# What changed in openai

> 338 changes in openai 6 to 7 from its official release notes, upgrade guides and advisories: 19 breaking, 3 deprecations, 22 security, 156 new APIs and options. Each one cited. Sources checked 2026-10-17. HTML: https://unstale.sh/npm/openai/

To see which of these apply to a repository, run `npx @unstale/cli init` in it.

## openai 7

### Breaking changes

- Streaming runTools now rejects unfinished turns. (7.17.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.17.0)
- WebSocket constructors for the Responses API now reject function-based API keys. (7.17.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.17.0)
- Rejects incomplete float32 vectors. (7.11.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.11.0)
- Responses WebSockets no longer follow redirects. (7.9.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.9.0)
- Hardens structured-output handling against format downgrade and preserves chat response-format integrity; Zod support is limited to provably JSON-native structured outputs. (7.6.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.6.0)
- Blocks custom credential headers on redirects. (7.6.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.6.0)
- The SDK now requires Node.js 22 and codifies version support, so developers must use Node.js 22 or later. (7.0.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.0.0)

### Security

- Safely decodes filenames inferred from URLs during uploads. (7.30.1) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.30.1)
- Callback credentials are kept local to each HTTP request. (7.16.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.16.0)
- WebSocket error messages from the Responses API are formatted safely. (7.16.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.16.0)
- Validates Azure metadata access token values. (7.11.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.11.0)
- First-class X.509 workload credentials are secured. (7.8.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.8.0)
- Rejects run-step identity hijacking before tool callbacks. (7.6.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.6.0)
- Redacts sensitive or malformed payloads from stream diagnostics and prevents sensitive assistant stream data from being retained or leaked. (7.6.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.6.0)
- Bounds SDK event streams and SSE/NDJSON buffering, validates streamed item identities, and rejects malformed Realtime WebSocket frames and named SSE provider error events. (7.6.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.6.0)
- Azure API keys are no longer disclosed across HTTP redirects. (7.5.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.5.0)
- Azure deployment path traversal is prevented. (7.5.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.5.0)
- Bedrock bearer credentials are no longer disclosed across redirects. (7.5.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.5.0)
- Bedrock credentials are contained to the configured endpoint origin. (7.5.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.5.0)
- Prototype manipulation in streamed chat snapshots is prevented. (7.5.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.5.0)
- Multipart upload content-type header injection is prevented. (7.5.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.5.0)
- runTools is hardened against inherited function names. (7.5.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.5.0)
- Streaming upload filename and header injection is prevented. (7.5.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.5.0)
- Redirects are rejected during workload identity token exchange. (7.5.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.5.0)
- Unsafe workload-identity token expirations are rejected. (7.5.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.5.0)

### Deprecations

- Deprecates connector_id for MCP. (7.19.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.19.0)
- The Sora video APIs are deprecated. (7.5.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.5.0)

### New APIs and options

- Image requests and responses now support safety identifiers. (7.36.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.36.0)
- Live sessions can now play audio fixtures. (7.35.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.35.0)
- Adds support for using Decisions in Live sessions. (7.35.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.35.0)
- Adds prompt caching options for Decisions, giving developers more control over prompt caching. (7.35.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.35.0)
- Adds support for updating project API key scopes. (7.35.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.35.0)
- Adds support for creating voices from prompts. (7.35.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.35.0)
- Adds response error diagnostics and model notices. (7.35.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.35.0)
- Adds options for regional processing residency, allowing developers to configure where processing occurs. (7.34.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.34.0)
- Adds beta session spending limits. (7.33.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.33.0)
- Adds billing_not_active as a session error code. (7.33.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.33.0)
- Adds browser and computer toolsets for developers to use. (7.33.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.33.0)
- Adds support for image URLs and detailed safety alert explanations. (7.33.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.33.0)
- Adds support for OCI as an external-storage provider. (7.33.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.33.0)
- Adds organization Decisions usage reporting. (7.33.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.33.0)
- Adds support for updating project API key expiry. (7.33.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.33.0)
- Adds shida to the named Live voices. (7.33.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.33.0)
- Adds vault updates and metadata filters. (7.33.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.33.0)
- Adds agent environment suspension and expiration. (7.32.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.32.0)
- Adds support for prewarmed hosted environments. (7.31.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.31.0)
- Adds standalone Decisions support to the API. (7.30.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.30.0)
- Typed function helpers support tool search with defer_loading=true. (7.29.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.29.0)
- Agents can observe local tool failures. (7.29.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.29.0)
- Adds support for agent session events. (7.28.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.28.0)
- Adds support for custom voice creation. (7.28.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.28.0)
- Agents can prepare hosted files. (7.27.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.27.0)
- Agents can download result artifacts. (7.27.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.27.0)
- Adds the ability to list agent session traces. (7.26.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.26.0)
- Beta Agents streams can collect their final output. (7.26.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.26.0)
- Typed application functions can be bound to Agents tools. (7.26.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.26.0)
- Agent streams can parse typed output. (7.26.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.26.0)
- Adds credential and session options for Agents. (7.24.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.24.0)
- Adds Cyber access programs to Responses. (7.24.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.24.0)
- Adds the GPT-6.1 Sol model identifier. (7.24.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.24.0)
- Adds a dedicated Node translation WebSocket. (7.24.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.24.0)
- Adds client secret creation for Realtime Translation. (7.24.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.24.0)
- Adds opt-in incremental output snapshots over WebSocket. (7.24.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.24.0)
- Adds support for GCP external storage. (7.23.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.23.0)
- Adds the GPT-Rosalind research model. (7.23.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.23.0)
- Adds the GPT-6 Luna model identifier for developers to use when selecting a model. (7.22.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.22.0)
- Adds the GPT-6 Sol model identifier for developers to use when selecting a model. (7.22.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.22.0)
- Adds session environment reset events. (7.21.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.21.0)
- Adds support for environment-variable vault credentials. (7.20.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.20.0)
- Adds API support for managing external storage configurations. (7.20.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.20.0)
- Adds support for retrieving safety cases. (7.20.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.20.0)
- Adds webhook events for safety warnings and deactivation. (7.20.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.20.0)
- Adds SIP media security information to incoming call events. (7.20.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.20.0)
- Adds API support for managing webhook endpoints. (7.19.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.19.0)
- Adds model settings for agent sessions. (7.18.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.18.0)
- Adds audio-mini model choices. (7.18.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.18.0)
- Adds managed WebSocket sessions for Responses. (7.18.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.18.0)
- Adds prompt-cache prewarming. (7.18.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.18.0)
- Adds compaction progress events so developers can track progress during compaction. (7.17.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.17.0)
- WebSocket iterators can have per-iterator limits on incoming events. (7.16.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.16.0)
- Adds the Agents API to the SDK. (7.15.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.15.0)
- Adds the Live API to the OpenAI Node SDK. (7.14.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.14.0)
- Adds controls for API key expiration. (7.13.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.13.0)
- Adds GPT Image 2.5 models and image options for use through the API. (7.12.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.12.1), [release notes](https://github.com/openai/openai-node/releases/tag/v7.12.0)
- Adds prompt cache diagnostics to the API, making cache information available to developers. (7.11.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.11.0)
- Adds expiration fields for service-account API keys. (7.11.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.11.0)
- The SDK adds gpt-6-astra and related features. (7.10.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.10.0)
- Responses and Chat Completions usage now include compute_units. (7.8.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.8.0)
- Adds the obfuscation field to ChatCompletionChunk, making it available to developers consuming chat completion chunks. (7.6.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.6.0)
- Adds support for named data-residency endpoints. (7.6.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.6.0)
- Adds support for backend-mediated Realtime WebRTC calls. (7.6.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.6.0)
- Adds project residency and usage quantity_unit fields to the API types. (7.6.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.6.0)
- Adds pinned X.509 workload token exchange and integrates attested X.509 workload identity into clients. (7.6.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.6.0)
- AzureClientOptions is exported for developers to use. (7.5.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.5.0)
- The SDK adds support for Bedrock Runtime endpoints, allowing developers to use the SDK with those endpoints. (7.5.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.5.0)
- A new Daybreak model identifier is available for use with the API. (7.5.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.5.0)
- A new gpt-5.6-cyber model identifier is available for use with the API. (7.5.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.5.0)
- An ImageDetail type is added to the SDK API. (7.5.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.5.0)
- Workload identity helpers are exported from openai/auth. (7.5.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.5.0)
- Realtime WebSocket connection URLs can now be overridden. (7.5.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.5.0)
- WebSocket events are separated in the API. (7.5.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.5.0)
- A ServiceTier type is added to the SDK API. (7.5.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.5.0)
- The API adds structured errors for MCP. (7.5.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.5.0)
- The API adds an Ultrafast tier. (7.5.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.5.0)
- The API adds structured WebSocket errors. (7.5.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.5.0)
- WebSocket stream IDs are added to the API. (7.5.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.5.0)
- The API adds an event for when a workload identity access token is issued. (7.5.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.5.0)
- Adds the gpt-5.5 model to the API. (7.4.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.4.0)
- Adds tool metadata fields to the API. (7.4.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.4.0)
- Adds content provenance checks to the API. (7.3.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.3.0)
- Adds fast tier support to the API. (7.2.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.2.0)

