PerformanceSuite/DevAssist

DevAssist MCP Server v2.2.0 🚀

An intelligent development assistant that provides runtime session management, architectural memory, and AI-powered code intelligence through the Model Context Protocol (MCP).

🌟 Overview

DevAssist transforms your development workflow by maintaining context across sessions, tracking decisions, preventing duplicate work, and providing intelligent assistance throughout your coding journey.

Key Benefits

• 40-60% Token Reduction: Through advanced tool masking
• Persistent Memory: Never lose context between sessions
• Intelligent Search: Vector-based semantic code search
• Automated Organization: Smart file cleanup and organization
• Session Intelligence: Warmup, monitoring, and cleanup

🚀 Quick Start

Installation

# Clone and install
git clone https://github.com/PerformanceSuite/DevAssist.git
cd DevAssist
npm install

# Initialize databases
npm run db:init

# Start the server
npm start

Claude Desktop Configuration

Add to ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "devassist": {
      "command": "node",
      "args": ["/path/to/DevAssist/index.js"]
    }
  }
}

📦 Features

🎨 UI Module (NEW!)

Enhanced visual development environment with:

• Live Preview: Split-screen code + browser preview with Playwright
• Visual Regression Testing: Pixel-perfect comparison and diff generation
• Design System Validation: Enforce design tokens and spacing scales
• Performance Tracking: Real-time metrics and budget enforcement
• Keyboard Shortcuts: Cmd+R refresh, Cmd+S save, Cmd+V validate
• Visual Checkpoints: Git integration with automatic screenshots

UI Mode Commands

# Enter UI Focus Mode
/ui-mode --component Button --viewport mobile

# Create visual checkpoint
/ui-checkpoint "Updated button styles"

# Exit UI mode with summary
/ui-mode --exit

Session Management

Complete lifecycle management for development sessions:

• Session Start: Warmup with context loading
• Heartbeat Monitoring: 5-minute interval checks
• Terminal Logging: Capture command history
• Session End: Intelligent cleanup and organization

Architectural Memory

Record and retrieve development decisions:

// Record a decision
devassist:record_architectural_decision {
  decision: "Use PostgreSQL for main database",
  context: "Need ACID compliance and complex queries",
  alternatives: ["MongoDB", "MySQL"],
  impact: "Requires schema migrations"
}

// Retrieve decisions
devassist:get_project_memory {
  query: "database decisions",
  category: "architecture"
}

Semantic Search

Powerful vector-based code search:

• Technology: LanceDB + all-mpnet-base-v2 embeddings
• Performance: ~50ms for 1000+ records
• Accuracy: 768-dimensional embeddings for precision

Progress Tracking

Monitor development milestones:

devassist:track_progress {
  milestone: "Authentication system complete",
  status: "completed",
  notes: "JWT implementation with refresh tokens"
}

Intelligent Cleanup

End-of-session file organization:

• Archives temporary files
• Detects and handles duplicates
• Organizes documentation
• Cleans build artifacts
• Updates .gitignore

🏗️ Architecture

Directory Structure

DevAssist/
├── index.js                    # Main MCP server
├── src/
│   ├── database/              # Data layer
│   │   ├── dataAccess.js     # Unified data access
│   │   ├── init.js           # Database initialization
│   │   └── migrate.js        # Migration scripts
│   ├── commands/              # MCP command handlers
│   │   ├── initproject.js    # Project initialization
│   │   └── ...               # Other commands
│   ├── session/               # Session management
│   │   ├── warmup.js         # Session warmup
│   │   ├── terminal-logger.js # Terminal capture
│   │   └── cleanup-manager.js # Cleanup logic
│   └── documentation/         # Documentation tools
│       └── tech-docs-fetcher.js
├── data/                      # Persistent storage
│   ├── vectors/              # LanceDB vectors
│   └── *.json                # Legacy JSON storage
└── masks/                     # Tool masking configs

Database Architecture

Hybrid Approach: SQLite + LanceDB

• SQLite: Structured data (projects, decisions, progress)
• LanceDB: Vector embeddings for semantic search
• Migration: Automatic migration from JSON to database

