Releases

Current stable release: 1.1.5. The entries below summarize the 1.0 line and preserve notable v0 history.

Recent changes

Version	Type	Summary
1.1.5	Patch	Supported Azure AI Foundry Responses streams that omit the function name from the terminal tool-argument event.
1.1.4	Patch	Updated upstream runtime and schema dependencies across the workspace.
1.1.3	Patch	Declared compatible workspace peer ranges so additive internal dependency releases do not force major releases of dependents.
1.1.2	Patch	completionModel() api now optional, defaulting to Chat Completions; gpt-6-astra added to the known model list (Grok provider tools stay Responses-only).
1.1.1	Patch	Bumped upstream runtime dependencies to their latest versions, aligned zod to 4.5.4 across all packages and workspaces, and updated the Core dependency to 1.1.1.
1.0.10	Patch	Updated the Core dependency to 1.0.10.
1.0.9	Patch	Added typed completion model controls with provider-neutral reasoning effort support, Agent defaults, per-run overrides, Studio selectors and persistence, and normalized observability attributes.
1.0.8	Patch	Updated the Core dependency to 1.0.8.
1.0.7	Patch	Updated the Core dependency to 1.0.7.
1.0.6	Patch	Updated the Core dependency to 1.0.6.
1.0.5	Patch	Declared and verified Bun 1.3.14 runtime compatibility across the OpenAI SDK transport and media paths, and stabilized structured tool-output branding across Core module instances.
1.0.4	Patch	Updated the Core dependency to 1.0.4.
1.0.3	Patch	Updated the Core dependency to 1.0.3.
1.0.2	Patch	Updated the Core dependency to 1.0.2.
1.0.1	Patch	Refreshed upstream SDK and runtime dependencies to their latest supported releases.
1.0.0	Major	Prepared the synchronized Anvia 1.0 release train.
0.5.1	Patch	Published updated upstream runtime dependencies.
0.5.0	Minor	Added model-aware context limits and provider-reported active context usage across normalized completion results and streams.
0.4.1	Patch	Improved complete-input tracing, safe/full capture behavior, prompt metadata, usage normalization, and observability integration.
0.4.0	Minor	Added provider-executed tool contracts and normalized provider tool/citation events used by compatible Responses integrations.

Compatibility and correctness milestones

The 0.3.x series contains several behaviorally important hardening changes:

• 0.3.25 emitted public streaming tool-call deltas by default.
• 0.3.23 stabilized compatible-provider reasoning stream identity.
• 0.3.22 retained authoritative usage on failed Responses streams when supplied.
• 0.3.21 rejected invalid streaming tool indices and incomplete terminal tool metadata.
• 0.3.20 rejected malformed JSON tool arguments.
• 0.3.16 preserved refusal text and terminal Responses failure states.
• 0.3.15 introduced autocomplete-friendly model name types while keeping custom IDs.
• 0.3.12 moved Core to a peer dependency to avoid duplicate private-type incompatibilities.
• 0.3.4 hardened embedding and image response validation.

These summaries are selective. Review the complete source history before upgrading across several versions.

• Full @anvia/openai changelog
• Compatibility and versioning
• API reference