Skip to content

redhat-ai-dev/rhess

Repository files navigation

Red Hat Enterprise Skills Server (RHESS)

Developer Preview — not yet ready for production use.

A self-hosted, OCI-containerized skills directory and server that lets enterprise teams curate, discover, and distribute AI agent skills internally — behind corporate firewalls — without relying on the public skills.sh registry.

RHESS is fully compatible with the npx skills CLI and the Agent Skills specification, so skills authored for RHESS work across 40+ supported AI coding agents (Claude Code, Cursor, GitHub Copilot, Codex, Windsurf, and more) without modification.

Overview

  • REST API — paginated skill listing, detail retrieval, and fuzzy search (unauthenticated)
  • Web directory — React + PatternFly UI for browsing, searching, and viewing skill details
  • Git-based ingestion — register any Git repository as a skill source; RHESS automatically clones, discovers, and indexes all SKILL.md files
  • Admin token auth — write operations (source registration, sync, deletion) require a single admin token; all reads are open
  • Agent Skills spec compliant — serves /.well-known/agent-skills/index.json for npx skills CLI discovery (v0.2.0)
  • OCI container — runs on OpenShift, Kubernetes, or locally via Podman/Docker

Quickstart (local)

podman run \
  -e RHESS_ADMIN_TOKEN=your-secret-token \
  -e DATABASE_PATH=/data/rhess.db \
  -p 3000:3000 \
  ghcr.io/redhat-ai-dev/rhess:latest

Then open http://localhost:3000 to browse the skills directory.

Quickstart (Kubernetes / OpenShift)

kubectl apply -f deploy/

See docs/developer-preview.md for known limitations and single-replica constraints.

Configuration

Environment Variable Required Description
RHESS_ADMIN_TOKEN Yes Bearer token for write operations. Server exits on startup if missing or empty.
DATABASE_PATH No Path to SQLite database file (default: ./rhess.db). Must be on a writable PVC in Kubernetes.
PORT No HTTP port (default: 3000).

Architecture

Single-process Node.js/TypeScript server using Fastify, serving both the REST API and the React/PatternFly web UI. SQLite via better-sqlite3 for the Developer Preview (PostgreSQL planned post-DP). Skills are ingested from registered Git repositories using simple-git with shallow clones.

See openspec/changes/rhess-enterprise-skills-server/design.md for full architectural decisions.

Tech Stack

  • Backend: Node.js, TypeScript, Fastify, SQLite (better-sqlite3), simple-git, Fuse.js
  • Frontend: React, PatternFly, Vite, react-markdown
  • Container: UBI9/nodejs-24, OpenShift Route / Kubernetes manifests

Contributing

This project is in Developer Preview. Contributions are welcome via pull requests against main.

# Fork and clone
git clone https://github.com/YOUR_USERNAME/rhess.git
cd rhess
git remote add upstream https://github.com/redhat-ai-dev/rhess.git

# Install dependencies (once package.json exists)
npm install

# Start dev server
npm run dev

License

Apache License 2.0 — see LICENSE.

About

Red Hat Enterprise Skills Server (RHESS) — self-hosted AI agent skills directory for enterprise teams

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages