# Heym

> AI-native workflow automation platform. Build, visualize, and run intelligent AI workflows without writing code.

Heym is an open-source, self-hostable automation platform purpose-built for AI workflows. It features a visual drag-and-drop canvas, multi-agent orchestration, built-in RAG pipelines, MCP support, and a portable skills system.

## Core Capabilities

- **Visual Workflow Editor**: Drag-and-drop canvas powered by Vue Flow with 30+ node types across 7 categories (Triggers, AI, Logic, Data, Integrations, Automation, Utilities).
- **AI Assistant**: Describe what you want in natural language or voice — the assistant generates nodes and edges and applies them to the canvas instantly.
- **Multi-Agent Orchestration**: One agent orchestrates named sub-agents and sub-workflows, all wired visually. Supports configurable reasoning effort, tool calling, and automatic model fallback.
- **Built-in RAG Pipeline**: Upload PDFs, Markdown, CSV, or JSON to managed QDrant vector stores. Wire a RAG node into any workflow for semantic search.
- **MCP Support**: Connect to MCP servers (as a client) or expose workflows as an MCP server for Claude Desktop, Cursor, and other clients.
- **Skills System**: Portable capability bundles (SKILL.md + optional Python tools) that extend agent context and toolbox via drag-and-drop.
- **Human-in-the-Loop**: Pause execution for user approval before proceeding.
- **Guardrails**: Content filtering, NSFW protection, and safety checks at the node level.
- **Parallel DAG Execution**: Independent nodes run concurrently based on graph structure.

## Tech Stack

- **Frontend**: Vue.js 3, TypeScript, Vite + Bun, Vue Flow, Pinia, Tailwind CSS, Shadcn Vue
- **Backend**: Python 3.11+, FastAPI, UV package manager, Alembic migrations
- **Database & Infra**: PostgreSQL 16, SQLAlchemy 2.0, Redis, RabbitMQ, Docker, QDrant
- **AI & LLM**: OpenAI, Ollama (local), vLLM, Cohere, MCP protocol
- **Security**: JWT auth, bcrypt, Pydantic v2 validation, Fernet encryption

## Node Types

### Triggers
- Input (Webhook), Cron, RabbitMQ Receive, Error Handler

### AI
- LLM (OpenAI, Ollama, vLLM, etc.), AI Agent (tool-calling with reasoning), Qdrant RAG (semantic search)

### Logic
- Condition, Switch, Loop, Merge

### Data
- Set (variables/transform), Variable (store/retrieve), DataTable (structured data), Execute (sub-workflows)

### Integrations
- HTTP (REST/GraphQL), Slack, Send Email (SMTP), Redis, RabbitMQ Send, Grist

### Automation
- Crawler (FlareSolverr), Playwright (browser automation)

### Utilities
- Wait, Output, Console Log, Throw Error, Disable Node, Sticky Note

## Use Cases

- **Customer Support**: RAG-powered knowledge base, human-in-the-loop escalation, multi-language support, Slack/email integration.
- **Data Processing**: Web crawling with Playwright/FlareSolverr, PDF extraction, LLM-powered data transformation, Redis/PostgreSQL/vector store output.
- **Content Generation**: Multi-agent research and writing pipelines, quality checks with guardrails, scheduled generation via cron.
- **DevOps**: Webhook triggers for CI/CD, AI-powered log analysis, automated incident response with approval gates.
- **Sales & Marketing**: Lead enrichment with web crawling, LLM-generated personalized outreach, CRM integration via HTTP/webhooks.

## Getting Started

```bash
git clone https://github.com/heymrun/heym.git
cd heym && cp .env.example .env
./run.sh
Ready at http://localhost:4017
```

## Links

- Website: https://heym.run
- GitHub: https://github.com/heymrun/heym
- Discord: https://discord.gg/aAw8E3rje
- Enterprise: enterprise@heym.run

## License

Commons Clause + MIT. Free to use, modify, and distribute. Not for commercial resale.