MemoWiki

A persistent memory layer for your codebase - Automated documentation generation with LLM-powered insights, semantic search, and agent-driven summaries.

Overview

MemoWiki transforms your code into searchable, up-to-date documentation automatically. It leverages Large Language Models (LLMs) to generate comprehensive docs, diagrams, and summaries while intelligently caching results to minimize API usage.

Key Features

• ✅ Automated Documentation - Generate docs from code with LLM analysis
• ✅ Visual Diagrams - Mermaid class and architecture diagrams
• ✅ Semantic Search - Natural language code search with vector embeddings
• ✅ Agent Summaries - Track AI agent implementation progress
• ✅ Smart Caching - Avoid redundant API calls
• ✅ Git Integration - Auto-detect changes and track history
• ✅ Multi-LLM Support - OpenAI, Anthropic, Gemini, OpenRouter, Ollama
• ✅ REST API - Expose functionality for IDE/CI integrations
• ✅ Web Dashboard - Visual interface for monitoring and management

Quick Start

Installation

npm install -g memowiki

Configuration

Create a .env file:

LLM_PROVIDER=openai
OPENAI_API_KEY=your-api-key-here
OPENAI_MODEL=gpt-4-turbo-preview

See Configuration Guide for all providers.

Basic Usage

# Generate documentation for your codebase
memowiki update

# Record an agent implementation summary
memowiki record --type feature --output my-feature

# Query summaries
memowiki summaries --recent 5

# Search code semantically
memowiki search "authentication logic"

# Analyze repository
memowiki analyze

# Start API server (with integrated dashboard)
memowiki serve

# Build dashboard and backend
npm run build:all

Web Dashboard

MemoWiki now includes a web dashboard for visualizing and managing your codebase documentation.

Features

• Overview - Stats cards showing documented files, cache hit rate, and pending TODOs
• Documentation Browser - Browse and filter documentation status by file
• Summaries Timeline - View agent implementation summaries with filtering
• Settings - Manage LLM provider configuration
• Activity Feed - Track recent documentation updates

Quick Start

# Build both backend and dashboard
npm run build:all

# Start the server (serves both API and dashboard)
memowiki serve

# Access dashboard at http://localhost:3000

Development Mode

# Terminal 1: Start backend API
memowiki serve

# Terminal 2: Start dashboard with hot-reload
cd dashboard
ng serve
# Access at http://localhost:4200

Build Scripts

| Script | Description | |--------|-------------| | npm run build | Build backend only | | npm run build:dashboard | Build dashboard only | | npm run build:all | Build both backend and dashboard |

Commands

Core Commands

| Command | Description | Documentation | |---------|-------------|---------------| | update | Generate documentation for codebase | Guide | | record | Create agent implementation summaries | Guide | | summaries | Query agent summaries | Guide | | search | Semantic code search | Guide | | analyze | Git repository insights | Guide | | serve | Start REST API server + dashboard | Guide | | intent | Analyze code change intent | Guide |

Command Details

Update - Generate Documentation

memowiki update [--full] [--verbose]

• Auto-detects changed files from git
• Generates markdown documentation
• Creates class and architecture diagrams
• Intelligently caches results

📖 Full Documentation

Record - Agent Summaries

memowiki record [-f files...] [-o name] [-t type] [--append]

• Track AI agent implementation progress
• Identify TODOs, stubs, and completed work
• Create structured summaries with LLM analysis
• Support incremental documentation

📖 Full Documentation

Summaries - Query Agent Work

memowiki summaries [-f file] [-t type] [-r recent]

• Efficient index-based filtering
• Query by file, type, or recency
• No need to parse all summary files

📖 Full Documentation

Search - Semantic Code Search

memowiki search <query> [-n limit]

• Natural language queries
• Vector embedding similarity search
• Returns ranked results with snippets

📖 Full Documentation

Analyze - Repository Insights

memowiki analyze

• View git repository status
• See recent commits and branch info
• Detect merge conflicts

📖 Full Documentation

Serve - API Server

memowiki serve [-p port]

