crystalyse-ai 1.0.3

CrystaLyse.AI - Autonomous AI agents for accelerated inorganic materials design through natural language interfaces

CrystaLyse.AI

Status: Research Preview - Fully Functional Materials Design Platform

CrystaLyse.AI is a computational materials design platform that accelerates materials exploration through AI-powered analysis and validation. Built on the OpenAI Agents framework with Model Context Protocol (MCP) integration, it provides a dual-mode system for rapid materials design workflows.

💡 For AI Assistants & Copilots: If you're using GitHub Copilot, Claude, ChatGPT, or other AI coding assistants, check out READMEFORLLMS.md - a comprehensive guide designed specifically for AI agents to quickly understand this project's capabilities, installation, and usage patterns. Simply copy-paste its contents into your AI assistant for instant project context!

Key Features

✅ Dual-Mode Analysis System

• Creative Mode: Fast exploration (~50 seconds) using Chemeleon + MACE
• Rigorous Mode: Complete validation (2-5 minutes) with SMACT + Chemeleon + MACE + Analysis Suite
• Real-time mode switching with unified interface

✅ Complete Materials Pipeline

• Composition Validation: SMACT screening for chemically reasonable materials
• Structure Prediction: Chemeleon crystal structure generation with multiple candidates
• Energy Calculations: MACE formation energy evaluation with uncertainty quantification
• Comprehensive Analysis: XRD patterns, RDF analysis, coordination studies
• 3D Visualisation: Interactive molecular viewers and professional analysis plots

✅ Advanced Interface Options

• Unified CLI: Single command interface with /mode and /agent switching
• Session Management: Persistent conversation history across multi-day projects
• Interactive Chat: Research-grade session-based workflows
• Batch Processing: High-throughput materials screening capabilities

Architecture Overview

High-Level System Architecture

graph TB
    subgraph "User Interface Layer"
        CLI[Enhanced CLI<br/>Rich Console + Themes]
        UNIFIED[Unified Interface<br/>Single Entry Point]
        LEGACY[Legacy CLI Commands<br/>Backward Compatibility]
    end
    
    subgraph "Agent Layer"
        SESSION_AGENT[Session-Based Agent<br/>OpenAI Agents SDK]
        LEGACY_AGENT[Legacy Agent<br/>Direct MCP Integration]
        VALIDATOR[Response Validator<br/>Anti-Hallucination]
    end
    
    subgraph "Memory System"
        SESSION_MEM[Session Memory<br/>SQLite Conversations]
        USER_MEM[User Memory<br/>Preferences & Research]
        DISCOVERY[Discovery Cache<br/>Computational Results]
        CROSS_SESSION[Cross-Session Context<br/>Research Continuity]
    end
    
    subgraph "MCP Server Architecture"
        CREATIVE[Chemistry Creative Server<br/>Chemeleon + MACE + Viz]
        UNIFIED_SERVER[Chemistry Unified Server<br/>SMACT + Chemeleon + MACE + Analysis]
        VIZ[Visualisation Server<br/>3Dmol.js + Pymatviz]
    end
    
    subgraph "Tool Dependencies (oldmcpservers)"
        SMACT[SMACT MCP<br/>Composition Validation]
        CHEMELEON[Chemeleon MCP<br/>Structure Prediction]
        MACE[MACE MCP<br/>Energy Calculation]
    end
    
    subgraph "Infrastructure"
        POOL[MCP Connection Pool<br/>Persistent Connections]
        SESSION_MGR[Session Manager<br/>Lifecycle Management]
        RESILIENT[Resilient Tool Caller<br/>Fault Tolerance]
    end
    
    CLI --> SESSION_AGENT
    UNIFIED --> SESSION_AGENT
    LEGACY --> LEGACY_AGENT
    
    SESSION_AGENT --> VALIDATOR
    SESSION_AGENT <--> SESSION_MEM
    SESSION_AGENT <--> USER_MEM
    SESSION_AGENT <--> DISCOVERY
    SESSION_AGENT <--> CROSS_SESSION
    
    SESSION_AGENT --> SESSION_MGR
    LEGACY_AGENT --> POOL
    SESSION_MGR --> POOL
    POOL --> RESILIENT
    
    RESILIENT --> CREATIVE
    RESILIENT --> UNIFIED_SERVER
    RESILIENT --> VIZ
    
    UNIFIED_SERVER --> SMACT
    UNIFIED_SERVER --> CHEMELEON
    UNIFIED_SERVER --> MACE
    CREATIVE --> CHEMELEON
    CREATIVE --> MACE