Embedding Models

// Available models
const EMBEDDING_MODELS = {
  'mpnet': {
    name: 'Xenova/all-mpnet-base-v2',
    dimensions: 768,
    description: 'Higher quality, better semantic understanding'
  },
  'minilm': {
    name: 'Xenova/all-MiniLM-L6-v2',
    dimensions: 384,
    description: 'Fast, lightweight, good for basic matching'
  }
};

🛠️ Available Tools

Core Tools

Session Lifecycle

graph LR
    A[Session Start] --> B[Warmup]
    B --> C[Active Development]
    C --> D[Heartbeat Monitoring]
    D --> C
    C --> E[Session End]
    E --> F[Cleanup]
    F --> G[Report Generation]

🧪 Testing

Run Tests

npm test

Test Coverage

• Database operations ✅
• Embedding generation ✅
• Semantic search ✅
• Session management ✅
• Cleanup operations ✅

Performance Benchmarks

npm run benchmark

Results:

• Embedding generation: ~160ms (first run with model loading)
• Semantic search: ~50ms for 1000 records
• Database operations: < 10ms

📊 Session Intelligence

Warmup Process

When starting a session, DevAssist:

1. Loads previous session context
2. Analyzes recent git changes
3. Checks terminal logs (last 50 commands)
4. Prepares search indices
5. Loads relevant documentation
6. Identifies pending TODOs
7. Starts heartbeat monitor
8. Activates terminal logger

Cleanup Manager

Intelligent file organization at session end:

// Cleanup rules
const CLEANUP_RULES = {
  archive: ['*.tmp', '*.log', '*.bak'],
  organize: {
    'docs/': ['*.md', '!README.md'],
    'tests/': ['*.test.js', '*.spec.js'],
    'archive/': ['old_*', 'backup_*']
  },
  ignore: ['.git', 'node_modules', '.devassist']
};

Terminal Logger

Captures development context:

• Command history
• Error patterns
• Workflow analysis
• Productivity metrics

🔧 Configuration

Environment Variables

Create .env file:

# Database
DB_PATH=./data/db
VECTOR_DB_PATH=./data/vectors

# Embeddings
EMBEDDING_MODEL=mpnet

# Session
HEARTBEAT_INTERVAL=300000
TERMINAL_LOG_SIZE=1000

# Cleanup
ARCHIVE_RETENTION_DAYS=30

Project Configuration

Each project gets a .devassist/config.json:

{
  "project": "my-app",
  "type": "node",
  "features": ["docker", "testing", "ci"],
  "settings": {
    "sessionWarmup": true,
    "heartbeatInterval": 300000,
    "terminalLogging": true,
    "intelligentCleanup": true
  }
}

🔄 Migration

From JSON to Database

# Automatic migration
npm run db:migrate

# Test migration without changes
npm run db:migrate test

🤝 Integration

With Prjctzr

DevAssist integrates seamlessly with Prjctzr:

1. Prjctzr creates project structure
2. Sets up DevAssist configuration
3. DevAssist handles runtime operations

With Claude Code

Custom slash commands available:

• /start-session - Begin with goals
• /end-session - End with cleanup
• /track-progress - Record achievements

📈 Roadmap

Coming Soon

• Real-time collaboration features
• Advanced analytics dashboard
• Plugin system for extensions
• Cloud sync for team sharing
• AI-powered code review

Future Enhancements

• Multi-model embedding support
• Distributed vector search
• Advanced cleanup strategies
• Session replay capabilities
• Team knowledge sharing

🐛 Troubleshooting

Common Issues

Database initialization fails

# Reset databases
rm -rf data/vectors data/db.sqlite
npm run db:init

Embedding model not loading

# Clear model cache
rm -rf node_modules/@xenova
npm install

Session not starting

• Check .devassist/ directory exists
• Verify project configuration
• Run with debug: DEBUG=* npm start

📄 License

MIT License - see LICENSE file

🔗 Links

• GitHub Repository
• MCP Documentation
• Report Issues

👥 Contributors

Built with ❤️ by the Performance Suite team.

DevAssist v2.2.0 - Intelligent Development Assistant with UI Module Part of the Custom MCP Servers Suite