Releases · mozilla-ai/otari-sdk-python

12 Jun 09:19

github-actions

otari-0.1.1

77c822a

otari: v0.1.1 Latest

Latest

0.1.1 (2026-06-12)

Features

independent release automation + surface gateway spec version (#12) (21b9b5b)
wrap /v1/messages/count_tokens (regenerate core + ergonomic method) (#10) (6704e56)

Bug Fixes

regenerate SDK client core so message.reasoning is a string (#14) (2cad75b)

Documentation

add AGENTS.md/CLAUDE.md agent guide, refresh README (#6) (56c3848)

Assets 2

10 Jun 15:01

njbrake

0.1.0

6704e56

0.1.0

First feature-complete release of the Python SDK. The client is now generated from the otari gateway's OpenAPI spec and covers the full API surface (inference and control plane), with a hand-written ergonomic layer on top. Both synchronous (OtariClient) and asynchronous (AsyncOtariClient) clients are included.

Still pre-1.0 (alpha): minor releases may include breaking changes while the surface settles.

Added

Full typed client core generated from the gateway OpenAPI spec, covering every endpoint: chat completions, responses, Anthropic-style messages, embeddings, moderations, rerank, models, and batches (#3).
Control-plane management with ergonomic methods for keys, users, budgets, pricing, and usage, so self-hosted gateways can be administered directly from the SDK. The raw generated methods remain available as an escape hatch (#7).
count_tokens() on the sync and async clients, wrapping POST /v1/messages/count_tokens. It counts the input tokens an Anthropic-style /messages request would consume without generating a response, and returns a typed CountTokensResponse (#10).
Typed responses across inference (ChatCompletion, embeddings, rerank, moderation, models) and streaming support for chat, messages, and responses.