• REST API for external integrations
• IDE plugins and CI/CD pipelines
• Programmatic access to all features

📖 Full Documentation

Intent - Change Classification

memowiki intent

• Analyze code change intent
• Classify as bug fix, feature, refactor, etc.
• Provide actionable suggestions

📖 Full Documentation

Documentation

Guides

• Configuration - LLM provider setup
• Dashboard - Web interface guide
• Update Command - Documentation generation
• Record Command - Agent summaries
• Summaries Command - Query summaries
• Search Command - Semantic search
• Analyze Command - Repository analysis
• Serve Command - API server
• Intent Command - Intent classification

Supported LLM Providers

| Provider | Setup | Best For | |----------|-------|----------| | OpenAI | Requires API key | Production quality | | Anthropic | Requires API key | Long context | | Gemini | Requires API key | Cost-effective | | OpenRouter | Requires API key | Model flexibility | | Ollama | Local install | Privacy, no limits | | Mock | No setup | Testing |

See Configuration Guide for detailed setup.

Output Structure

.codewiki/
├── memory/           # Individual file documentation
├── diagrams/         # Class diagrams
├── flows/            # Architecture diagrams
├── summaries/        # Project and agent summaries
│   ├── index.json   # Summary metadata index
│   └── *.md         # Summary files
└── cache.json        # Documentation cache

Use Cases

For Developers

• Onboarding - New team members understand codebase quickly
• Documentation - Automated, always up-to-date
• Code Review - Context for pull requests
• Architecture - Visual system diagrams

For AI Agents

• Memory - Persistent knowledge of codebase
• Context - Understand system architecture
• Progress Tracking - Document implementation work
• TODO Management - Track incomplete work

For Teams

• Knowledge Sharing - Centralized code knowledge
• Change Tracking - Git-integrated documentation
• Search - Find code by natural language
• API Access - Integrate with existing tools

Programmatic Usage

import {
  LLMService,
  WikiManager,
  SummaryGenerator,
  SearchService,
  GitService
} from 'memowiki';

// Generate documentation
const llmService = new LLMService(config);
const doc = await llmService.generateDocumentation(codeStructure, gitState);

// Create agent summary
const generator = new SummaryGenerator(llmService, config);
const summary = await generator.generateImplementationSummary(
  ['src/auth/login.ts'],
  'feature'
);

// Search codebase
const searchService = new SearchService();
await searchService.initialize();
const results = await searchService.search('authentication', 10);

// Analyze git
const gitService = new GitService();
const gitState = await gitService.getGitState();

Integration Examples

VS Code Extension

// Generate docs on save
vscode.workspace.onDidSaveTextDocument(async (document) => {
  const response = await fetch('http://localhost:3000/api/v1/documentation', {
    method: 'POST',
    body: JSON.stringify({ file: document.fileName })
  });
});

GitHub Actions

- name: Generate Documentation
  run: |
    memowiki update
    git add .codewiki
    git commit -m "docs: update documentation"
    git push

Git Hook

#!/bin/bash
# .git/hooks/post-commit
memowiki update
memowiki record --type feature

Performance

• Caching - Avoids redundant API calls (10-100x faster on unchanged files)
• Incremental Updates - Only processes changed files
• Parallel Processing - Concurrent LLM requests
• Smart Filtering - Respects .gitignore and .memowikiignore

Requirements

• Node.js 16+
• Git repository (optional but recommended)
• LLM provider API key (or Ollama for local)
• ChromaDB (optional, for semantic search)

Troubleshooting

# Clear cache
rm -rf .codewiki/cache.json

# Regenerate all docs
memowiki update --full

# Enable verbose logging
memowiki update --verbose

# Test configuration
memowiki analyze

Contributing

Contributions welcome! See CONTRIBUTING.md for guidelines.

License

MIT License - see LICENSE for details.

Support

• 📖 Documentation
• 🐛 Issues
• 💬 Discussions
• 📧 Email

Built with ❤️ for developers and AI agents