8treenet/goraven

GoRaven – Self‑hosted AI Agent platform

What it does – GoRaven is a self‑hosted runtime that lets each team member work with an autonomous AI agent in its own “workspace”. The agent can read and write files, run shell commands, call APIs, query a RAG knowledge base, and produce concrete artefacts such as code, reports, or charts. The focus is on getting work done rather than just returning a chat response.

Key features

• Team‑oriented workspaces – isolated environments per user, shared projects, and admin‑controlled model quotas and tool permissions.
• Skill marketplace – reusable prompts, scripts and workflows (called skills) that can be installed with a click and maintained centrally.
• RAG‑enabled reasoning – policies, docs and other business material are indexed; agents retrieve relevant context at execution time, so answers are sourced rather than hallucinated.
• Plugin hooks – custom logic can be injected before/after conversations, tool calls or SSE streams, allowing deep integration without forking the core.
• Full self‑hosting – a single Docker command runs the whole stack (Go backend, React/TS frontend, Tailwind CSS, plus bundled runtimes like Python, Git, etc.). Data never leaves the server and any LLM‑compatible API (OpenAI, Claude, DeepSeek, Qwen, GLM, …) can be used.
• Observability‑first design – execution steps are exposed so users can see what the agent is doing, making debugging and iterative improvement easier.

How to try it

# Pull the official image
docker pull 8treenet/goraven:latest

# Run it (expose port 8000)
 docker run -d --restart=always --name goraven-agent \
   -p 8000:8000 8treenet/goraven:latest

Mount a host directory (/opt/goraven:/goraven/data) to keep workspaces persistent. After the container starts, open http://localhost:8000 and follow the wizard to create an admin account.

Architecture at a glance – The README includes a diagram showing a Go backend that orchestrates LLM calls, tool execution, and RAG retrieval, while a React + TypeScript frontend provides the UI. The system is packaged with Docker, making deployment straightforward.

Who it’s for – Teams that need an internal AI assistant capable of performing real tasks (code generation, data analysis, report writing, etc.) while keeping data on‑premise and retaining control over models, tools, and permissions.

License – Apache 2.0, allowing commercial use and modification.

---

All information above is taken directly from the repository’s README.