a2a_with_mcp

# Multi-Agent Market Research & Financial Analysis System This project demonstrates a sophisticated multi-agent system designed for market research and financial analysis. It combines the power of large language models (LLMs), specialized agents, and external data sources via the Model Context Protocol (MCP). ## Overview The system features an **Orchestrator Agent** that receives user requests and delegates subtasks to specialized agents. The key specialized agent is the **Financial Analysis Agent**, which has been enhanced to connect to multiple data sources. ## Architecture 1. **User Interface (UI):** (Assumed) Interacts with the Orchestrator Agent. 2. **Orchestrator Agent (`orchestrator_agent`):** * Built with Google ADK. * Receives user requests. * Uses an LLM (Gemini Flash) to determine the required subtasks. * Delegates tasks to specialized agents via A2A calls using specific tools. * Synthesizes results from specialized agents into a final response. 3. **Specialized Agents:** * **Financial Analysis Agent (`financial_agent_langgraph`):** * Built with LangGraph. * Connects to a **Postgres MCP server** for internal database queries (current prices, company info). * Connects to the **`mcp-crypto-price` MCP server** (Node.js) for external crypto market/historical analysis. * Exposes its capabilities via an A2A-compliant FastAPI server. * **Sentiment Analysis Agent (`sentiment_analysis_agent`):** * Built with CrewAI (in this example). * Connects to an **MCP Reddit server** to fetch posts. * Analyzes sentiment related to specific companies/cryptocurrencies. * Exposes its capabilities via an A2A-compliant FastAPI server. * **Other Agents (Competitor Analysis, Visualization, Templates):** * Placeholder agents demonstrating potential extensions (currently return mock data in the orchestrator). 4. **MCP Servers:** * **Postgres MCP Server:** (Assumed `MCP-servers/postgres_mcp.py`) Provides tools to interact with a financial PostgreSQL database. * **Crypto Price MCP Server:** (`mcp-crypto-price` Node.js package) Provides tools (`get-market-analysis`, `get-historical-analysis`) using the CoinCap API. * **Reddit MCP Server:** (Located within `sentiment_analysis_agent`) Provides tools to fetch Reddit data. ## Key Features & Concepts - **Multi-Agent Collaboration:** Demonstrates how an orchestrator can manage and leverage multiple specialized agents. - **A2A Protocol:** Uses a standardized protocol for inter-agent communication (JSON-RPC based). - **Model Context Protocol (MCP):** Enables agents (like the Financial Agent) to securely and reliably use external tools and data sources (Postgres DB, CoinCap API via `mcp-crypto-price`, Reddit API) served via MCP. - **Multi-Backend MCP:** The Financial Agent showcases connecting to multiple, different MCP servers simultaneously. - **LangGraph & CrewAI:** Utilizes different frameworks for building the core logic of specialized agents. ## Running the System 1. **Prerequisites:** Python 3.10+, Node.js/npx, PostgreSQL database, API Keys (OpenAI, CoinCap, potentially Reddit/PRAW). 2. **Setup:** * Clone the repository. * Install Python dependencies (`pip install -r requirements.txt` - ensure a consolidated requirements file exists or install per-agent). * Set required environment variables (see individual agent READMEs). * Ensure MCP server scripts/packages are accessible (`MCP-servers/postgres_mcp.py`, `mcp-crypto-price` via npx). 3. **Launch Agents (in separate terminals):** * `python -m financial_agent_langgraph` (Handles its MCP servers internally) * `uvicorn sentiment_analysis_agent.agent_server:app --reload --port 10000` (Or however the sentiment agent is run) * `python -m orchestrator_agent` * (Launch other specialized agents if implemented). 