app.py

from fastapi import FastAPI, Query
from fastapi.middleware.cors import CORSMiddleware
import pandas as pd

from filtered_search_engine import SmartRecommender
from reranker import Reranker
from intent_classifier import IntentClassifier
from keyword_boosting_layer import apply_keyword_boost

# ------------------------------
# Initialize App
# ------------------------------
app = FastAPI(
    title="Salahkar AI Recommender",
    description="Smart cultural, heritage & food recommendation engine for BharatVerse",
    version="1.0.0"
)

from fastapi.staticfiles import StaticFiles

# CORS Support (allows frontend browser access)
app.add_middleware(
    CORSMiddleware,
    allow_origins=["*"],  
    allow_credentials=True,
    allow_methods=["*"],
    allow_headers=["*"],
)

# Mount images folder to serve static files
app.mount("/images", StaticFiles(directory="images"), name="images")

# ------------------------------
# Load Core Components Once
# ------------------------------
print("📌 Loading dataset...")
df = pd.read_csv("salahkar_enhanced.csv")

print("📌 Loading smart recommendation engine...")
engine = SmartRecommender()

print("📌 Loading reranker model...")
reranker = Reranker()

print("📌 Loading intent recognizer...")
intent_detector = IntentClassifier()

print("🚀 Salahkar AI Ready!")


# ------------------------------
# Routes
# ------------------------------

@app.get("/")
def root():
    return {
        "message": "🇮🇳 Welcome to Salahkar AI – BharatVerse Intelligent Recommendation System",
        "usage": "/recommend?query=your text"
    }


@app.get("/recommend")
def get_recommendation(query: str = Query(..., description="User's search text"), k: int = 7):

    print(f"\n🔍 User Query: {query}")

    # 1️⃣ Detect intent
    detected_intent = intent_detector.predict_intent(query)
    print(f"🧠 Intent Detected: {detected_intent}")

    # 2️⃣ FAISS + Filter Search
    results = engine.recommend(query, k=k)

    # 3️⃣ Prepare results for reranker
    prepared = []
    for name, domain, category, region, score in results:
        row = df[df["name"] == name].iloc[0]
        prepared.append({
            "name": name,
            "domain": domain,
            "category": category,
            "region": region,
            "embedding_score": float(score),
            "text": row["search_embedding_text"],
            "image": row["image_file"]
        })

    # 4️⃣ Re-rank using cross encoder
    reranked_results = reranker.rerank(query, prepared)

    # 5️⃣ Apply keyword boosting
    final_results = apply_keyword_boost(query, reranked_results)

    # 6️⃣ Format response for frontend
    response = [
        {
            "name": item["name"],
            "category": item["category"],
            "domain": item["domain"],
            "region": item["region"],
            "score": float(item["final_score"]),
            "image": f"/images/{item['image']}" if item.get("image") else None
        }
        for item in final_results[:k]
    ]

    return {
        "query": query,
        "intent": detected_intent,
        "results": response
    }


# -------------------------------------------
# Run (Ignored by HuggingFace — needed only for local testing)
# -------------------------------------------
if __name__ == "__main__":
    import uvicorn
    uvicorn.run(app, host="0.0.0.0", port=7860)