L4ntern0 / oh-my-tang

What is oh-my-tang?

oh-my-tang is a TypeScript plugin for OpenCode that structures AI tasks as a governed workflow inspired by the Tang Dynasty's Three Departments and Six Ministries: draft plans, review them, dispatch to specialized agents (ministries like personnel, revenue, rites, military, justice, works), execute via runtime sessions or local fallback, and audit outcomes. Developers get tools like tang_process to kick off requests, tang_audit for execution provenance, tang_pipeline for flow snapshots, and tang_doctor for health checks—all persisting state for cross-session inspection. It tackles the opacity of multi-agent AI by enforcing review gates, retries, and token budgeting.

Why is it gaining traction?

In a sea of experimental GitHub projects like proton experimental github or ue4ss experimental github, oh-my-tang stands out with its intuitive bureaucracy metaphor that maps to real audit needs, plus operator-facing views for anomalies, hotspots, and diagnostics without digging into raw payloads. The hybrid runtime-local execution ensures workflows never fully fail, and config via .oh-my-tang.json plus env vars like TANG_HEALTH_RISK_PROFILE keeps it tweakable. Developers dig the structured visibility over ad-hoc agent chains.

Who should use this?

OpenCode users orchestrating complex tasks like code generation, testing pipelines, or multi-agent coordination who need built-in reviews and audit trails. Ideal for AI workflow experimenters frustrated by untraceable failures, or teams in experimental design github setups wanting fallback-heavy reliability without losing provenance.

Verdict

With 13 stars and 1.0% credibility score, this experimental GitHub gem is a vibe-coded prototype—fun for audit-focused tinkering, backed by multilingual docs and fixture tests, but skip for production until more battle-testing. Try it if Tang-style governance clicks for your OpenCode flows.

(198 words)