### Performance

- Bedrock avoids copying repeated query values, improving performance. (7.11.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.11.0)
- Avoids quadratic comparisons when accumulating long text in Responses. (7.6.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.6.0)
- Quadratic EventStream FIFO draining is eliminated. (7.5.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.5.0)
- Quadratic SSE buffer copying is eliminated. (7.5.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.5.0)

### Behavior and default changes

- Exposes image tool error and moderation details. (7.33.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.33.0)
- Local tools can run during session creation streams. (7.29.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.29.0)
- Computer use is now available to beta agents. (7.25.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.25.0)
- Usage APIs were updated; developers may need to adjust how they use them. (7.9.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.9.0)
- WebSocket connections now send a default User-Agent. (7.8.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.8.0)
- Function call output no longer requires call IDs, so developers can omit call_id. (7.7.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.7.0)
- Transcription models were updated. (7.1.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.1.0)

### Fixes and other changes

- Declares Realtime response error messages in the SDK types. (7.35.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.35.0)
- Aligns beta pagination limits and non-null parameters. (7.33.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.33.0)
- Preserves detailed token usage in totalUsage. (7.32.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.32.0)
- Corrects the parameters for custom voice creation. (7.30.1) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.30.1)
- Structured text is parsed only in final message phases. (7.30.1) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.30.1)
- Chat Completions now allows the original image detail value. (7.26.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.26.0)
- Corrects the endpoint used to cancel an eval run. (7.26.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.26.0)
- Lane IDs are reserved across WebSocket session replacements. (7.26.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.26.0)
- Credentials are refreshed when a WebSocket reconnects. (7.26.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.26.0)
- Preserves queries on the base URL when joining relative endpoints. (7.24.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.24.0)
- Prevents replay of queued writes whose delivery is uncertain. (7.24.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.24.0)
- Prevents hangs when transcript grouping deadlines are fractional. (7.24.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.24.0)
- Preserves supported streamed logprobs and provisional metadata in Responses. (7.24.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.24.0)
- Preserves model choices and improves request handling. (7.21.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.21.0)
- Preserves request options in legacy GET calls. (7.20.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.20.0)
- Validates WebSocket results and preserves header defaults. (7.18.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.18.0)
- Chat runners now preserve abort reasons. (7.17.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.17.0)
- Realtime handling now preserves native WebSocket error causes. (7.17.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.17.0)
- Strict schemas now omit an impossible optional branch. (7.17.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.17.0)
- Streaming now emits terminal SSE events even when they lack a trailing blank line. (7.17.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.17.0)
- Malformed WebSocket events are handled, and buffering is improved. (7.16.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.16.0)
- Unread events are discarded when an iterator returns. (7.16.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.16.0)
- Rejects root IDs in run-step deltas. (7.12.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.12.0)
- Preserves suppressed workload authorization. (7.12.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.12.0)
- Uses ALSA card identifiers when recording audio. (7.11.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.11.0)
- Preserves API error classification for streamed errors. (7.11.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.11.0)
- Waits for input cleanup to finish before reporting recording success. (7.11.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.11.0)
- Prefers readable inputs over body metadata when handling audio inputs. (7.11.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.11.0)
- Signals only recording processes that have started. (7.11.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.11.0)
- Cancels Bedrock requests that are awaiting AWS credentials. (7.11.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.11.0)
- Rejects zero batch upload concurrency. (7.11.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.11.0)
- Treats cancelled vector store files as terminal rather than continuing to poll. (7.11.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.11.0)
- Accepts ArrayBuffers originating in another realm for uploads. (7.11.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.11.0)
- Supports the Edge Runtime when Node process is unavailable. (7.11.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.11.0)
- Preserves numeric embedding output when request bodies are overridden. (7.11.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.11.0)
