🎬 Movie Q&A Agent — Hybrid RAG + Live Search Powered by OpenRouter & ChromaDB

A smart, Retrieval-Augmented Generation (RAG) based chatbot that answers natural language questions about movies and TV series using real movie metadata and live internet search fallback. Built with ChromaDB for semantic vector search over the TMDB 5000 Movie Dataset, OpenRouter for LLM access, and ARES API for live internet queries.

---

🚀 Features

• 🔍 Hybrid RAG + Live Search: Answers come from vector similarity search over TMDB data or live internet results when data isn’t found locally.
• 🧠 Query Classification: Routes queries intelligently to:
  • Movie/TV metadata search (ChromaDB)
  • OpenRouter LLM chat
  • ARES live internet search
• 🌐 Flexible LLM Access: Use GPT-4, Mixtral, Claude, or other OpenRouter-supported models seamlessly.
• 🎥 Real Movie Dataset: Integrates TMDB metadata with cast, crew, genres, and overviews.
• 🛠️ Easy to Extend: Modular code for classification, retrieval, generation, and web search fallback.

---

📂 Dataset

This project uses the following Kaggle dataset files:

• tmdb_5000_movies.csv: Core movie metadata including titles, genres, and plot summaries.
• tmdb_5000_credits.csv: Cast and crew details.

📥 Download the TMDB Movie Dataset from Kaggle

---

⚙️ How It Works

1. Data Preparation

   • Merge movies and credits datasets
   • Clean and extract key fields (title, genres, overview, top cast & crew)
   • Build single-document text blobs for each movie/series

2. Embedding & Vector DB

   • Use sentence-transformers (all-MiniLM-L6-v2) to create vector embeddings for each movie document
   • Store embeddings and metadata in ChromaDB for fast semantic search

3. Query Classification

   • Use an LLM-based router prompt to classify queries into:
     • MOVIE_QUERY (search in ChromaDB)
     • INTERNET_QUERY (fetch live info using ARES API)
     • GENERIC_CHAT (fallback to chat with LLM)

4. Query Routing & Response

   • For movie queries, retrieve top relevant movie chunks and generate answer via OpenRouter
   • For internet queries, forward the question to ARES live search API and return clean, readable responses
   • For generic queries, respond with LLM chat completions

---

🧪 Example Usage

query = "When is the next season of Stranger Things coming?"

response = movie_agent(query)
print(response['recommendations'])

# Clean output:
# The next season of "Stranger Things," Season 5, is expected in 2025.