# Tokz Labs

> Small, fast models for AI agents: the Decisions API and the open OpenDecision models.

- [Introduction](https://docs.tokz.si/introduction.md): Tokz Labs builds small, fast models for AI agents. One API key, one balance.
- [Quickstart](https://docs.tokz.si/quickstart.md): Get a key, install the SDK, ask your first question.
- [Decisions API](https://docs.tokz.si/decisions.md): Typed answers about text — routing, tagging, yes/no, scores — from the open OpenDecision models.
- [Fine Tune](https://docs.tokz.si/fine-tune.md): Train your own decision model on your text, then run it on our servers or on your machine.
- [TypeScript SDK](https://docs.tokz.si/sdk/typescript.md): @tokz/sdk — typed decisions about text over HTTP, and your fine tunes on your own machine.
- [Python SDK](https://docs.tokz.si/sdk/python.md): tokz for Python — typed decisions about text, sync and async, and your fine tunes on your own machine.
- [Billing](https://docs.tokz.si/billing.md): Prepaid credits, with low-balance alerts before you run out.
- [Troubleshooting](https://docs.tokz.si/troubleshooting.md): What comes up, and the fix for each.
- [Liveness and readiness](https://docs.tokz.si/api-reference/operations/liveness-and-readiness.md): Reports process liveness and database reachability. Unauthenticated.
- [Answer typed decisions about texts](https://docs.tokz.si/api-reference/decisions/answer-typed-decisions-about-texts.md): Asks one or more typed questions (choice, multilabel, bool, score) about up to 64 texts and returns, per text and question, the best option, the selected options, a probability distribution over the options, and for scores the expected value. This runs an OpenDecision model: the text is sent to a To…
- [Decision models and prices](https://docs.tokz.si/api-reference/decisions/decision-models-and-prices.md): The models behind /decide with their price per million billed tokens and the request limits. Unauthenticated and cacheable.
- [Decision usage for the authenticated key](https://docs.tokz.si/api-reference/decisions/decision-usage-for-the-authenticated-key.md): Counter totals, a per-day series and a per-model split for /decide. Counters only; no text, question or option is recorded.
- [List a fine tune's model files](https://docs.tokz.si/api-reference/fine-tunes/list-a-fine-tunes-model-files.md): The files a local run needs, with their hashes. decideLocal and decide_local call this to download a model, and about once an hour to learn it still exists. It does not count against the daily download limit.
- [Download one of a fine tune's model files](https://docs.tokz.si/api-reference/fine-tunes/download-one-of-a-fine-tunes-model-files.md): Streams one file listed by the bundle route, with its hash in x-tokz-sha256. Every file counts against the organization's daily download limit, so use GET and not HEAD, which counts the same bytes.

## OpenAPI Specs

- [openapi](/openapi.json)


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.