- Preserves numeric responses after request hooks run. (7.11.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.11.0)
- Validates the response collection before decoding embeddings. (7.11.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.11.0)
- Preserves explicitly empty filenames during uploads. (7.11.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.11.0)
- Corrects the fields reported in function argument completion events. (7.11.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.11.0)
- Preserves alternate Headers inputs. (7.11.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.11.0)
- Parses JSON media types without case sensitivity. (7.11.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.11.0)
- Parses namespaced function tools in Responses. (7.11.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.11.0)
- Follows next cursors through empty pages so pagination can continue. (7.11.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.11.0)
- Bounds partial number recovery to its token. (7.11.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.11.0)
- Honors zero as a custom polling interval. (7.11.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.11.0)
- Preserves routing components in the Realtime base URL. (7.11.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.11.0)
- Validates indexed response collections. (7.11.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.11.0)
- Omits parsed values from replayed Responses history. (7.11.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.11.0)
- Rejects invalid retry counts. (7.11.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.11.0)
- Preserves retry delays when work runs in parallel workers. (7.11.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.11.0)
- Interrupts pending reads when a stream is cancelled. (7.11.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.11.0)
- Surfaces server error frames and settles streams even after listeners throw. (7.11.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.11.0)
- Retains terminal message snapshots so they remain available after completion. (7.11.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.11.0)
- Accounts for elapsed time during token exchange. (7.11.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.11.0)
- Serializes tool-call IDs once per completion. (7.11.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.11.0)
- Treats a cancelled vector store file as a terminal polling state. (7.11.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.11.0)
- Recognizes the incomplete status for web search calls. (7.11.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.11.0)
- Carries declared options through withOptions and fixes Azure deployment paths. (7.11.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.11.0)
- CommonJS now loads under browser export conditions. (7.9.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.9.0)
- CommonJS package imports are supported with Jest 28. (7.9.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.9.0)
- Multipart bodies are detected from own properties only. (7.8.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.8.0)
- Native browser ESM imports are restored. (7.8.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.8.0)
- WebSocket waiters are settled when listeners throw. (7.8.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.8.0)
- The workload-identity refresh buffer is clamped to the token lifetime. (7.8.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.8.0)
- Cancellation is now honored while decoding SSE responses. (7.7.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.7.0)
- Preserves Blob files and multipart media types during uploads. (7.6.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.6.0)
- Honors cancellation during polling delays and Bedrock bearer credential resolution, and stops delayed or additional tool callbacks after cancellation. (7.6.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.6.0)
- Preserves large streaming responses. (7.6.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.6.0)
- Preserves array order when creating multipart uploads. (7.6.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.6.0)
- Prevents logprobs from being double-counted in streamed chat completions. (7.6.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.6.0)
- Bounds the work performed during webhook signature verification. (7.6.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.6.0)
- The audio upload metadata requirements are clarified. (7.5.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.5.0)
- Mid-stream errors are surfaced when iterating buffered streams. (7.5.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.5.0)
- Custom tool calls are supported in ChatCompletionStream. (7.5.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.5.0)
- Custom tool calls are supported in the parser and ChatCompletionStream. (7.5.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.5.0)
- Azure preserves model-to-deployment mapping for images.edit. (7.5.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.5.0)
- Streamed Responses output text is kept linear and canonical. (7.5.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.5.0)
- Responses streams reject with OpenAIError when the stream emits an error event. (7.5.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.5.0)
- SSE parsing handles mixed line endings. (7.5.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.5.0)
- SSE streams finish immediately after [DONE]. (7.5.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.5.0)
- toFile no longer throws a TypeError for Responses without a URL. (7.5.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v7.5.0)

