Lighthouse Agent Tooling

Lighthouse Agent Tooling – A set of connectors, IDE extensions, and AI-native developer tools to seamlessly integrate Lighthouse into AI-powered workflows. Supports MCP (Model Context Protocol) connectors, Cursor/IDE extensions for pinning and encrypting datasets/models.

Code Style

This repo uses ESLint (flat config) and Prettier for consistent code quality and formatting.

• Run lint: pnpm lint
• Auto-fix: pnpm lint:fix
• Format: pnpm format

Pre-commit hooks via Husky run lint-staged to format and lint changed files.

Prerequisites

• Node.js: v18+ (recommended: v22+)
• pnpm: v8+ (for workspace management)
• TypeScript: v5+ (installed automatically)

Package Structure

This monorepo contains the following packages:

• @lighthouse-tooling/types - Shared TypeScript types and interfaces ✅
• @lighthouse-tooling/shared - Shared utilities and common code
• @lighthouse-tooling/sdk-wrapper - Unified SDK wrapper for Lighthouse and Kavach
• @lighthouse-tooling/mcp-server - MCP (Model Context Protocol) server
• @lighthouse-tooling/vscode-extension - VSCode extension
• @lighthouse-tooling/cursor-extension - Cursor IDE extension

Build Process

Workspace Setup (pnpm)

# Install all dependencies for the workspace
pnpm install

# Build all packages
pnpm run build

# Type check all packages
pnpm run type-check

Individual Package Build

For individual package development, use npm within each package directory:

# Navigate to types package
cd packages/types

# Type checking
npm run type-check

# Build
npm run build

# Development mode (watch)
npm run dev

# Clean build artifacts
npm run clean

Note: TypeScript is installed at the workspace level via pnpm. Individual packages use npm for their build scripts.

Types Package Implementation

The @lighthouse-tooling/types package provides comprehensive TypeScript interfaces:

• Core interfaces: UploadResult, Dataset, AccessCondition, ProgressUpdate
• MCP types: MCPToolDefinition, MCPRequest, MCPResponse
• Authentication: AuthConfig, TokenInfo, APICredentials
• Error handling: LighthouseError, RetryConfig
• Workspace context: WorkspaceContext, ProjectFile, GitInfo

Usage

import { UploadResult, Dataset, AuthConfig, LIGHTHOUSE_MCP_TOOLS } from "@lighthouse-tooling/types";

// Create upload result
const uploadResult: UploadResult = {
  cid: "QmHash...",
  size: 1024,
  encrypted: true,
  uploadedAt: new Date(),
};

Development

# Start development mode (watch all packages)
pnpm run dev

# Lint all packages
pnpm run lint

# Run tests
pnpm run test

# Run tests with coverage
pnpm run test:coverage

# Clean all build artifacts
pnpm run clean

CI/CD

This project uses GitHub Actions for continuous integration and deployment.

Pipelines

Workflow	Trigger	Description
CI	Push/PR to main	Lint, build, type-check, and test
Release	Tag v* or manual	Build, test, package extensions, create GitHub release

CI Jobs

• Lint & Format: ESLint and Prettier checks
• Build: Compile all TypeScript packages
• Type Check: TypeScript type validation
• Test: Run all tests with coverage

Turbo Remote Caching (Optional)

Enable remote caching for faster CI builds:

1. Create a Vercel account (free tier works)
2. Link your repo: npx turbo link
3. Get your token from the Vercel dashboard
4. Add GitHub secrets:
   • TURBO_TOKEN: Your Vercel remote cache token
   • TURBO_TEAM (variable): Your Vercel team slug

Releasing

1. Update version in relevant package.json files
2. Create and push a tag: git tag v1.0.0 && git push origin v1.0.0
3. The release workflow will:
   • Run full CI validation
   • Package VSCode and Cursor extensions
   • Create a GitHub release with .vsix artifacts

VS Marketplace Publishing (Optional)

To auto-publish to VS Marketplace:

1. Get a Personal Access Token
2. Add VSCE_PAT secret to GitHub repository
3. Non-prerelease tags will auto-publish

License

MIT License - see LICENSE file for details.