app.py

import nh3
from fastapi import FastAPI, Query
from doc_searcher_v2 import DocSearcherV2
from huggingface_hub import login
from config import HUGGING_FACE_API_KEY, COLLECTION_NAME, API_KEY
import time
import logging

# Setup logging
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)

login(HUGGING_FACE_API_KEY)

app = FastAPI(
    title="Senatus Legal Search API",
    description="Semantička pretraga pravnih dokumenata sa Qwen3 reranker-om",
    version="2.0"
)

# Globalno inicijalizuj DocSearcher (učitava se samo jednom)
doc_searcher_v2 = DocSearcherV2(collection_name=COLLECTION_NAME)

ALLOWED_API_KEY = str(API_KEY)

@app.on_event("startup")
async def warm_up():
    """Warm up modela pri pokretanju aplikacije"""
    logger.info("🔥 Warming up models...")
    start = time.time()
    try:
        # Pokreni dummy query da učita sve modele u memoriju
        _ = await doc_searcher_v2.search_semantic("test", qdrant_limit=5, top_k=1)
        elapsed = time.time() - start
        logger.info(f"✅ Models ready! Warmup took {elapsed:.1f}s")
    except Exception as e:
        logger.error(f"❌ Warmup failed: {e}")

@app.get("/")
async def root():
    """Root endpoint za provjeru da li aplikacija radi"""
    return {
        "status": "ok",
        "message": "Senatus API is running",
        "version": "2.0",
        "endpoints": {
            "search": "/api/v2/search",
            "docs": "/docs",
            "health": "/health"
        }
    }

@app.get("/health")
async def health():
    """Health check endpoint"""
    return {
        "status": "healthy",
        "models_loaded": True,
        "version": "2.0"
    }

@app.get("/api/v2/search")
async def v2_search(
    q: str = Query(..., description="Query tekst za pretragu"),
    limit: int = Query(3, ge=1, le=10, description="Broj rezultata (1-10, default: 3)"),
    qdrant_limit: int = Query(20, ge=10, le=50, description="Broj chunk-ova iz Qdrant-a (10-50, default: 20)")
):
    """
    Semantička pretraga pravnih dokumenata
    
    - **q**: Tekstualni upit
    - **limit**: Broj rezultata za vraćanje (default: 3, max: 10)
    - **qdrant_limit**: Broj dokumenata iz Qdrant-a prije rerankinga (default: 20, max: 50)
    
    Returns: Lista rezultata sa skorovima, sortirana po relevantnosti
    """
    logger.info(f"Search query: '{q}' | limit={limit} | qdrant_limit={qdrant_limit}")
    start = time.time()
    
    # Sanitizacija inputa
    query = q.lower()
    xss = nh3.clean(query)
    
    # Pretraga sa parametrima
    data = await doc_searcher_v2.search_semantic(
        text=xss,
        qdrant_limit=qdrant_limit,
        top_k=limit
    )
    
    elapsed = time.time() - start
    logger.info(f"Search completed in {elapsed:.2f}s | Results: {len(data)}")
    
    return data