## openai 6

### Breaking changes

- Promises emitted through EventEmitter are now rejected when an error occurs. (6.45.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.45.0)
- Incomplete structured output is no longer parsed. (6.45.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.45.0)
- The migrate CLI has been removed. (6.40.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.40.0)
- Removes web_search_call.results from ResponseIncludable. (6.34.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.34.0)
- Makes the type field required in ResponseInputMessageItem. (6.33.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.33.0)
- The detail field was removed from ResponseInputFile. (6.28.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.28.0)
- The detail field was removed from ResponseInputFileContent. (6.28.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.28.0)
- The GA ComputerTool now uses the CompuerTool class. (6.27.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.27.0)
- The computer_use_preview tool is moved to ComputerUsePreview. (6.27.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.27.0)
- Updates the Responses API parameter for prompt_cache_key. (6.26.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.26.0)
- The model parameter is now required for the responses/compact endpoint. (6.11.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.11.0)
- InputAudio is removed from ResponseInputContent. (6.8.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.8.0)

### Security

- Replayed messages are now sanitized by runTools. (6.49.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.49.0)
- Adds validation for workload identity access tokens. (6.45.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.45.0)
- API-key headers are redacted in debug logs. (6.37.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.37.0)
- Tightens auth header selection. (6.36.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.36.0)

### Deprecations

- Marks assistants as deprecated. (6.17.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.17.0)