Dual-Mode Operation Flow

graph LR
    subgraph "Entry Points"
        USER_QUERY[User Query]
        MODE_SELECT{Mode Selection<br/>/mode creative or rigorous}
    end
    
    subgraph "Creative Mode (~50s)"
        CREATIVE_SERVER[Chemistry Creative Server]
        FAST_STRUCT[Chemeleon Structure<br/>Generation]
        FAST_ENERGY[MACE Energy<br/>Calculation]
        BASIC_VIZ[Basic 3D<br/>Visualisation]
        CREATIVE_RESULT[Fast Results<br/>Structure + Energy]
    end
    
    subgraph "Rigorous Mode (2-5min)"
        UNIFIED_SERVER[Chemistry Unified Server]
        SMACT_VAL[SMACT Composition<br/>Validation]
        FULL_STRUCT[Chemeleon Complete<br/>Structure Generation]
        FULL_ENERGY[MACE Full Energy<br/>Analysis]
        COMPREHENSIVE[XRD + RDF +<br/>Coordination Analysis]
        RIGOROUS_RESULT[Complete Results<br/>Publication Ready]
    end
    
    subgraph "Memory Integration"
        CACHE[Discovery Cache<br/>Avoid Redundancy]
        SESSION[Session History<br/>Research Context]
        INSIGHTS[Cross-Session<br/>Learning]
    end
    
    USER_QUERY --> MODE_SELECT
    MODE_SELECT -->|Creative| CREATIVE_SERVER
    MODE_SELECT -->|Rigorous| UNIFIED_SERVER
    
    CREATIVE_SERVER --> FAST_STRUCT
    FAST_STRUCT --> FAST_ENERGY
    FAST_ENERGY --> BASIC_VIZ
    BASIC_VIZ --> CREATIVE_RESULT
    
    UNIFIED_SERVER --> SMACT_VAL
    SMACT_VAL --> FULL_STRUCT
    FULL_STRUCT --> FULL_ENERGY
    FULL_ENERGY --> COMPREHENSIVE
    COMPREHENSIVE --> RIGOROUS_RESULT
    
    CREATIVE_RESULT --> CACHE
    RIGOROUS_RESULT --> CACHE
    CACHE --> SESSION
    SESSION --> INSIGHTS

Session-Based Research Flow

graph TB
    subgraph "Session Lifecycle"
        START[Session Start<br/>crystalyse chat -s project]
        RESUME[Session Resume<br/>crystalyse resume project]
        CONTEXT[Context Loading<br/>Previous Conversations]
    end
    
    subgraph "Interactive Commands"
        MODE_SWITCH["/mode creative|rigorous"]
        AGENT_SWITCH["/agent chat|analyse"]
        HISTORY["/history - Show Past"]
        STATUS["/status - Current State"]
        THEME["/theme - Change UI"]
    end
    
    subgraph "Research Memory"
        CONV_HIST[Conversation History<br/>SQLite Storage]
        DISCOVERIES[Discovery Database<br/>Computational Results]
        USER_PROF[User Profiles<br/>Research Interests]
        WEEKLY[Weekly Summaries<br/>Research Progress]
    end
    
    subgraph "Multi-Day Continuity"
        PROJECT[Long-term Projects]
        COLLAB[Collaborative Sessions]
        INSIGHTS[Pattern Recognition]
        EXPORT[Research Export]
    end
    
    START --> CONTEXT
    RESUME --> CONTEXT
    CONTEXT --> MODE_SWITCH
    CONTEXT --> AGENT_SWITCH
    
    MODE_SWITCH --> CONV_HIST
    AGENT_SWITCH --> CONV_HIST
    HISTORY --> CONV_HIST
    STATUS --> USER_PROF
    
    CONV_HIST --> DISCOVERIES
    DISCOVERIES --> WEEKLY
    WEEKLY --> PROJECT
    PROJECT --> COLLAB
    COLLAB --> INSIGHTS
    INSIGHTS --> EXPORT

Repository Structure

CrystaLyse.AI/
├── crystalyse/                    # Core package
│   ├── agents/                    # AI agent implementations
│   ├── infrastructure/            # Connection pooling, session management
│   ├── memory/                    # Memory system and caching
│   ├── output/                    # Dual formatter and visualisation
│   ├── validation/                # Anti-hallucination system
│   └── cli.py                     # Unified command-line interface
├── chemistry-unified-server/      # Rigorous mode (SMACT + Chemeleon + MACE)
├── chemistry-creative-server/     # Creative mode (Chemeleon + MACE)
├── visualization-mcp-server/      # 3D structures and analysis plots
├── oldmcpservers/                 # Individual tool servers (required dependencies)
│   ├── smact-mcp-server/          # SMACT composition validation
│   ├── chemeleon-mcp-server/      # Chemeleon structure prediction
│   └── mace-mcp-server/           # MACE energy calculations
├── docs/                          # Comprehensive documentation
│   ├── guides/                    # Installation and usage guides
│   ├── concepts/                  # Analysis modes and architecture
│   ├── tools/                     # Individual tool documentation
│   └── reference/                 # Complete API reference

Quick Start

Installation

# Clone repository  
git clone https://github.com/ryannduma/CrystaLyse.AI.git
cd CrystaLyse.AI

# Create conda environment
conda create -n crystalyse python=3.11
conda activate crystalyse

# Install core package
pip install -e .

# Install individual tool servers (required dependencies)
pip install -e ./oldmcpservers/smact-mcp-server
pip install -e ./oldmcpservers/chemeleon-mcp-server
pip install -e ./oldmcpservers/mace-mcp-server

