Novion

A comprehensive medical research and analysis platform with specialized agent-based reasoning capabilities.

Project Overview

Novion is a medical research and analysis platform with a sophisticated multi-agent system using LangChain, enhanced with MCP (Model Context Protocol) tools and a flexible chat interface.

Key Features

Chat Interface

• Comprehensive chat interface in chat_interface.py
• Supports multiple LLM providers (OpenAI and Google)
• Streaming chat responses
• Direct tool execution via chat commands
• Specialized agent-based reasoning with chain-of-thought explanation

Specialized Medical Agents

Novion implements a team of specialized agents with chain-of-thought reasoning:

1. Pharmacist Agent: Expert in medication management, drug interactions, and pharmaceutical care
2. Researcher Agent: Specialist in clinical trials, research methodologies, and evidence-based medicine
3. Medical Analyst Agent: Focused on analyzing patient data, diagnostic information, and treatment outcomes

All agents show their detailed reasoning process within <think></think> tags before providing final recommendations.

MCP Tool Integration

• MCP installer in mcp/installer.py
• Flexible tool execution system
• Dynamic MCP server installation
• Tool discovery and help functionality

Server Endpoints

• /chat: Direct LLM interaction
• /chat/stream: Streaming chat responses
• /tools/execute: Direct MCP tool execution
• /tools: List available tools
• /mcp/install: Install new MCP servers

Getting Started

Prerequisites

• Python 3.8+
• Node.js (for MCP tools)
• npm/npx

Installation

1. Clone the repository
2. Install Python dependencies: pip install -r requirements.txt
3. Set up environment variables in .env.local

Running the Server

python backend/server.py

For a simplified demo without external dependencies:

python simple_server.py

Testing

Use the test_client.py script to test server functionality:

python test_client.py

Or run specific tests:

python test_client.py tools  # Test the tools endpoint
python test_client.py chat "Hello"  # Test basic chat
python test_client.py ask pharmacist "What are common side effects of ibuprofen?"  # Test specialized agents
python test_client.py tool list_fhir_resources  # Test a specific tool

Frontend

The system includes a simple web-based chat interface for interacting with the LLM and MCP tools.

• Access at: http://localhost:8000/
• Supports: Tool execution, agent consultation, streaming responses

Development

Adding New MCP Servers

Use the installer API or chat command:

/tool install_mcp_server server_name=package_name args=comma,separated,list env=KEY1=VAL1,KEY2=VAL2

Adding New Specialized Agents

Extend the agents in chat_interface.py by adding new system messages and agent types.

Project Structure

• backend/: Core server functionality
  • chat_interface.py: Main chat interface with LLM integration
  • server.py: FastAPI server with API endpoints
  • mcp/: MCP integration modules
    • installer.py: MCP server installation management
    • client.py: Novion MCP client implementation
    • fhir_server.py: FHIR MCP server implementation
• frontend/: User interface components
  • chat_ui.html: Web-based chat interface
• simple_server.py: Simplified demo server
• test_client.py: API testing client

License

Proprietary - All Rights Reserved