### New APIs and options

- Support was added for spend_limit admin APIs. (6.49.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.49.0)
- Helpers now support standard schemas. (6.49.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.49.0)
- A helper for Zod realtime functions was added. (6.49.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.49.0)
- Zod now supports schema definitions. (6.49.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.49.0)
- Adds an endpoint for managing API keys for a service account within an organization project. (6.48.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.48.0)
- Adds owner_project_access to APIKeyListParams. (6.47.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.47.0)
- Adds async event iterators for consuming events. (6.47.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.47.0)
- Adds fromReadableStream to ResponseStream. (6.47.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.47.0)
- Adds support for streaming file uploads. (6.47.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.47.0)
- API updates for gpt-5.6-sol; developers can use the updated model support. (6.46.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.46.0)
- Adds an afterCompletion hook to runTools. (6.45.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.45.0)
- Realtime now supports sideband connections using call_id. (6.45.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.45.0)
- Adds a helper for response input items. (6.45.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.45.0)
- Adds a helper to accumulate response streams. (6.45.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.45.0)
- Adds the chat completions moderation API. (6.42.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.42.0)
- Adds the responses moderation API. (6.42.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.42.0)
- Adds support for using Responses with Amazon Bedrock. (6.41.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.41.0)
- Responses now include an additional_tools item. (6.40.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.40.0)
- The responses compact method accepts the service_tier parameter. (6.38.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.38.0)
- Admin organization usage responses now include a quantity field. (6.37.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.37.0)
- Realtime translation is launched. (6.37.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.37.0)
- Adds support for Admin API Keys per endpoint. (6.36.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.36.0)
- Adds group_type and user metadata fields and updates types across admin resources. (6.36.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.36.0)
- Adds the OAuthErrorCode type for representing OAuth error codes. (6.35.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.35.0)
- Adds the prompt_cache_retention parameter to responses compact. (6.35.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.35.0)
- Responses now support results as an output option for web_search_call. (6.35.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.37.0), [release notes](https://github.com/openai/openai-node/releases/tag/v6.35.0)
- WebSocket clients support binary messages. (6.35.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.35.0)
- WebSocket clients support path parameters. (6.35.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.35.0)
- WebSockets are supported in the browser when using simple auth. (6.35.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.35.0)
- The underlying WebSocket type is exposed for use by developers. (6.35.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.35.0)
- Adds the phase field to Message in conversations. (6.34.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.34.0)
- The client adds support for short-lived tokens. (6.34.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.34.0)
- Adds a keys field to computer action types. (6.33.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.33.0)
- Adds stream() and async iterator support to WebSocket classes. (6.33.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.33.0)
- Added the 5.4 mini model slug. (6.32.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.32.0)
- Added the 5.4 nano model slug. (6.32.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.32.0)
- Adds in and nin filter types to ComparisonFilter, allowing developers to use these filters. (6.31.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.31.0)
- Adds the defer_loading field to NamespaceTool. (6.30.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.30.0)
- Batches can now use the /v1/videos endpoint. (6.30.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.30.0)
- Adds support for custom voices. (6.29.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.29.0)
- Sora API improvements add a character API. (6.28.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.28.0)
- Sora API improvements support higher resolution exports. (6.28.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.28.0)
- Sora API improvements add video edits. (6.28.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.28.0)
- Sora API improvements add video extensions. (6.28.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.28.0)
- Adds a new computer tool to the API. (6.26.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.26.0)
- Adds gpt-5.4 support to the API. (6.26.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.26.0)
- Adds the tool search tool to the API. (6.26.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.26.0)
- Adds phase to the API. (6.25.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.25.0)
- Added gpt-audio-1.5 to realtime. (6.24.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.24.0)
- Added gpt-realtime-1.5 to realtime. (6.24.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.24.0)
- WebSockets are now available for the Responses API. (6.23.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.23.0)
- Container API support was added for network_policy. (6.22.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.22.0)
- Container API support was added for skills. (6.22.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.22.0)
- Images are now supported in the batch API, allowing developers to include them in batch requests. (6.21.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.21.0)
- Adds API support for hosted shell. (6.20.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.20.0)
- Adds API support for skills. (6.20.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.20.0)
- Adds context_management support to the responses API. (6.19.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.19.0)
- Responses support image generation actions. (6.18.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.18.0)
- Adds the name field to ResponseFunctionCallArgumentsDoneEvent. (6.18.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.18.0)
- Adds a status field to shell_call_output. (6.17.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.17.0)
- Adds the completed_at prop to Response. (6.16.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.16.0)
- Adds support for gpt-image-1.5. (6.14.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.14.0)
- Dated slugs were added for sora-2. (6.13.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.13.0)
- Adds support for gpt-5.2 in the API. (6.12.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.12.0)
- Adds support for the gpt-5.1-codex-max model. (6.10.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.10.0)
- Adds the compact Responses API endpoint. (6.10.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.10.0)
- Adds API support for gpt-5.1. (6.9.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.9.0)
- Hybrid searching adds ranking options. (6.8.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.8.0)
- The Realtime API adds token_limits. (6.8.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.8.0)
- Adds support for zod 4 schemas. (6.7.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.7.0)
- Adds responses.input_tokens.count to the API, allowing developers to access the input token count. (6.6.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.6.0)
- The audio/transcriptions endpoint now supports gpt-4o-transcribe-diarize. (6.4.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.4.0)

