v1.3.0

stepfun-ai/Step-Audiov1.3.0Aug 7, 2026by cdoern

AI Summary

This release introduces new remote inference providers including Meta AI, Mistral, and DeepSeek, alongside a Neo4j vector provider and classifier reranker. It includes significant refactoring of the reasoning implementation and provider types, along with fixes for container bootstrapping, authentication warnings, and various security patches.

Key Highlights

  • Added new remote inference providers: Meta AI, Mistral, and DeepSeek
  • Introduced Neo4j vector provider and classifier reranker type
  • Refactored reasoning implementation and cleaned up type annotations
  • Fixed container startup issues and removed deprecated passthrough providers
  • Enhanced HTTP client configuration with connection pool limits

Breaking Changes

  • Removed the 'remote::passthrough' inference provider
  • Removed sklearn dependency

New Features

  • Meta AI remote inference provider
  • Mistral remote inference provider
  • DeepSeek remote inference provider
  • Neo4j vector provider
  • Bedrock auto-discovery of models
  • Durable out-of-process job scheduling for file_processors
  • HTTPx connection pool limits via network config
  • Classifier reranker type for post-retrieval chunk filtering

Full Release Notes

## What's Changed
* fix(ci): use --no-project in SDK Makefile to avoid circular dependency during release by @leseb in https://github.com/ogx-ai/ogx/pull/6273
* feat(inference): add Meta AI remote inference provider by @raghotham in https://github.com/ogx-ai/ogx/pull/6275
* feat(inference, meta): inject network/TLS config into Meta provider HTTP clients by @mattf in https://github.com/ogx-ai/ogx/pull/6277
* refactor(inference, messages): extract shared SSE passthrough stream helper for Anthropic-compatible providers by @mattf in https://github.com/ogx-ai/ogx/pull/6269
* fix(containers): pass --insecure to entrypoint to restore starter image boot by @derekhiggins in https://github.com/ogx-ai/ogx/pull/6288
* chore: remove unused prompt_for_config by @mattf in https://github.com/ogx-ai/ogx/pull/6286
* fix(inference): strip duplicate /v1 prefix in vLLM Anthropic message URLs by @nathan-weinberg in https://github.com/ogx-ai/ogx/pull/6294
* feat(inference, messages): add anthropic_count_tokens to OpenAIMixin via translation by @mattf in https://github.com/ogx-ai/ogx/pull/6268
* fix(ci): generate ogx-client in-repo instead of cloning external repo by @cdoern in https://github.com/ogx-ai/ogx/pull/6297
* chore: fix type annotations in core by @mattf in https://github.com/ogx-ai/ogx/pull/6296
* fix(ci): publish ogx-client-typescript now that its release branches exist by @cdoern in https://github.com/ogx-ai/ogx/pull/6301
* docs: add OpenAI-compatible gateway example with DaoXE by @seven7763 in https://github.com/ogx-ai/ogx/pull/6298
* docs(blog): add multi-tenant capabilities blog post by @leseb in https://github.com/ogx-ai/ogx/pull/6129
* feat(vector-io): Add the Neo4j vector provider by @franciscojavierarceo in https://github.com/ogx-ai/ogx/pull/6274
* fix(vector_io): prevent duplicate libomp crash when importing faiss on macOS by @leseb in https://github.com/ogx-ai/ogx/pull/6304
* fix(server): warn at startup when authentication is not configured by @nathan-weinberg in https://github.com/ogx-ai/ogx/pull/6307
* chore: clean up inference provider type annotations and remove exclusions by @mattf in https://github.com/ogx-ai/ogx/pull/6313
* fix(inference): remove duplicate per-provider model id resolution by @mattf in https://github.com/ogx-ai/ogx/pull/6324
* fix(inference): return requested model id in streamed completion chunks by @sohumt123 in https://github.com/ogx-ai/ogx/pull/6320
* fix(ui): improve vector store and citation workflows by @franciscojavierarceo in https://github.com/ogx-ai/ogx/pull/6325
* chore: bump fallback_version to 1.2.1.dev0 by @cdoern in https://github.com/ogx-ai/ogx/pull/6276
* chore(github-deps): bump actions/stale from 10.3.0 to 10.4.0 by @dependabot[bot] in https://github.com/ogx-ai/ogx/pull/6281
