@principal/core

Core library for Principal AI - GitHub repository parsing and LLM services for the hackathon.

Features

• GitHub Repository Loading: Fetch and cache repository contents via GitHub API
• MemoryPalace Integration: Access codebase views, notes, and guidance from .alexandria/ directories
• Groq LLM Integration: Ultra-fast LLM inference for intelligent responses

Setup

1. Install dependencies:

bun install

2. Set your GitHub token (for API access):

export GITHUB_TOKEN=your_github_token_here

Demo

Run the demo to see MemoryPalace loading a GitHub repository:

bun run src/demo.ts

Architecture

Adapters

• GitHubFileSystemAdapter: Pre-fetches repository files and implements MemoryPalace's FileSystemAdapter interface
• GitHubGlobAdapter: Fetches file trees from GitHub and implements glob pattern matching

How It Works

1. Pre-fetch: Downloads .alexandria/ directory structure from GitHub
2. Cache: Stores files in memory to support synchronous file operations
3. MemoryPalace: Uses cached files to read views, notes, and guidance
4. LLM Context: Sends structured context to Groq for intelligent responses

Next Steps

• Integrate with Groq LLM for Q&A
• Connect to voice interface (Agora/ElevenLabs/HeyGen)
• Deploy to Appwrite