### Behavior and default changes

- The API now accepts None for prompt_cache_key. (6.49.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.49.0)
- The API now accepts None for safety_identifier. (6.49.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.49.0)
- Passes context to runTools callbacks. (6.47.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.47.0)
- Workload identity is now included in audit logs. (6.40.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.40.0)
- Messages can be queued while waiting for a connection. (6.35.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.35.0)
- WebSocket connections can automatically reconnect. (6.35.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.35.0)

### Fixes and other changes

- Azure endpoints now have trailing slashes normalized. (6.49.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.49.0)
- Parsed tool callback arguments are now inferred. (6.49.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.49.0)
- Stream handling now accounts for audio completion markers. (6.49.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.49.0)
- Schema definition references are escaped before strict validation. (6.49.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.49.0)
- Schema types are preserved while references are escaped. (6.49.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.49.0)
- Header names are normalized before appending, fixing inconsistent header handling. (6.48.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.48.0)
- Handles Azure filter stream chunks. (6.47.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.47.0)
- Avoids deep Zod types in Deno. (6.47.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.47.0)
- Preserves readable stream deltas in Assistants. (6.47.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.47.0)
- Emits stream finalization errors. (6.47.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.47.0)
- Adds support for zod v4 mini schemas. (6.47.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.47.0)
- Array delta entries are placed by index instead of appended, preserving their intended positions. (6.46.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.46.0)
- The runner normalizes missing tool call IDs. (6.46.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.46.0)
- Sanitizes extracted definition references in the Zod integration. (6.45.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.45.0)
- Fully resolves extracted wrapper definitions in the Zod integration. (6.45.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.45.0)
- Optional bodies set to null are no longer serialized. (6.45.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.45.0)
- Restores output_text on streamed final responses. (6.45.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.45.0)
- Leading whitespace chunks no longer break the partial parser for structured outputs. (6.45.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.45.0)
- ActionSearch.query is now optional. (6.40.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.40.0)
- Treat text/plan with format: binary as a raw upload. (6.39.1) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.39.1)
- Improves guidance for undici dispatcher mismatches. (6.39.1) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.39.1)
- Types now allow runtime fetch options. (6.39.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.39.0)
- Image 2 is updated. (6.37.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.37.0)
- The size enum regression is fixed. (6.37.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.37.0)
- Fixes support for admin API key authentication. (6.36.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.36.0)
- InputFileContent has additional detail in its API definition. (6.35.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.35.0)
- Corrects the prompt_cache_retention enum value in chat/completions and responses. (6.35.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.35.0)
- Allows a single message to exceed the WebSocket queue size. (6.35.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.35.0)
- Aligns SDK response types with expanded item schemas. (6.33.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.33.0)
- Fixes the phase enum. (6.25.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.25.0)
- Prevents abort listeners from being removed too early. (6.18.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.18.0)
- Fixes a memory leak involving abort signals. (6.18.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.18.0)
- The client no longer parses responses with an empty content-length. (6.18.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.18.0)
- Updates the type for the find_in_page action. (6.18.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.18.0)
- Undoes a change to the web search Find action. (6.18.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.18.0)
- The grader input list was fixed, which may affect how developers provide grader inputs. (6.13.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.13.0)
- Corrects the API endpoint used by the code tool. (6.11.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.11.0)
- Returns the correct lines when TypeScript errors occur. (6.11.0) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.11.0)
- Aligns the TypeScript types for input items and output items. (6.9.1) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.9.1)
- The nullability of logprobs was fixed, clarifying its type for developers. (6.8.1) — [release notes](https://github.com/openai/openai-node/releases/tag/v6.8.1)