lyra
Health Warn
- License — License: NOASSERTION
- Description — Repository has a description
- Active repo — Last push 0 days ago
- Low visibility — Only 5 GitHub stars
Code Pass
- Code scan — Scanned 12 files during light audit, no dangerous patterns found
Permissions Pass
- Permissions — No dangerous permissions requested
No AI report is available for this listing yet.
Chat-driven AI agent that builds full-stack apps through an autonomous SDLC pipeline. 18 specialist agents, self-learning loop, any LLM provider.
Lyra
Describe it. Approve the design. Get a working app.
Lyra is an AI agent that builds full-stack applications through conversation. It runs an autonomous SDLC pipeline — from requirements gathering to deployment — using 18 specialist agents that coordinate, learn, and improve with every project.
Built on top of Hermes Agent — uses the hermes CLI under the hood.
What makes Lyra different
Most AI coding tools autocomplete lines or generate files. Lyra runs the entire software development lifecycle:
| What you do | What Lyra does |
|---|---|
| Describe your app in plain English | Interviews you for requirements, one question at a time |
| Approve a requirements summary | Generates a visual UI/UX preview for your approval |
| Say "looks good" | Architects, plans tasks, writes code, runs QA, deploys |
| Use your app | Learns from the session and gets better next time |
The SDLC Pipeline
Lyra doesn't just write code — it runs a full software development lifecycle with 18 specialist agents:
You ──► Requirements ──► Architecture ──► Task Planning ──► Development
Engineer Architect Planner Developer
│
Deployed ◄── DevOps ◄── Security ◄── Code Review ◄── QA ◄────┘
Engineer Auditor Reviewer Engineer
Each specialist has deep domain knowledge. The coordinator orchestrates the pipeline, verifies each phase's output, and advances automatically — stopping only at approval checkpoints where your input matters.
Self-Learning Loop
Lyra creates reusable skills from complex tasks and improves them during use. It remembers what worked across sessions, builds a model of your preferences, and gets faster and more accurate over time.
Any LLM, Any Platform
Use whatever model fits your budget and needs:
| Provider | Models |
|---|---|
| OpenRouter | Claude, GPT, Gemini, Llama, + hundreds more |
| Google AI Studio | Gemini 3 Flash/Pro |
| Ollama | Any local model (Llama, Qwen, Mistral, etc.) |
| Fireworks AI | Fast inference for open models |
| DeepInfra | 100+ open models, pay-per-use |
| HuggingFace | 20+ open models via unified endpoint |
| + 12 more | Kimi, GLM, MiniMax, Arcee, Upstage, NovitaAI... |
Switch providers anytime — no code changes, no lock-in.
Talk to Lyra from anywhere: CLI | Web Dashboard | Desktop App | Telegram | Discord | Slack | WhatsApp | Signal | Teams | Google Chat | Email
Quick Start
Prerequisites
- Python 3.11+
- Node.js 20+
- uv (Python package manager)
1. Clone and configure
git clone https://github.com/ub1979/lyra.git
cd lyra
cp .env.example .env
# Edit .env — add at least one LLM provider API key (e.g. OPENROUTER_API_KEY)
2. Start Lyra
./start.sh
That's it. On first run it installs dependencies, enables the builder plugin, and opens the dashboard at http://localhost:9119.
Click New Project, describe your app, and Lyra handles the rest.
Compare with the August 28 version
When the codex/aug28-ui-brain comparison worktree is installed on the same
computer, launch either version directly from this folder:
./start.sh # current Lyra — http://localhost:9119
./old_start.sh # August 28 + new UI + Project Brain — http://localhost:9120
Use two terminal windows to run both at the same time. Each version keeps its
generated projects in its own checkout. Press Ctrl+C in its terminal to
stop that version. To choose a different comparison port, runLYRA_OLD_PORT=9130 ./old_start.sh.
Using the CLI instead
uv sync # install dependencies (first time only)
uv run hermes setup # pick a provider and model
uv run hermes # start the CLI
> Build me a todo app with categories and due dates
Stopping
./stop.sh
Architecture
lyra/
├── agent/ # Core agent loop, tool execution, display
├── tools/ # Built-in tools (terminal, browser, files, skills, memory)
├── plugins/
│ ├── ultimate-builder/ # Guided app builder with dashboard UI
│ ├── platforms/ # Telegram, Discord, Slack, WhatsApp, etc.
│ ├── image_gen/ # AI image generation
│ └── ... # 20+ plugins
├── skills/ # Built-in skill library
├── optional-skills/ # Community skills (blockchain, creative, data-science, etc.)
├── web/ # React dashboard (Vite + TypeScript)
├── apps/desktop/ # Electron desktop app
└── ui-tui/ # Terminal UI (Ink)
Key Concepts
- Specialist Agents: 18 domain-specific agents (requirements, architecture, development, QA, security, DevOps, etc.) each with deep playbooks
- SDLC Coordinator: Orchestrates specialists in sequence, verifies phase outputs, stops at approval gates
- Guided Builder: Dashboard plugin that turns the SDLC pipeline into a chat-driven experience with visual previews
- Skill System: Agents create, store, and improve reusable skills from experience
- Terminal Backends: Run commands locally, in Docker, via SSH, on Modal (serverless), Singularity, or Daytona
- Memory: Agent-curated persistent memory with FTS5 search and cross-session recall
Run Anywhere
| Backend | Use Case |
|---|---|
| Local | Your machine, zero setup |
| Docker | Isolated container, reproducible |
| SSH | Remote server — agent can't read your .env |
| Modal | Serverless — hibernates when idle, near-zero cost |
| Singularity | HPC clusters |
| Daytona | Serverless dev environments |
Dashboard
The web dashboard provides a clean, progressive-disclosure interface:
- Guided Builder: Chat-driven app creation with approval gates
- Sessions: View and continue conversations
- Models: Configure providers and switch models
- Analytics: Token usage and cost tracking
- Plugins: Enable/disable capabilities
- Cron: Schedule autonomous tasks in natural language
Contributing
We welcome contributions! Please read our Contributing Guide before submitting a PR.
Quick version:
- Fork the repo
- Create a feature branch (
git checkout -b feat/my-feature) - Make your changes
- Run tests (
cd web && npm test) - Submit a PR — all PRs require review before merge
See CONTRIBUTING.md for the full process, coding standards, and what makes a good PR.
Community
- GitHub Discussions — questions, ideas, show & tell
- Issues — bug reports and feature requests
Acknowledgments
Lyra builds on the open-source Hermes Agent framework by Nous Research. Lyra adds the guided builder, progressive-disclosure dashboard, and enhanced SDLC workflow on top of that foundation.
License
Original material controlled by ub1979 is designated CC BY-NC-ND 4.0
for noncommercial use, with education and research as the intended audience.
Attribution is required; sharing adapted versions is not licensed.
See LICENSE and licensing scope and exceptions.
Inherited Hermes material remains MIT under LICENSE-MIT.upstream;
other third-party licenses and earlier grants remain in effect.
The restricted original material is source-available rather than OSI open source.
Reviews (0)
Sign in to leave a review.
Leave a reviewNo results found