Bio Agents MCP

🚀 Biological Agent MCP Project

A collection of microservices and clients for interacting with biological databases through natural language.

✨ Features

Components

• LLM Client: A natural language interface with a web interface and terminal mode.
• PDB MCP Server: An API service for the Protein Data Bank.
• ChEMBL MCP Server: An API service for the chemical database.

Architecture

┌─────────────┐     ┌──────────────┐
│   LLM Interface │     │ Ollama Large Language Model │
│ (Chainlit)    │     │              │
└─────┬───────┘     └───────┬──────┘
      │                     │
┌─────┴─────────────────────┴──────┐
│           LLM Client             │
└─────┬─────────────────────┬──────┘
      │                     │
┌─────┴───────┐     ┌──────┴───────┐
│  PDB MCP    │     │ ChEMBL MCP  │
│   Server    │     │    Server    │
└─────────────┘     └──────────────┘

🚀 Quick Start

1. Configure the environment:

cp .env.example .env

2. Start the services:

make build
make up

3. Start the web interface:

make run-chainlit

Visit http://localhost:8000 to start querying biological data.

📚 Documentation

Development

• Use make help to view available commands.
• Each service has its own README file containing detailed documentation.
• Configuration files are located in the conf/ directory.

Explanation

This project consists of multiple modules that use the FastMCP framework to interact with various services and APIs. Each module can be run independently or together using Docker Compose. The main focus of this project is the biological agent, which provides tools and services for interacting with biological data sources such as the Protein Data Bank (PDB).

Modules

LLM Client

The llm-client module provides a client for interacting with the large language model server, processing queries, and utilizing available tools. It is built using the FastMCP framework and supports asynchronous operations (using aiohttp).

For more information, see LLM Client README.

Protein Data Bank

The protein_data_bank_mcp module provides a server for interacting with the Protein Data Bank (PDB) API to obtain structure assembly descriptions, chemical components, drugbank annotations, branched entities, non-polymer entities, polymer entities, uniprot annotations, structures, pubmed annotations, pdb clustering data aggregations, clustering group sources, pdb clustering data aggregation methods, and pairwise multimer interface descriptions. It is built using the FastMCP framework and supports asynchronous operations (using aiohttp).

For more information, see Protein Data Bank README.

📦 Installation

Docker

A Dockerfile is provided for each module to build Docker images.

• Build a Docker image:

  docker build -t <module name> .
  

• Run a Docker container:

  docker run --env-file .env <module name>
  

Docker Compose

A docker-compose.yml file is provided to run all services together.

• Start all services:

  docker-compose up -d
  

• Stop all services:

  docker-compose down
  

Makefile

A Makefile is provided to simplify common tasks.

• Available targets:
  • setup-env: Set up the initial environment.
  • build: Build all Docker images.
  • up: Start all services.
  • down: Stop all services.