Back to Explore

openai/openai-node

GitHub
27 updates · last 90 days1 watchersOpen source

Last release:

Official OpenAI TypeScript and JavaScript client library that lets you call the OpenAI REST API from Node.js or other JS/TS environments. It provides convenient methods for common APIs such as the Responses API, supports streaming via SSE, and includes helpers for features like file uploads and workload identity authentication.

Project status

  • The repository is actively maintained, with an upstream push on 2026-09-17 and multiple recent package updates (v7.13.0 through v7.17.0).
  • Apparent update cadence is fast recently, with consecutive updates spaced roughly 1 to 2 days apart (for example v7.16.0 on 2026-09-15 and v7.17.0 on 2026-09-16).

AI summary generated

AI-generated from public sources. May be inaccurate. Report

Recent updates

  • v7.20.0

    v7.20.0 adds new Admin APIs for external storage configuration management, plus several safety and telephony-related event features, and introduces an environment-variable based vault credential option. It also includes a bug fix described as preserving request options in legacy GET calls. The code diff shows a broader, more opinionated change to how query parameters vs request options are handled across many Admin endpoints.

    Features
  • v7.19.0

    v7.19.0 adds webhook endpoint management to the `client.webhooks` resource, along with related API surface for managing webhook endpoints. It also deprecates MCP `connector_id` in favor of newer connector configuration fields. Beyond the release notes, the SDK changes debug logging behavior and redaction to summarize payloads rather than logging full JSON bodies.

    SecurityFeatures
  • v7.18.0

    v7.18.0 adds several new API capabilities, including agent session model settings, audio-mini model choices, prompt-cache prewarming, and managed Responses WebSocket sessions. It also includes a WebSocket bug fix focused on validating WebSocket results and preserving header defaults.

    Features
  • v7.17.0

    v7.17.0 adds API support for compaction progress events and includes multiple bug fixes around streaming and error preservation (chat runner, realtime WebSocket errors, responses, and streaming runTools). It also tightens CI behavior by retrying transient Deno download failures and updates SDK tooling for linting and vendored TypeScript transformer behavior.

    Features
  • v7.16.0

    v7.16.0 introduces per-iterator WebSocket buffering limits and improves robustness of WebSocket event handling, including better behavior around malformed events and buffering. It also includes Live transcript grouping performance improvements related to acknowledgment normalization and suffix handling, plus some internal request cancellation and credential scoping fixes.

    Features
  • v7.15.0

    v7.15.0 adds the new beta Agents API surface area, including environments, sessions (with streaming support), and related sub-resources (artifacts, events, items, turns, vaults, and more). The release notes only mention this feature at a high level, but the diff also includes supporting pagination infrastructure used by the Agents endpoints.

    Features
  • v7.14.0

    v7.14.0 introduces the new Live API surface in the SDK (new Live resources and an OpenAI.live client handle). The diff also adds substantial new implementation code for Live WebRTC/DataChannel plumbing and transcript grouping logic, alongside small type expansions for image generation call payloads. Release notes only explicitly call out the Live API feature and some dependency bumps in the ecosystem-tests fixtures.

    Features
  • v7.13.0

    v7.13.0 adds API key expiration controls. The code changes introduce new optional expiration fields for project API keys and service account API keys, and allows specifying an initial expiration window when creating service accounts.

    Features
  • v7.12.1

    Release v7.12.1 primarily updates the package version and release metadata, with CI-related changes to the Cloudflare worker ecosystem tests. The release notes claim GPT Image 2.5 support is included, but the provided code diff does not show any API source changes for that functionality.

    Features
  • v7.12.0

    v7.12.0 adds support for GPT Image 2.5 models and related image options, plus fixes around assistant run-step delta identity validation and file-processing timeout handling. It also adjusts workload identity authorization header handling to preserve intentionally suppressed Authorization values. Most runtime changes are limited to stream validation, timeout measurement, and auth header injection logic.

    Features
  • v7.11.0

    v7.11.0 adds several API-level enhancements, including prompt cache diagnostics and additional service-account API key expiration fields. The release also includes many bug fixes across audio, auth, embeddings, polling, responses, and realtime behaviors, along with documentation and test/CI infrastructure updates.

    SecurityFeatures
  • v7.10.0

    v7.10.0 adds support for the new model id gpt-6-astra, along with additional response-related types and behaviors. The release notes also mention adding the repository's canonical SDK security model documentation, but the code diff includes several other API surface additions (notably Safety endpoints/webhooks and new Responses item/event types).

    Features
  • v7.9.0

    v7.9.0 updates several API surface areas and documentation, including usage-related API types. It also fixes Responses WebSocket redirect behavior and improves compatibility for CommonJS loading under browser export conditions and Jest 28.

    BreakingFeatures
  • v7.8.0

    v7.8.0 adds `compute_units` to usage objects for Responses and Chat Completions, and adjusts browser and WebSocket behavior (including default WebSocket User-Agent). It also includes authentication and event-handling bug fixes, plus several SDK reliability tweaks.

    Features
  • v7.7.0

    v7.7.0 introduces an API change that makes function tool call output call IDs optional, and it fixes SSE streaming so cancellation is properly honored while decoding events. It also includes a significant internal rewrite of the SSE iterator to ensure aborts stop decoding and cleanup happens deterministically.

    BreakingFeatures
  • v7.6.0

    v7.6.0 adds new API surface fields (ChatCompletionChunk obfuscation, project residency/usage quantity fields) and new capabilities for Realtime WebRTC and named data residency. The release notes also describe a large set of security hardening and stream-parsing bounds, plus multiple bug fixes across streaming, auth, assistants, audio, and uploads.

    SecurityFeatures
  • v7.5.0

    v7.5.0 adds Bedrock Runtime endpoint support, new model identifiers, and additional API types and realtime/websocket stream metadata. It also includes many robustness and security fixes around streaming, event parsing, credential handling, and SSE/WebSocket behavior. The diff indicates broad generated SDK updates plus substantial CI and ecosystem test harness adjustments.

    BreakingSecurityFeatures
  • v7.4.0

    v7.4.0 adds the gpt-5.5 model to the SDK and extends response and beta response tool metadata types with optional tool name and namespace fields. It also includes an internal test/CI workflow reshuffle (including a shift of unit tests toward Vitest) plus several dependency bumps and an mTLS documentation update.

    Features
  • v7.3.0

    v7.3.0 adds an API surface for content provenance checks and updates a large portion of repository documentation and tooling. The release notes only mention the new content provenance checks feature, but the diff also shows packaging/publishing changes and several developer-facing API naming updates in examples and streaming/realtime documentation.

    Features
  • v7.2.0

    Release v7.2.0 introduces support for a new API service tier mode, Fast. The main code impact is updates to the TypeScript types and inline documentation for `service_tier` across Responses and Chat Completions related interfaces.

    Features