Experimental lab to test runtime skill matching and inspect the private structured JSON returned by the agent against a real published HEA.
lab-onlyreal published HEA groundingno live product path changes
This lab keeps the visible conversation natural while always evaluating next-help skills from skillsContent.json. The panel below shows the private structured JSON that the agent returns turn by turn. It does not write leads, call Shopify, or alter the live widget path.
Select a published HEA to start.
Agent Structured JSON History
private state returned by the agent across turns
No structured agent JSON yet.
Last Runtime Trace
matched next-help skills + top grounding matches + parse status
No trace yet.
Last Compiled Prompt
exact OpenAI message array used for the last turn
No compiled prompt yet.
Conversation
Client-managed history, real HEA grounding, runtime skill matching, SSE streaming response.
creator: (unknown)hea: (none)
Start a conversation to see the HEA answer with knowledge plus runtime skill matching. The left column will accumulate the private structured JSON returned by the agent on each turn.