Multi-surface AI agent workflow platform — a product of HodeTech.
Relavium lets you design multi-agent, multi-model LLM workflows as git-committable
YAML and run them across a Tauri desktop app, a VS Code extension, and a CLI —
local-first in Phase 1, with cloud execution and a control-plane portal in
Phase 2. The workflow engine is a pure-TypeScript package shared by every surface,
and multi-provider LLM access goes through Relavium's own @relavium/llm
abstraction over the official provider SDKs (no Vercel, no LangChain).
The canonical documentation lives in docs/ — start at
docs/README.md, which is organized by the kind of question each
section answers.
| Start here | |
|---|---|
| Vision · Product constraints · UVP | What and why |
| Tech stack · Project structure | What it's built with |
| Architecture · Decisions (ADRs) · Reference | How it works |
| Roadmap · Standards | Where it's going, and the rules |
Pre-implementation. The documentation, architecture decisions, and phased roadmap are complete; engineering begins at Phase 0 — Foundations.