Turn a rough app idea into a fully scoped engineering spec — locally, privately, and fast enough to do it before you lose the spark.
AuraForge is a native desktop app built on Tauri + React + Rust. It acts as your AI planning partner during the thinking phase: you describe what you want to build, and AuraForge guides you through scope, architecture, data model, and edge cases until it can emit a comprehensive spec document that AI coding tools like Claude Code can execute with minimal guesswork.
Everything runs locally via Ollama. No data leaves your machine.
- Structured planning conversations — One topic at a time with a planning-focused system prompt that prevents scope creep and vague requirements
- Codebase import — Point AuraForge at an existing project and it incorporates the current state into planning context
- Artifact generation — Emits spec documents, architecture plans, task breakdowns, and coverage/confidence reports
- Document linting — Reviews generated artifacts for quality, completeness, and consistency before you hand them off
- Diff tracking — Compares artifact versions across sessions to show exactly what changed in the plan
- Full local AI — Powered by Ollama; pulls and switches models from within the app
- Node.js 18+
- Rust stable toolchain (
rustup) - Ollama installed and running locally
- Tauri system dependencies: tauri.app/start/prerequisites
git clone https://github.com/saagpatel/AuraForge
cd AuraForge
npm install# Start in development mode
npm run dev:tauri
# Or lean dev mode (lower disk usage, slower restarts)
npm run dev:leanOn first launch, the onboarding wizard guides you through connecting Ollama and pulling a model.
| Layer | Technology |
|---|---|
| Desktop shell | Tauri 2 |
| Frontend | React 19, TypeScript, Tailwind CSS 4, Zustand |
| Backend | Rust — document generation, linting, artifact diffing, search |
| Local AI | Ollama (any compatible model) |
| Storage | SQLite via rusqlite, local app data dir |
| Testing | Vitest, Testing Library |
AuraForge uses a clean Rust backend with dedicated modules for each concern: docgen generates structured artifacts, lint validates them, artifact_diff tracks changes across sessions, importer ingests codebases, and llm manages streaming Ollama conversations. The React frontend streams responses in real time and stores session history in a local SQLite database.
MIT