what stackpulse tracks
OpenAI Node releases from GitHub
StackPulse watches OpenAI Node release notes and keeps the original source link close to every summary.
Official Node.js library for the OpenAI API StackPulse turns upstream changelogs into scannable summaries with risky changes, deprecations, migration notes, and source links.
what stackpulse tracks
StackPulse watches OpenAI Node release notes and keeps the original source link close to every summary.
upgrade risk
Risky changes are separated from normal feature notes so you can scan upgrade impact before changing production dependencies.
migration notes
Migration steps and recommended actions are only shown when the upstream release notes support them.
This release updates API types for Agents, Live, Images, and admin. The notes report no other changes.
Users of the OpenAI Node SDK working with Agents, Live, Images, or admin APIs are affected by the type updates.
This release adds API typing for reasoning input transformations on Responses. It also updates development dependencies used by the Cloudflare Worker and browser Webpack ecosystem tests, and simplifies review handoff documentation.
Developers using the Node SDK's Responses API may benefit from improved types for reasoning input transformations.
This release adds typed async function tools to the API. It also updates several dependencies, including the runtime dependency Zod; no security fixes or breaking changes are described.
Node SDK users who work with async function tools can use the new typed support.
This release adds recognition of gpt-6-rosalind in beta Responses. It also retains agent scheduling support and consolidates repository guidance documentation.
Users working with beta Responses can use the newly recognized gpt-6-rosalind model.
This release adds support for delivering session errors as stream events in Agents. It also updates several development dependencies and the Azure identity dependency; no breaking changes, security fixes, or migration instructions are noted.
Users of the Agents streaming functionality may receive session errors as stream events.
This release adds a `model` option for the Responses API. It also updates the `follow-redirects` development dependency used by the Cloudflare Worker ecosystem tests.
Developers using the Responses API can now specify a model through the added option.
This release adds API support for Australia, India, and Japan data residency options, adds GPT-6 Rosalind to the model catalog, and expands image, audio, and agent-turn options. It also updates a development dependency used in browser Webpack ecosystem tests.
Node.js SDK users can access new residency settings, a newly cataloged model, and expanded image, audio, and agent-turn API options.
This release adds 12-hour caching for Responses, including cache-write usage reporting, and introduces project service tier settings. It also updates the Puppeteer development dependency used by browser tests and clarifies generated-file ownership documentation.
OpenAI Node users can use the new Responses caching and project service tier capabilities; browser-test maintainers receive a Puppeteer dependency update.
Upgrade to v7.50.0 to use the new Responses caching and project service tier features.
This release adds API support for workload identity providers and federation rules. The release notes do not describe further details or migration requirements.
Users of the OpenAI Node SDK who need workload identity provider or federation rule APIs are affected.
This release expands Live Responses delegation tools and choices, introducing new variants that may require updates to exhaustive TypeScript switches. It also improves retention of sparse WebSocket output facts; remaining changes update development dependencies in ecosystem tests.
Users of Live Responses with exhaustive TypeScript switches may need to handle the newly added variants.
Review exhaustive TypeScript switches for Live Responses delegation and add handling for new variants.
This release adds API support for typed warmup of Responses WebSocket connections. No other changes are described in the release notes.
Users of the OpenAI Node SDK who warm up Responses WebSocket connections may use the new typed support.
This release adds a broad set of API capabilities, including agent context compaction, session review and tool settings, experimental time awareness, and session metadata filtering. It also expands image editing, organization key management, SIP/Realtime functionality, and hosted-tool usage reporting, while fixing Realtime WebSocket session-target selection.
OpenAI Node users adopting agent, image, organization, SIP, Realtime, or Responses APIs can use the added settings and capabilities.
Upgrade to v7.46.0 to use the new API capabilities or receive the Realtime WebSocket session-target fix.
This release clarifies the API behavior for prompt cache breakpoint TTLs and adds compatibility with google-auth-library v11 for Vertex. No breaking changes or security fixes are noted.
Users integrating the SDK with Vertex and google-auth-library v11 are affected; the release also clarifies prompt cache breakpoint TTL behavior for API users.
Upgrade if you use google-auth-library v11 with Vertex or need the clarified prompt cache breakpoint TTL behavior.
This release adds the `gpt-6.1-terra` model identifier to the API. It also updates development and dependency packages, including `undici`; the notes describe no user-facing fixes, breaking changes, or security patches.
Node SDK users who need to reference the `gpt-6.1-terra` model can use its newly added model identifier.
Upgrade to v7.45.0 if you need the `gpt-6.1-terra` model identifier.
Adds agent automation APIs, including invocation endpoints, delegation controls, and prompt-cache lifetimes. Live Decision events now support message and tool-call variants, changing the event item types; the release also updates image-model guidance following the shutdown of gpt-image-1.
SDK users handling Live Decision events may need code changes, and applications explicitly selecting gpt-image-1 need to choose a replacement model.
Update Live Decision event handling for the new item unions and replace any explicit gpt-image-1 selection with a supported model.
This release expands Live Decisions API types: `instructions` can now be a string or an array of text, image, or JSON parts, and Decisions inputs support additional message roles and tool-history items. These type changes can break code that assumes instructions are always strings or that Decisions inputs have a fixed shape; the release also includes dependency updates in ecosystem tests.
Applications using Live Decisions instructions or typing Decisions inputs may need updates; existing string instructions and user-message inputs continue to work.
Review Live Decisions code for string-only instruction handling and assumptions about input types, then update it for the new union types.
This release adds an API for deleting voices. It also reduces repeated ancestor-path hashing and avoids excessive work when redacting long log values.
OpenAI Node users can now delete voices through the API; users with agent path processing or long log values may also benefit from the performance fixes.
This release adds Vertex AI support for the Responses API. The release notes do not describe additional changes.
Users integrating the OpenAI Node SDK with Vertex AI and the Responses API are affected.
Upgrade to v7.41.0 if you need Vertex AI support for the Responses API.
This release adds API support for 30-minute cache-write usage counts, compaction trigger IDs, and custom voice consent management. It also fixes WebSocket handling for large messages, option overrides, lifecycle state, and connection retries, alongside request/response encoding, multipart headers, and Windows audio input fixes.
OpenAI Node users working with usage reporting, compaction, custom voices, WebSockets, or Windows audio may benefit from these additions and fixes.
Upgrade to v7.40.0 to access the new API support and WebSocket and audio fixes.
This release adds opt-in bounded connection recovery for realtime connections and updates moderation examples to be more neutral. No breaking changes were introduced.
Developers using realtime connections may benefit from the new recovery feature.
Consider enabling bounded connection recovery if using realtime features.