OHMind Documentation

A Language-Model Multi-Agent Framework for Accelerating Hydroxide Exchange Membrane (HEM) Discovery

Welcome to the OHMind documentation. This guide provides comprehensive information for researchers, developers, and administrators working with the OHMind platform for AI-driven cation design and HEM workflows.

Table of Contents

Overview

OHMind is developed by the PolyAI team from the Changchun Institute of Applied Chemistry, Chinese Academy of Sciences. It combines advanced machine learning models with domain-specific computational chemistry tools to accelerate the discovery of high-performance hydroxide exchange membranes.

The platform integrates:

  • Generative molecular design using Junction Tree Variational Autoencoders (JT-VAE)
  • Multi-objective optimization via Particle Swarm Optimization (PSO)
  • Quantum chemistry calculations through ORCA integration
  • Molecular dynamics simulations via GROMACS
  • Wavefunction analysis using Multiwfn
  • Literature retrieval through RAG-based search

System Components

OHMind consists of five main components:

Component Directory Description
Core Library OHMind/ VAE models, PSO optimization, QM/MD integration, reaction models, and HEM property prediction
Multi-Agent System OHMind_agent/ LangGraph-based multi-agent orchestration with 8 specialized agents
CLI Application OHMind_cli/ Textual-based TUI for interactive agent communication
Web UI OHMind_ui/ Chainlit + LangGraph frontend for browser-based access
MCP Servers OHMind_agent/MCP/ Five domain-specific servers (Chemistry, ORCA, Multiwfn, GROMACS, HEMDesign)

Architecture Highlights

  • LangGraph Workflow: Supervisor pattern with conditional routing and task planning
  • MCP Integration: Model Context Protocol for tool access via langchain-mcp-adapters
  • Persistent Sessions: Long-lived MCP connections for efficient tool execution
  • Task Planning: Automatic decomposition of complex queries into multi-step execution plans
  • RAG System: Qdrant-based vector search for HEM literature retrieval
  • Streaming Interface: Real-time token streaming with agent activity visualization

Documentation Chapters

Getting Started

Architecture

Agent Reference

MCP Server Reference

Core Library

CLI Application

Configuration

Tutorials

API Reference

Troubleshooting

Task Link
Install OHMind Quick Start Guide
Run HEM optimization HEM Optimization Tutorial
Configure MCP servers MCP Configuration
Troubleshoot issues Troubleshooting Overview
API integration API Overview

MCP Server Quick Reference

Server Tools Purpose
OHMind-Chem 17+ SMILES operations, functional groups, molecular properties
OHMind-HEMDesign 7 PSO optimization, backbone/cation management, job control
OHMind-ORCA 10+ QM calculations, geometry optimization, properties
OHMind-Multiwfn 16 Wavefunction analysis, orbital visualization, electron density
OHMind-GROMACS 25+ MD simulations, system preparation, trajectory analysis

Getting Help

  • Issues: Report bugs and feature requests on the project repository
  • Documentation: This documentation is continuously updated
  • Community: Join discussions with other OHMind users

*Last updated: 2025-12-23 OHMind v1.0.0*

PolyAI Team
Copyright © 2009-2025 Changchun Institute of Applied Chemistry, Chinese Academy of Sciences
Address: No. 5625, Renmin Street, Changchun, Jilin, China. Postal Code: 130022