# Install unified MCP servers
pip install -e ./chemistry-unified-server
pip install -e ./chemistry-creative-server
pip install -e ./visualization-mcp-server

Configuration

# Set OpenAI API key
export OPENAI_API_KEY="sk-your-key-here"

# Verify installation
crystalyse --help
crystalyse config show

Usage Examples

Quick Analysis

# Creative mode (fast exploration)
crystalyse analyse "Find stable perovskite materials for solar cells" --mode creative

# Rigorous mode (complete validation)
crystalyse analyse "Analyse CsSnI3 for photovoltaic applications" --mode rigorous

Interactive Research Sessions

# Start a research session
crystalyse chat -u researcher -s solar_project -m creative

# Resume previous work
crystalyse resume solar_project -u researcher

# List all sessions
crystalyse sessions -u researcher

Unified Interface

# Launch interactive interface with mode switching
crystalyse

# In-session commands:
# /mode creative     - Switch to creative mode
# /mode rigorous     - Switch to rigorous mode
# /agent chat        - Switch to chat agent
# /agent analyse     - Switch to analysis agent
# /help              - Show available commands
# /exit              - Exit interface

Example Output

Creative Mode Results:

╭─────────────────────── Discovery Results ────────────────────────────╮
│ Generated 5 perovskite candidates with formation energies:            │
│                                                                        │
│ 1. CsGeI₃ - Formation energy: -2.558 eV/atom (most stable)           │
│ 2. CsPbI₃ - Formation energy: -2.542 eV/atom                         │
│ 3. CsSnI₃ - Formation energy: -2.529 eV/atom                         │
│ 4. RbPbI₃ - Formation energy: -2.503 eV/atom                         │
│ 5. RbSnI₃ - Formation energy: -2.488 eV/atom                         │
│                                                                        │
│ 3D visualisations created: CsGeI3_3dmol.html, CsPbI3_3dmol.html      │
╰────────────────────────────────────────────────────────────────────────╯

Rigorous Mode Output:

• Complete SMACT composition validation
• Multiple structure candidates per composition
• Professional analysis plots (XRD, RDF, coordination analysis)
• 3D interactive visualisations
• Publication-ready results

Applications

Energy Materials

• Battery cathodes and anodes (Li-ion, Na-ion, solid-state)
• Solid electrolytes and ion conductors
• Photovoltaic semiconductors and perovskites
• Thermoelectric materials

Electronic Materials

• Ferroelectric and multiferroic materials
• Magnetic materials and spintronics
• Semiconductor devices and memory materials
• Superconductors and quantum materials

Research Workflows

• High-throughput materials screening
• Structure-property relationship studies
• Materials optimisation and design
• Experimental validation planning

Performance Characteristics

Operation	Creative Mode	Rigorous Mode
Simple query	~50 seconds	2-3 minutes
Complex analysis	1-2 minutes	3-5 minutes
Batch processing	5-10 minutes	15-30 minutes

System Requirements:

• Python 3.11+
• 8GB RAM minimum (16GB recommended)
• OpenAI API key
• Optional: NVIDIA GPU for MACE acceleration

Documentation

Comprehensive documentation is available in the docs/ directory:

• Quickstart Guide - Get started in minutes
• Installation Guide - Detailed setup instructions
• CLI Usage Guide - Complete command reference
• Analysis Modes - Creative vs Rigorous workflows
• Tool Documentation - SMACT, Chemeleon, MACE, Visualisation
• API Reference - Complete API documentation

Scientific Integrity

CrystaLyse.AI maintains computational honesty:

• 100% Traceability: Every result traces to actual tool calculations
• Zero Fabrication: No estimated or hallucinated numerical values
• Complete Transparency: Clear distinction between AI reasoning and computational validation
• Anti-Hallucination System: Response validation prevents fabricated results

Acknowledgments

CrystaLyse.AI builds upon exceptional open-source tools:

• SMACT: Semiconducting Materials by Analogy and Chemical Theory
• Chemeleon: Crystal structure prediction with AI
• MACE: Machine learning ACE force fields
• Pymatviz: Materials visualisation toolkit
• OpenAI Agents SDK: Production-ready agent framework

Citation

If you use CrystaLyse.AI in your research, please cite the underlying tools:

• SMACT: Davies et al., "SMACT: Semiconducting Materials by Analogy and Chemical Theory" JOSS 4, 1361 (2019)
• Chemeleon: Park et al., "Exploration of crystal chemical space using text-guided generative artificial intelligence" Nature Communications (2025)
• MACE: Batatia et al., "MACE: Higher Order Equivariant Message Passing Neural Networks for Fast and Accurate Force Fields" NeurIPS (2022)
• Pymatviz: Riebesell et al., "Pymatviz: visualization toolkit for materials informatics" (2022)

License

MIT License - see LICENSE for details.