A Python notebook covering the three primitives, state shapes, confidence math, ten-question fan-out, composite scoring, typed function calls, async requests, retries, and a running token ledger.
Original by Asif RazzaqGetting startedMarkTechPost tutorialOriginal published Source reviewed
Before you dive in
What you’ll find in the original
Compare the same question over a bare message and structured state.
Recompute a Choice’s confidence from its probability distribution.
Keep policy weights, argument validation, and exact counting in code.
Worth knowing
The notebook requires an API key and prints live answers; the examples are a tutorial, not a benchmark on labeled data.