The Self-Evolving AI Developer
Typical coding assistants reset every session. Autogenesis remembers, learns, and grows with you.
Autogenesis is an autonomous software engineer that lives in your browser.
It doesn't just autocomplete code; it plans, architects, builds, tests, and deploys full-stack applications from a single prompt.
But the real magic is in its growth mechanics:
- Memory: It remembers every project you've built.
- XP System: Every feature you implement earns XP.
- Skill Tree: As it levels up, it unlocks new capabilities (e.g., Python Lv1 -> Python Lv5).
- RLHF Training: Rate its code (1–5 stars) to train a custom model adapter via Oumi.
- Export: Export the entire code-base as .zip file.
| Feature | Description | Tech Provider |
|---|---|---|
| ⚡ Instant Builds | Generates full-stack apps in seconds using LPUs. | Groq |
| 🧠 Deep Logic | Handles complex architecture & reasoning. | Gemini 2.0 |
| 💾 Long-Term Memory | Tracks project history & evolution. | Kestra |
| 🎯 Self-Correction | Auto-fixes bugs & dependency errors. | CodeRabbit |
| 🚀 Instant Deploy | One-click public hosting. | Vercel |
| 🎓 Adaptive Learning | Learns from your feedback ratings. | Oumi |
Watch code turn into reality instantly with full-screen previews.
Watch your agent evolve from "Baby" to "Sage" as you build more app.
Clean, well-structured code with automated error checking.
- Node.js 18+
- Python 3.9+
cd backend
python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
cp .env.example .env # Add your GROQ_API_KEY
python -m uvicorn api:app --reloadcd frontend
npm install
npm run devVisit http://localhost:3000 and start building!
Autogenesis uses a multi-agent system to simulate a real dev team:
- The Architect (Gemini): Plans the project structure.
- The Coder (Groq): Writes high-speed code logic.
- The Reviewer (CodeRabbit): Checks for bugs and security.
- The DevOps (Vercel): Handles deployment pipelines.
Autogenesis is built on a hybrid multi-model architecture that combines ultra-fast inference, high-context reasoning, orchestration, RL training, and automated CI/CD.
| Technology | Purpose |
|---|---|
| Groq LLaMA‑3.1 / 3.3 70B | Primary coding engine for multi-file generation, architecture, and debugging. |
| Gemini 2.0 Flash | High-context planner and fallback reasoning model for complex workflows. |
| Oumi (RLHF + DPO) | Learns from user ratings and optimizes code generation preferences. |
| CodeRabbit | Automated PR reviews, security scanning, and code quality enforcement. |
| Cline AI CLI | Autonomous command execution, dev automation, and simulated agent workflow. |
| Technology | Purpose |
|---|---|
| FastAPI | Backend API for orchestrating agents and tools. |
| Uvicorn | ASGI server enabling low-latency streaming responses. |
| Server-Sent Events (SSE) | Real-time event, log, and file streaming to the UI. |
| Python 3.9+ | Primary backend language and orchestration logic. |
| Pydantic | Request/response validation and type-safe schemas. |
| Docker | Containerization for reproducible backend deployments. |
| python-dotenv | Environment variable and secret management. |
| Module | Functionality |
|---|---|
| Architect Agent | Plans folder structures, system architecture, and task breakdown. |
| Coder Agent | Generates code file-by-file across frontend and backend. |
| Reviewer Agent | Detects issues, suggests fixes, and enforces quality gates. |
| Memory Engine | Stores project history, XP, and skill progression. |
| Intelligence Engine | Implements levels, XP, and skill evolution logic. |
| Oumi RL Engine | Produces and consumes feedback data for preference optimization. |
| Technology | Purpose |
|---|---|
| Next.js 14/15 | React meta-framework for the Autogenesis web app. |
| TypeScript | Type-safe frontend development and API contracts. |
| Tailwind CSS | Utility-first styling for rapid UI iteration. |
| Framer Motion | Animations and micro-interactions for the dashboard and previews. |
| Web Speech API | Voice-to-text prompts for hands-free tasking. |
| React Hooks + Context | Client-side state and session management. |
| Technology | Role |
|---|---|
| Vercel | One-click deployment for the frontend experience. |
| GitHub Actions | Automated CI/CD pipelines and checks on every push/PR. |
| Kestra | Long-term memory summarization, workflows, and scheduling. |
| Dockerfile Generator | Generates production-ready Docker builds for services. |
| Tool | Purpose |
|---|---|
| pnpm / npm | Frontend dependency and script management. |
| ESLint + Prettier | Optional linting and auto-formatting of the codebase. |
| Vitest / Pytest | Auto-generated unit testing for frontend and backend. |
| GitHub Codespaces | Cloud-based development environment support. |
| Technology | Purpose |
|---|---|
| JSON Memory Store | Persistent agent memory for sessions, XP, and history. |
| Kestra Buckets | Workflow state, logs, and artifact storage. |
| In-memory caching | Latency reduction for hot paths and repeated queries. |
Built for the 2025 AI Hackathon.
- Groq for speed.
- Gemini for intelligence.
- Kestra for orchestration.
- Oumi for training.
Built with 🫶 by Nikhil Jones A & SH Nihil Mukkesh
© 2025 Autogenesis Team. Open Source. Licensed under the MIT License.[3]