Typed decision/classification service with an in-process gte-small ONNX embedding engine and lexical fallback, JS/Python SDKs, adapters, and the otsone model family. Apache-2.0 (sub-packages keep their MIT licenses). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
1.6 KiB
1.6 KiB
title, icon
| title | icon |
|---|---|
| Changelog | lucide/history |
Changelog
OpenJev fork (2026-09-23)
Renamed to openjev-adapters (import openjev_adapters) and rebranded to the
OpenJev ecosystem. Based on
@typesafe-ai/system-one-adapter-python
(MIT); see NOTICE.
Features
- env-configurable endpoint: base URL and credential resolve from
OPENJEV_BASE_URL/TYPESAFE_BASE_URLandOPENJEV_API_KEY/TYPESAFE_API_KEY, defaulting to the hosted TypeSafe API (https://api.typesafe.ai/v1) and staying wire-compatible with a self-hosted OpenJev deployment. Addsresolve_base_url,resolve_api_key,build_endpoint_provider, andbuild_async_endpoint_provider.
The request/response shapes (model + state + questions-as-dict → Choice/Score/Noul)
are unchanged, so it remains a drop-in for typesafe_sdk.
v0.2.1 (2026-09-22)
Bug fixes
- reject incomplete chat completions
- allow OpenAI responses without token usage
- stop retrying provider refusals as malformed output
- support both
httpxandhttpx2provider SDKs
Features
- support Gemini provider
v0.2.0 (2026-09-18)
Breaking Changes
- ser/de library has been changed from
msgspectopydanticastypesafe-sdkdid inv0.7.0
Features
- support
typesafe-sdk>=0.7.0
v0.1.5 (2026-09-18)
Bug fixes
- constrained
typesafe-sdkversion to>=0.6.0,<0.7.0
v0.1.4 (2026-09-16)
Bug fixes
- reuse providers and close owned SDK clients. Thanks @AbdelStark!
v0.1.3 (2026-09-15)
Initial release.