pi-typesafe
Pi extension with a Jev evaluation tool, terminal playground, and typed API for other extensions.
Pi extension with a Jev evaluation tool, terminal playground, and typed API for other extensions.
Pi provides `typesafe_evaluate`, which submits caller-provided JSON state and independent Choice, Score, and Noul questions in one request. The tool requires operator opt-in and returns structured answers to the agent.
Which team should handle this message?
Illustrative state: a customer says their subscription was charged twice and requests help today.
Set up a TypeSafe key, then enable the tool for the current Pi session or explicitly enable it for headless use.
Provide one JSON state and up to 32 typed questions; the extension validates and normalizes the request.
Send the state and independent questions through the TypeSafe client.
Render the returned Choice, Score, and Noul answers in the terminal without placing playground output in the agent context.
The extension puts typed, batched judgments inside an agent tool while keeping the data submission explicit and requiring session-level opt-in.