* chore(python-deps): bump sentence-transformers from 5.5.1 to 5.6.0 by @dependabot[bot] in https://github.com/ogx-ai/ogx/pull/6282
* refactor: run starter as non-root by @rhdedgar in https://github.com/ogx-ai/ogx/pull/6319
* chore(api-deps): bump opentelemetry-exporter-otlp-proto-http from 1.43.0 to 1.44.0 in /src/ogx_api by @dependabot[bot] in https://github.com/ogx-ai/ogx/pull/6322
* fix: bump pyasn1 for CVE-2026-59885 by @cdoern in https://github.com/ogx-ai/ogx/pull/6330
* chore(python-deps): bump mcp from 1.27.1 to 1.28.1 by @dependabot[bot] in https://github.com/ogx-ai/ogx/pull/6310
* chore(python-deps): bump torch from 2.12.0 to 2.13.0 by @dependabot[bot] in https://github.com/ogx-ai/ogx/pull/6317
* feat(inference): add Mistral remote inference provider by @NishchayMahor in https://github.com/ogx-ai/ogx/pull/6239
* fix(inference, anthropic): default json_schema.strict to true by @derekhiggins in https://github.com/ogx-ai/ogx/pull/6024
* fix(responses): only dedup user messages when function_call_outputs are present by @derekhiggins in https://github.com/ogx-ai/ogx/pull/6022
* fix(responses): populate file_citation annotations when models don't cite inline by @elinacse in https://github.com/ogx-ai/ogx/pull/6292
* feat(inference): add DeepSeek remote inference provider by @NishchayMahor in https://github.com/ogx-ai/ogx/pull/6240
* fix(responses): delimit untrusted web_search/file_search tool output before feeding it back to the model by @mittalpk in https://github.com/ogx-ai/ogx/pull/6337
* fix(inference): add _enforce_credentials=False to passthrough AsyncOpenAI client by @NickGagan in https://github.com/ogx-ai/ogx/pull/6344
* feat(inference): expose httpx connection pool limits via network config by @elinacse in https://github.com/ogx-ai/ogx/pull/6306
* feat(bedrock): auto-discover models via ListFoundationModels and mantle API by @Artemon-line in https://github.com/ogx-ai/ogx/pull/6315
* docs: fix Gemini provider auth docs to match mutually-exclusive behavior by @Aawegg in https://github.com/ogx-ai/ogx/pull/6341
* test(inference): update the Cerebras model ID by @ryanl-cerebras in https://github.com/ogx-ai/ogx/pull/6366
* feat(cli): rename 'ogx letsgo' to 'ogx go' with backward-compatible alias by @mattf in https://github.com/ogx-ai/ogx/pull/5815
* refactor: clean up reasoning implementation by @robinnarsinghranabhat in https://github.com/ogx-ai/ogx/pull/5429
* chore: remove remote::passthrough inference provider by @mattf in https://github.com/ogx-ai/ogx/pull/6368
* fix(inference, anthropic): emit tool messages before user text in translation by @LuShadowX in https://github.com/ogx-ai/ogx/pull/6367
* fix: remove sklearn dep by @mattf in https://github.com/ogx-ai/ogx/pull/6359
* feat(vector_io): add classifier reranker type for post-retrieval chunk filtering by @varshaprasad96 in https://github.com/ogx-ai/ogx/pull/5890
* feat: durable out-of-process job scheduling for file_processors by @cdoern in https://github.com/ogx-ai/ogx/pull/6160
* chore: updated discord invite link by @dominikkawka in https://github.com/ogx-ai/ogx/pull/6374
* fix: bump aiohttp for CVE-2026-69244 by @cdoern in https://github.com/ogx-ai/ogx/pull/6375
* fix: make AsyncOgxClient functional by inheriting from sync ApiClient by @aegeiger in https://github.com/ogx-ai/ogx/pull/6372

## New Contributors
* @seven7763 made their first contribution in https://github.com/ogx-ai/ogx/pull/6298
* @sohumt123 made their first contribution in https://github.com/ogx-ai/ogx/pull/6320
* @NishchayMahor made their first contribution in https://github.com/ogx-ai/ogx/pull/6239
* @mittalpk made their first contribution in https://github.com/ogx-ai/ogx/pull/6337
* @Aawegg made their first contribution in https://github.com/ogx-ai/ogx/pull/6341
* @ryanl-cerebras made their first contribution in https://github.com/ogx-ai/ogx/pull/6366
* @LuShadowX made their first contribution in https://github.com/ogx-ai/ogx/pull/6367
* @dominikkawka made their first contribution in https://github.com/ogx-ai/ogx/pull/6374

**Full Changelog**: https://github.com/ogx-ai/ogx/compare/v1.2.2...v1.3.0