.prompt.strip():
            raise HTTPException(status_code=400, detail="Prompt cannot be empty")
        
        if not request.json_schema:
            raise HTTPException(status_code=400, detail="JSON schema cannot be empty")
        
        # Decode image if provided
        image = None
        if request.image_base64:
            image = decode_base64_image(request.image_base64)
        
        # Generate response
        result = llm_client.generate_structured_response(
            prompt=request.prompt,
            json_schema=request.json_schema,
            image=image,
            use_grammar=request.use_grammar
        )
        
        # Format response
        if "error" in result:
            return StructuredOutputResponse(
                success=False,
                error=result["error"],
                raw_response=result.get("raw_response")
            )
        else:
            return StructuredOutputResponse(
                success=True,
                data=result.get("data"),
                raw_response=result.get("raw_response")
            )
            
    except HTTPException:
        raise
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Internal server error: {str(e)}")

@api_app.post("/generate_with_file", response_model=StructuredOutputResponse)
async def generate_with_file(
    prompt: str = Form(...),
    json_schema: str = Form(...),
    image: Optional[UploadFile] = File(None),
    use_grammar: bool = Form(True)
):
    """
    Alternative endpoint for uploading image as file
    
    Args:
        prompt: Text prompt
        json_schema: JSON schema as string
        image: Uploaded image file
        use_grammar: Whether to use grammar-based structured output
    
    Returns:
        StructuredOutputResponse: Structured response or error
    """
    # Check model initialization
    if llm_client is None:
        raise HTTPException(
            status_code=503, 
            detail="LLM model not initialized. Check server configuration."
        )
    
    try:
        # Validate input data
        if not prompt.strip():
            raise HTTPException(status_code=400, detail="Prompt cannot be empty")
        
        if not json_schema.strip():
            raise HTTPException(status_code=400, detail="JSON schema cannot be empty")
        
        # Parse JSON schema
        try:
            parsed_schema = json.loads(json_schema)
        except json.JSONDecodeError as e:
            raise HTTPException(status_code=400, detail=f"Invalid JSON schema: {str(e)}")
        
        # Process image if provided
        pil_image = None
        if image:
            # Check file type
            if not image.content_type.startswith('image/'):
                raise HTTPException(status_code=400, detail="Uploaded file must be an image")
            
            # Read and convert image
            image_data = await image.read()
            pil_image = Image.open(BytesIO(image_data))
        
        # Generate response
        result = llm_client.generate_structured_response(
            prompt=prompt,
            json_schema=parsed_schema,
            image=pil_image,
            use_grammar=use_grammar
        )
        
        # Format response
        if "error" in result:
            return StructuredOutputResponse(
                success=False,
                error=result["error"],
                raw_response=result.get("raw_response")
            )
        else:
            return StructuredOutputResponse(
                success=True,
                data=result.get("data"),
                raw_response=result.get("raw_response")
            )
            
    except HTTPException:
        raise
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Internal server error: {str(e)}")

@api_app.get("/health")
async def health_check():
    """API health check"""
    model_status = "loaded" if llm_client is not None else "not_loaded"
    return {
        "status": "healthy" if llm_client is not None else "degraded",
        "model_status": model_status,
        "message": "API is working correctly" if llm_client is not None else "API is working, but model is not loaded"
    }

@api_app.get("/")
async def root():
    """Root endpoint with API information"""
    return {
        "message": "LLM Structured Output API",
        "version": "1.0.0",
        "model_loaded": llm_client is not None,
        "endpoints": {
            "/generate": "POST - main endpoint for generating structured response",
            "/generate_with_file": "POST - endpoint with image file upload",
            "/health": "GET - health check",
            "/docs": "GET - automatic Swagger documentation"
        }
    }

if __name__ == "__main__":
    from config import Config
    uvicorn.run(
        "api:api_app",
        host=Config.HOST,
        port=Config.API_PORT,
        reload=True
    )