app.py

"""
Yelp Reviews Semantic Search Engine
AIPI 510 - Week 13 Project
"""

import gradio as gr
from sentence_transformers import SentenceTransformer
import chromadb
from pathlib import Path

# Load the embedding model
print("Loading embedding model...")
model = SentenceTransformer('sentence-transformers/all-MiniLM-L6-v2')

# Resolve ChromaDB storage
print("Preparing ChromaDB storage...")
BASE_DIR = Path(__file__).resolve().parent
REPO_ROOT = BASE_DIR.parent

client = None
collection = None
tried_paths = []

print("Loading ChromaDB collection...")
candidate_client = chromadb.PersistentClient(path=str(BASE_DIR / "chromadb"))
collection = candidate_client.get_or_create_collection("yelp_reviews")
client = candidate_client

if collection is None or client is None:
    raise RuntimeError(
        "Could not find collection 'yelp_reviews'. Checked the following locations: "
        + (", ".join(tried_paths) if tried_paths else "<none>")
    )

def semantic_search(query, n_results=3):
    """
    Search Yelp reviews using semantic similarity
    
    Args:
        query: Search query string
        n_results: Number of top results to return
    
    Returns:
        Formatted string with top matching reviews
    """
    if not query.strip():
        return "Please enter a search query."
    
    # Encode the query
    query_embedding = model.encode([query])
    
    # Query the collection
    results = collection.query(
        query_embeddings=query_embedding.tolist(),
        n_results=n_results
    )
    
    # Format the results
    output = []
    for i, (doc, metadata) in enumerate(zip(results['documents'][0], results['metadatas'][0])):
        output.append(f"### Result {i+1}")
        output.append(doc)
        output.append("-" * 80)
    
    return "\n\n".join(output) if output else "No results found."

def search_interface(query, num_results):
    """Wrapper function for Gradio interface"""
    return semantic_search(query, n_results=int(num_results))

# Create Gradio interface
demo = gr.Interface(
    fn=search_interface,
    inputs=[
        gr.Textbox(
            label="Search Query",
            placeholder="e.g., 'great pizza' or 'romantic atmosphere' or 'friendly staff'",
            lines=2
        ),
        gr.Slider(
            minimum=1,
            maximum=10,
            value=3,
            step=1,
            label="Number of Results"
        )
    ],
    outputs=gr.Textbox(label="Top Matching Reviews", lines=20),
    title="🔍 Yelp Reviews Semantic Search",
    description="""Search through Yelp business reviews using AI-powered semantic search. 
    Enter natural language queries to find relevant reviews based on semantic meaning, not just keywords.""",
    examples=[
        ["great food and friendly service", 3],
        ["romantic atmosphere perfect for date night", 3],
        ["fast service and good prices", 3],
        ["disappointed with the quality", 3],
        ["authentic Italian cuisine", 3],
    ],
    theme=gr.themes.Soft(),
    article="""
    ### About This App
    This semantic search engine uses:
    - **Dataset**: Yelp Review Full (650K reviews with 1-5 star ratings)
    - **Embeddings**: sentence-transformers/all-MiniLM-L6-v2
    - **Vector DB**: ChromaDB
    - **Interface**: Gradio
    
    Built for AIPI 510: Data Sourcing for Analytics - Week 13
    """
)

if __name__ == "__main__":
    demo.launch()