4. **Interact:** Use the UI or send requests directly to the Orchestrator Agent (default port 8000). ## Project Structure - `orchestrator_agent/`: Code for the central orchestrator. - `financial_agent_langgraph/`: Code for the LangGraph-based financial agent. - `sentiment_analysis_agent/`: Code for the CrewAI-based sentiment agent. - `MCP-servers/`: (Example location) Contains MCP server implementations (e.g., `postgres_mcp.py`). - `README.md`: This file. - (Potentially `requirements.txt`, `.env.example`, etc.) *(See individual agent directories for their specific README files)* ## Requirements - Python 3.9 or higher - PostgreSQL database server - OpenAI API key (for GPT models) ## Installation 1. Clone this repository ```bash git clone <repository-url> cd <repository-directory> ``` 2. Install dependencies ```bash pip install -r requirements.txt ``` 3. Set up your PostgreSQL database: ```sql -- Create the companies table CREATE TABLE companies ( ticker TEXT PRIMARY KEY, name TEXT NOT NULL, sector TEXT, industry TEXT ); -- Create the stocks table CREATE TABLE stocks ( id SERIAL PRIMARY KEY, symbol TEXT NOT NULL REFERENCES companies(ticker), price NUMERIC(10, 2), volume INTEGER, timestamp TIMESTAMP DEFAULT CURRENT_TIMESTAMP ); -- Create the financial_metrics table CREATE TABLE financial_metrics ( id SERIAL PRIMARY KEY, ticker TEXT NOT NULL REFERENCES companies(ticker), metric TEXT NOT NULL, value NUMERIC(15, 2), period TEXT ); -- Create the financial_snapshot table CREATE TABLE financial_snapshot ( id SERIAL PRIMARY KEY, data JSONB, created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP ); -- Insert sample data as needed ``` 4. Create a `.env` file with your OpenAI API key: ``` OPENAI_API_KEY=your_api_key_here ``` ## Configuration 1. Update the PostgreSQL connection settings in `MCP-servers/postgres_mcp.py`: ```python db_url = "postgresql://postgres:password@localhost:5432/testdb" ``` Replace with your database credentials. ## Usage ### 1. Start the MCP Postgres Server ```bash python3 MCP-servers/postgres_mcp.py ``` ### 2. Run the Financial Agent You can run the agent in several ways: #### Testing Direct Queries ```bash python test_financial_agent.py ``` #### Testing MCP Connection ```bash python test_mcp_connection.py ``` #### Testing Async MCP Tool Invocation ```bash python test_async_mcp.py ``` ### 3. Example Queries You can ask the agent questions like: - "What's the current stock price for AAPL?" - "Show me financial metrics for Microsoft" - "Compare the P/E ratios of tech companies" - "What are the latest market trends from the financial snapshot?" ## MCP Tools Available The agent can access the following MCP tools: 1. **query**: Run arbitrary SQL queries against the database ``` Example: SELECT * FROM stocks WHERE symbol = 'AAPL' ``` 2. **fetch_financial_snapshot**: Get the latest financial snapshot data as a JSON object ``` This returns a comprehensive market overview from the financial_snapshot table ``` ## Advanced Configuration ### Event Loop Management The agent implements a shared event loop system for all MCP operations to prevent "Event loop is closed" errors during async operations. This is handled in the `get_mcp_event_loop()` function. ### Error Handling The system includes robust error handling with fallback mechanisms for: - Database connection issues - MCP server communication failures - Async event loop errors ## Troubleshooting 1. **Event Loop Errors**: If you experience "Event loop is closed" errors, restart the MCP server and the agent. 2. **Database Connection Issues**: Ensure your PostgreSQL server is running and the credentials in `postgres_mcp.py` are correct. 3. **Tool Execution Failures**: Check the logs in `financial_agent.log` for detailed error messages. ## License [Your License Information] ## Acknowledgments - LangChain and LangGraph for the agent framework - The Model Context Protocol (MCP) for standardizing tool access ## Environment Variables The system requires various API keys and configuration settings to be provided as environment variables. Create a `.env` file in the root directory with the following variables: ``` # OpenAI API for financial agent OPENAI_API_KEY=your_openai_api_key_here # Google API for sentiment and orchestrator agents GOOGLE_API_KEY=your_google_api_key_here # CoinCap API for cryptocurrency data COINCAP_API_KEY=your_coincap_api_key_here # Database configuration for financial agent DB_HOST=localhost DB_PORT=5432 DB_USER=postgres DB_PASSWORD=your_password_here DB_NAME=financial_db # Agent ports (optional, defaults shown) SENTIMENT_AGENT_PORT=10000 FINANCIAL_AGENT_PORT=10001 ORCHESTRATOR_PORT=8000 ``` Each agent has specific environment variable requirements: 1. **Financial Agent**: Requires `OPENAI_API_KEY`, database configuration, and `COINCAP_API_KEY` 2. **Sentiment Agent**: Requires `GOOGLE_API_KEY` 3. **Orchestrator Agent**: Requires `GOOGLE_API_KEY` 4. **Visualization Agent**: Requires `GOOGLE_API_KEY` Environment variables can also be set directly in your shell if you prefer not to use a `.env` file.