Agent-to-Agent-MCP

# SQL Database Agent with Google ADK Framework This project implements an intelligent SQL agent that can interpret natural language queries about a database, translate them into SQL, execute them, and provide human-friendly responses using the Google ADK (Agent Development Kit) framework. ## Project Overview  The agent uses a combination of components: - **Main Agent**: Handles natural language understanding and generates responses - **SQL Query Tool**: A tool that generates and executes SQL queries - **MCP Server**: Multi-Call Protocol server that provides tools to the main agent - **Few-Shot Examples**: Pre-loaded examples to help improve SQL query generation - **SQLite Database**: The Chinook music store sample database ## Features - Natural language to SQL conversion - Database schema introspection - Error handling and query refinement - Example-enhanced prompting for better SQL generation - Conversational presentation of query results ## Prerequisites - Python 3.10+ - Google Cloud account with Vertex AI API enabled - Google Cloud authentication configured ## Installation 1. Clone this repository: ```bash git clone <repository-url> cd sql-database-agent ``` 2. Install dependencies: ```bash pip install -r requirements.txt ``` 3. Set up Google Cloud authentication: ```bash # Option 1: Set up Application Default Credentials gcloud auth application-default login # Option 2: Use a service account key export GOOGLE_APPLICATION_CREDENTIALS=/path/to/your/service-account-key.json ``` 4. Configure environment variables in a `.env` file: ``` GOOGLE_CLOUD_PROJECT=your-project-id # Optional: Pre-configured example store (after running ingestion) EXAMPLE_STORE=your-example-store-resource-name ``` ## Running the Application ### Step 1: Prepare Few-Shot Examples (Optional) To improve the agent's performance, you can ingest few-shot examples: ```bash python injest_examples.py --project your-project-id ``` After running, you'll get an example store resource name that you should set in your `.env` file. ### Step 2: Start the Agent Run the main agent script: ```bash python agent.py ``` This will initialize the agent and allow you to interact with it. The script includes a sample question to demonstrate functionality. ## Implementation Details ### Key Components - **agent.py**: Main entry point and agent orchestration - **sql_query_mcp_server.py**: MCP server that provides SQL tools - **sqlite_database.py**: Database wrapper for the Chinook SQLite database - **injest_examples.py**: Tool for ingesting few-shot examples ### Architecture 1. The user sends a natural language query 2. The main agent processes the query and calls the SQL tool 3. The SQL tool uses the MCP server to: - List tables and get schema information - Generate an appropriate SQL query - Execute the query against the database 4. The main agent interprets the results and responds in natural language ## Customization ### Using a Different Database To use a different SQLite database: 1. Update the `setup_database` function in `sqlite_database.py` 2. Update the few-shot examples in `injest_examples.py` to match your schema ### Modifying Agent Behavior You can adjust the agent's behavior by modifying the instructions in: - `agent.py`: Change the main agent's instruction prompt - `sql_query_mcp_server.py`: Modify the SQL generation agent's prompt