Reasoning API
Kadai Inference
Get a grounded answer.
Kadai is the reasoning API. You query Kadai and get back an answer grounded in the evidence available to the system. Kadai does not act as the authority. It synthesizes across what Kura contains and what Ledger can support. The caller sees a single API. Behind it, the full bounded-synthesis pipeline runs — Compiler, Crosswalk, generation, Ledger, Boundary Gate — before anything is returned.
Without this: the caller talks directly to a model. The model answers from whatever it has — training data, context window, improvisation. There is no evidence boundary, no claim checking, and no output state. The burden of trust falls entirely on the person reading the response.
What Kadai does
Kadai is the single entry point for governed synthesis. A question enters Kadai. The Compiler rewrites it. The Crosswalk retrieves scoped evidence. The generation layer produces a proposal. The Ledger evaluates it. The Boundary Gate assigns a state. Kadai returns the result.
- One API call for the full bounded-synthesis pipeline
- Same contract across Workshop, Refinery, and Clean Room
- Returns response text + output state + claim-level audit trail
Who this is for
Application developers
integrate Kadai as the answer API in their workflows. One API call replaces direct model access and returns governed, classified responses.
The end user
receives a response that has already been evaluated and assigned an explicit state — not raw model text that still has to be defended by hand.