← Back to Discover

WFGY

An open-source collection of protocols, text packs and debugging guides for people working with AI assistants, agents and retrieval pipelines, including a Problem Map for broken workflows.

Observability & evals

What it does

WFGY describes itself as an open-source ecosystem for engineered language, human-machine interaction and reusable AI execution components. Its current flagship is the WFGY 5.0 Polaris Protocol, which is being released in stages rather than all at once. The first usable piece is Polaris Goal Compiler, a portable protocol you download and paste into an AI assistant, an agent rule file or a workflow policy. It aims to make the assistant settle the goal before building, show the active task and what is blocked, and avoid claiming completion too early. For workflows that are already failing, the site points to Problem Map 3.0, which bundles a troubleshooting atlas, a router, a Global Debug Card and a Global Fix Map for broken RAG or agent pipelines. Earlier generations stay available, including a text pack for WFGY 3.0 that you upload to a strong LLM and start by typing run, then go. The project publishes evidence packages with raw outputs, verdicts and checksums, and runs a Discord for updates. Deeper engine and runtime components are planned for later releases.

Capabilities

  • Polaris Goal Compiler, a portable protocol to paste into an assistant or agent rules
  • Problem Map 3.0 with a troubleshooting atlas, router and Global Debug Card
  • Downloadable text pack for WFGY 3.0 that you run inside an LLM
  • Public evidence packages with raw outputs and checksums
  • Staged releases, with deeper engine components planned later

When to use it

  • A RAG or agent workflow is failing and you want a structured route to debug it
  • You want a portable ruleset that makes an AI assistant clarify the goal before building
  • You want to try open-source protocol ideas with your own LLM