DocumentationOfficial
TypeSafe AI documentation
Start here if you are writing code. The docs are organised around the three question primitives, then around what to do with the confidence value: threshold it, route on it, or decompose a hard question into several easy ones. There is an llms.txt index if you would rather hand the whole thing to an agent.
- quickstart
- primitives
- patterns
- confidence
- From
- TypeSafe AI (US)
- Published
- 2026-09
- Listed here
- 2026-09-26
Sources checked
Also in this category
Documentation- DocumentationOfficial
Jev 1.13 known failure modes
A vendor page documenting where its own model is uneven, which is rarer than it should be. The examples matter in practice: the same question asked as a boolean and as a choice can disagree sharply, and a question and its negation, asked separately, do not have to sum to one. If you are about to put a threshold on a probability, read this first.
TypeSafe AI2026-09-17 - DocumentationOfficial
System One HTTP API reference
One endpoint, one POST, three top-level fields: the state to judge, the model, and a map of named questions. The response comes back keyed by the same names you sent, with token usage attached. Worth reading even if you use an SDK, because it is the shortest complete description of the protocol, including which status codes you are expected to retry with backoff.
TypeSafe AI2026-09 - DocumentationOfficial
TypeSafe cookbooks
End-to-end recipes grouped by the shape of the problem: self-consistency, batching many questions into one call, re-ranking a retrieval shortlist, recovering structure from flattened text, checking citations against their source, and screening input before it reaches a model that will act on it. The batching recipe is the one that most changes how you write the first call.
TypeSafe AI2026-09
