helixml / helix

SaaS • Private Deployment • Docs • Discord HelixML - AI Agents on a Private GenAI Stack 👥 Discord **Deploy AI agents in your own data center or VPC and retain complete data security & control.** HelixML is an enterprise-grade platform for building and deploying AI agents with support for RAG (Retrieval-Augmented Generation), API calling, vision, and multi-provider LLM support. Build and deploy LLM applications by writing a simple configuration file. Our intelligent GPU scheduler packs models efficiently into available GPU memory and dynamically loads and unloads models based on demand, optimizing resource utilization. ✨ Key Features 🤖 AI Agents • **Easy-to-use Web UI** for agent interaction and management • **Session-based architecture** with pause/resume capabilities • **Multi-step reasoning** with tool orchestration • **Memory management** for context-aware interactions • **Support for multiple LLM providers** (OpenAI, Anthropic, and local models) 🛠️ Skills and Tools • **REST API integration** with OpenAPI schema support • **MCP (Model Context Protocol) server** compatibility • **GPTScript integration** for advanced scripting • **OAuth token management** for secure third-party access • **Custom tool development** with flexible SDK 📚 Knowledge Management • **Built-in document ingestion** (PDFs, Word, text files) • **Web scraper** for automatic content extraction • **Multiple RAG backends**: Typesense, Haystack, PGVector, LlamaIndex • **Vector embeddings** with PGVector for semantic search • **Vision RAG support** for multimodal content **Main use cases:** • Upload and analyze corporate documents • Add website documentation URLs to create instant customer support agents • Build knowledge bases from multiple sources 🔍 Tracing and Observability Context is everything. Agents can process tens of thousands of tokens per step—Helix provides complete visibility under the hood: **Tracing features:** • View all agent execution steps • Inspect requests and responses to LLM providers, third-party APIs, and MCP servers • Real-time token usage tracking • Pricing and cost analysis • Performance metrics and debugging 🚀 Additional Features • **Multi-tenancy** with organization, team, and role-based access control • **Scheduled tasks** and cron jobs • **Webhook triggers** for event-driven workflows • **Evaluation framework** for testing and quality assurance • **Payment integration** with Stripe support • **Notifications** via Slack, Discord, and email • **Keycloak authentication** with OAuth and OIDC support 🏗️ Architecture HelixML uses a microservices architecture with the following components: **Three-layer agent hierarchy:** • **Session**: Manages agent lifecycle and state • **Agent**: Coordinates skills and handles LLM interactions • **Skills**: Group related tools for specific capabilities • **Tools**: Individual actions (API calls, functions, scripts) 💻 Tech Stack Backend • **Go 1.24.0** - Main backend language • **PostgreSQL + PGVector** - Data storage and vector embeddings • **GORM** - ORM for database operations • **Gorilla Mux** - HTTP routing • **Keycloak** - Identity and access management • **NATS** - Message queue • **Zerolog** - Structured logging Frontend • **React 18.3.1** - UI framework • **TypeScript** - Type-safe JavaScript • **Material-UI (MUI)** - Component library • **MobX** - State management • **Vite** - Build tool • **Monaco Editor** - Code editing AI/ML • **OpenAI SDK** - GPT models integration • **Anthropic SDK** - Claude models integration • **LangChain Go** - LLM orchestration • **GPTScript** - Scripting capabilities • **Typesense / Haystack / LlamaIndex** - RAG backends Infrastructure • **Docker & Docker Compose** - Containerization • **Kubernetes + Helm** - Orchestration • **Flux** - GitOps operator 🚀 Quick Start Install on Docker Use our quickstart installer: The installer will prompt you before making changes to your system. By default, the dashboard will be available on . For setting up a deployment with a DNS name, see or read the detailed docs. We've documented easy TLS termination for you. **Next steps:** • Attach your own GPU runners per runners docs • Use any external OpenAI-compatible LLM Install on Kubernetes Use our Helm charts for production deployments: • Control Plane Helm Chart • Runner Helm Chart 🔧 Configuration All server configuration is done via environment variables. You can find the complete list of configuration options in . **Key environment variables:** • - OpenAI API credentials • - Anthropic API credentials • - Database connection settings • - Authentication settings • - Public URL for the deployment • - GPU runner configuration See the configuration documentation for detailed setup instructions. 👨‍💻 Development For local development, refer to the Helix local development guide. **Prerequisites:** • Docker Desktop (or Docker + Docker Compose) • Go 1.24.0+ • Node.js 18+ • Make **Quick development setup:** See for comprehensive setup instructions. 📖 Documentation • **Overview** - Platform introduction • **Getting Started** - Build your first agent • **Control Plane Deployment** - Production deployment guide • **Runner Deployment** - GPU runner setup • **Agent Architecture** - Technical specification • **API Reference** - REST API documentation • **Contributing Guide** - How to contribute • **Upgrading Guide** - Migration instructions 🤝 Contributing We welcome contributions! Please see our Contributing Guide for details. By contributing, you confirm that: • Your changes will fall under the same license • Your changes will be owned by HelixML, Inc. 📄 License Helix is licensed under a similar license to Docker Desktop. You can run the source code (in this repo) for free for: • **Personal Use:** Individuals or people personally experimenting • **Educational Use:** Schools and universities • **Small Business Use:** Companies with under $10M annual revenue and less than 250 employees If you fal…