Movie Search Agent

A Langchain and Streamlit powered movie search agent that helps users find and get information about movies.

Features

• Movie search functionality with detailed information retrieval
• Integration with Google Search, DuckDuckGo and YouTube for comprehensive results
• Real-time streaming responses
• Interactive Streamlit interface
• Powered by LangChain and OpenAI for intelligent processing

Architecture

The application follows a modular architecture:

┌─────────────────┐     ┌───────────────────┐     ┌───────────────────┐
│                 │     │                   │     │                   │
│  Streamlit UI   │────▶│  MovieSearchAgent │────▶│  Search Tools     │
│                 │     │                   │     │                   │
└─────────────────┘     └───────────────────┘     └───────────────────┘
                               │                          │
                               ▼                          ▼
                        ┌─────────────┐           ┌─────────────────┐
                        │             │           │                 │
                        │  LLM Model  │           │  External APIs  │
                        │             │           │                 │
                        └─────────────┘           └─────────────────┘

• Streamlit UI: Provides the user interface for interacting with the agent
• MovieSearchAgent: Core component that processes user queries and coordinates responses
• Search Tools: Collection of tools for retrieving information from various sources
• LLM Model: Language model that powers the agent's understanding and response generation
• External APIs: Google Search, DuckDuckGo and YouTube APIs for retrieving movie information

Setup

Prerequisites

• Python 3.8 or higher
• API keys for:
  • OpenAI or OpenRouter (for the language model)
  • Google Search API
  • Google Custom Search Engine ID

Installation

1. Clone the repository:

   git clone https://github.com/yourusername/movie-search-agent.git
   cd movie-search-agent

2. Create a virtual environment:

   python -m venv .venv

3. Activate the virtual environment: Windows (PowerShell):

   .venv\Scripts\Activate.ps1

   Mac/Linux:

   source .venv/bin/activate

4. Install dependencies:

   pip install -r requirements.txt

5. Create a .env file and add your API keys, you can find required keys in the .env.example file in the root of the project.

Note: Google Search API and Custom Search Engine ID are optional, you can use DuckDuckGo as a search engine.

Running the Application

Start the Streamlit app:

streamlit run src/streamlit.py

The application will be available at http://localhost:8501

Challenges Faced

• Incomplete Search Results: Search queries sometimes returned incomplete information (e.g., missing movie ratings), requiring prompt engineering to get more detailed information.

• Tool Naming Conventions: Resolved issues with function calling by implementing proper naming conventions for tools to ensure correct invocation and prompt engineering.

• StreamlitCallBackHandler Compatibility: Overcame integration challenges between the latest LangChain agent implementation and Streamlit's callback handler by implementing a custom solution based on community contributions. Github Issue

• API Rate Limiting: Faced rate limiting issues with the APIs, so implemented multiple providers e.g Google Search and DuckDuckGo.

Usage Examples

Finding Movie Information

1. Enter a query like "Tell me about Inception"
2. The agent will search for information about the movie and return details including:
   • Release date
   • Director and cast
   • IMDB rating
   • Genre
   • Plot summary
   • Trailer link

Comparing Movies

1. Enter a query like "Compare The Matrix and Inception"
2. The agent will search for information about both movies and provide a comparison

Finding Movies by Criteria

1. Enter a query like "What are some good sci-fi movies from the 90s?"
2. The agent will search for and recommend movies matching your criteria

Contributing

1. Fork the repository
2. Create your feature branch
3. Commit your changes
4. Push to the branch
5. Create a new Pull Request

License

This project is licensed under the MIT License - see the LICENSE file for details.