Developer Documentation

Everything you need to build with ÆtherLight Lumina. From quick start to advanced patterns.

Help Build the Future of Coding

We're Open Source (MIT License) • Join our beta, report bugs, suggest features, and contribute code. Shape the future of voice-powered development.

Try Beta

Quick Start

Get up and running in under 5 minutes

1

Install Extension

code --install-extension aetherlight.lumina

Or search "ÆtherLight Lumina" in VS Code Extensions marketplace

2

Start Coding

Press backtick (`) to activate voice:

Speak your command
Select target (Claude/Cursor)
Voice transcribed server-side

Unified Sprint Management System

1. Generate Sprint Plan

Voice command: "Analyze this codebase and generate a sprint plan"

Creates ACTIVE_SPRINT.toml with phases, tasks, dependencies

2. Select Active Sprint

Dropdown lets you choose from multiple sprints

Switch between ACTIVE_SPRINT, BACKLOG, KEY_AUTHORIZATION, etc.

3. Start Next Task

Click "Start Next Task" button

Generates AI-enhanced living prompt for any terminal

AI-Enhanced Living Task Prompts

When you click "Start Next Task", Lumina generates a context-aware prompt that:

  • Analyzes time elapsed since the task was written (e.g., "7 days elapsed")
  • Reviews all commits that happened between task creation and execution
  • Updates context automatically to reflect codebase changes
  • Reduces errors by ensuring AI has current project state
  • Delivers to any terminal selected above the text area (Claude Code, Cursor, etc.)

Technical Architecture

ÆtherLight is designed as a 3-layer system that progressively scales from local-first to global mesh networking.

Layer 1: Local Database

Q1 2026

Ultra-fast local storage with semantic search capabilities.

  • SQLite: Lightweight relational database for device state
  • ChromaDB: Vector embeddings for semantic pattern search
  • PostgreSQL: Optional cloud sync for team collaboration
  • Performance: <1ms local lookups

Layer 2: Mesh Network

Q1 2026

Peer-to-peer synchronization for local area networks.

  • mDNS Discovery: Automatic device discovery on LAN
  • WebRTC Sync: Direct peer-to-peer pattern sharing
  • Conflict Resolution: CRDT-based merge strategies
  • Performance: <10ms LAN lookups

Layer 3: DHT

Planned 2026

Distributed hash table for global knowledge sharing.

  • Kademlia DHT: Content-addressable pattern storage
  • IPFS Integration: Immutable pattern versioning
  • Censorship Resistance: No central authority required
  • Performance: <200ms global lookups

What Works Today

Voice capture in VS Code extension
Server-side transcription via OpenAI Whisper API (no user API key needed)
Token tracking in Supabase for usage limits
Sprint data in local TOML files (Git-friendly)
Patterns stored in GitHub repo (not database yet)
Context injection client-side before sending to Claude/Cursor

Join Our Open Beta

Help us build the future of voice-powered development

What's Included

  • Voice-powered coding with Whisper transcription
  • Sprint management with AI-generated task breakdown
  • Context injection for enhanced Claude/Cursor commands
  • Open source under MIT license (fork, modify, contribute)
  • Direct Discord access to dev team
  • Early access to new features

How to Contribute

  • Use it daily - Real-world testing in your workflow
  • Report bugs - Help us squash issues (GitHub Issues)
  • Share feedback - What works? What doesn't? What's missing?
  • Suggest features - Shape the roadmap with your ideas
  • Contribute code - VS Code extension, CLI tools, or APIs
  • Share patterns - Help others learn from your workflows
Join Beta → Start Contributing

Free during beta • Open source • No credit card required

Contributing

Help us build the future of AI-assisted development

Code Contributions

We're open source (MIT license). Fork, modify, submit PRs.

• VS Code Extension: TypeScript, React
• Backend: Rust (performance-critical), TypeScript
• Database Layer: SQLite, PostgreSQL, ChromaDB
• Embeddings: Python (ONNX Runtime, ChromaDB)

Pattern Contributions

Share your best AI prompts. Help the community ship faster.

• Opt-in only: You control what gets shared
• Anonymized: No personal info, no code snippets
• Credited: Your patterns show your username
• Rewarded: Top contributors get swag + bonuses
Contribute Patterns →

Need Help?

We're here for you. Multiple support channels for different needs.

Discord Community

Real-time chat with other developers

Join Discord →

GitHub Issues

Bug reports and feature requests

Open Issue →

Email Support

Direct support from our team

Email Us →