LangGraph MongoDB Movie Agent

A conversational AI agent built with LangGraph and MongoDB Atlas that can search movies and maintain conversation memory. This tutorial demonstrates how to build an intelligent agent with vector search, full-text search, and persistent memory capabilities.

Features

• Vector Search: Semantic search through movie plots using embeddings
• Full-text Search: Direct title-based movie lookup
• Conversation Memory: Persistent memory across sessions using MongoDB
• Tool Selection Logging: See which tools the agent chooses to use
• Session Management: Maintain conversation context with thread IDs

Prerequisites

• Python 3.8+
• MongoDB Atlas cluster with the sample_mflix dataset
• OpenAI API key (for Azure OpenAI)
• Voyage AI API key (for embeddings)

Setup

1. Clone the Repository

git clone <your-repo-url>
cd langgraph-agent-0

2. Install Dependencies

pip install -r requirements.txt

3. Environment Variables

Create a .env file in the project root:

MONGODB_URI=mongodb+srv://username:[email protected]/
OPENAI_API_KEY=your_openai_api_key
AZURE_OPENAI_ENDPOINT=your_azure_endpoint
AZURE_OPENAI_API_KEY=your_azure_api_key
VOYAGEAI_API_KEY=your_voyage_api_key

4. MongoDB Setup

Ensure your MongoDB Atlas cluster has:

• The sample_mflix database with embedded_movies collection
• Movie documents with plot, title, and fullplot fields
• Existing plot embeddings in plot_embedding_voyage_3_large field

Usage

Running the Agent

python main.py

The application will:

1. Set up search indexes (first run may take ~1-2 minutes)
2. Prompt for a session ID
3. Start the conversation interface

Example Interactions

Enter a session ID: user123
Ask me about movies! Type 'quit' to exit.

Your question: recommend a funny movie
🔧 Using tools: plot_search

Answer: I recommend "Caddyshack" - a hilarious golf comedy with great characters and memorable humor!

Your question: save that I like comedies
🔧 Using tools: save_memory

Answer: Got it! I've saved that you enjoy comedies.

Your question: what do you know about my preferences?
🔧 Using tools: retrieve_memories

Answer: I remember that you like comedies.

Architecture

Components

• main.py: Entry point and conversation loop
• config.py: MongoDB setup, models, and index creation
• agent.py: LangGraph workflow definition
• tools.py: Movie search tools (vector and full-text)
• memory.py: Conversation memory tools

Tools Available

1. plot_search: Vector similarity search through movie plots
2. title_search: Full-text search by movie title
3. save_memory: Store important conversation information
4. retrieve_memories: Recall stored memories

LangGraph Workflow

User Input → Agent → Tool Selection → Tool Execution → Response
     ↑                                                      ↓
     ←─────────── Memory Persistence ←─────────────────────

Configuration

Models Used

• LLM: Azure OpenAI GPT-4
• Embeddings: Voyage AI voyage-3-large (2048 dimensions)
• Vector Store: MongoDB Atlas Vector Search
• Memory: MongoDB Atlas with vector indexing

Search Indexes

The application automatically creates:

• Vector Index: For semantic plot search
• Full-text Index: For title-based search

Development

Project Structure

langgraph-agent-0/
├── main.py              # Application entry point
├── config.py            # Configuration and setup
├── agent.py             # LangGraph agent definition
├── tools.py             # Movie search tools
├── memory.py            # Memory management tools
├── requirements.txt     # Python dependencies
├── .env                 # Environment variables (create this)
├── .gitignore          # Git ignore rules
└── README.md           # This file

Adding New Tools

1. Define your tool function with the @tool decorator
2. Add it to the appropriate tools list (TOOLS or MEMORY_TOOLS)
3. The agent will automatically discover and use it

Customizing the Agent

• Modify the system prompt in agent.py
• Adjust search parameters in tools.py
• Change memory configuration in memory.py

Troubleshooting

Common Issues

1. Index Creation Timeout: Increase wait_until_complete values in config.py
2. Memory Errors: Check MongoDB connection and permissions
3. Search Issues: Verify sample_mflix dataset is properly loaded

Logs

The agent shows tool selection for transparency:

🔧 Using tools: plot_search, save_memory

License

MIT License - see LICENSE file for details.

Contributing

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Add tests if applicable
5. Submit a pull request