Packages

@ai-sdk/openai

4.0.108 · 3 Nov 2026Apache-2.022M / weekRepositoryHomepagenpmAtom feed

Upgrading @ai-sdk/openai From 3.0.76 to 3.0.136

59 releases in betweenevery note read

Reading releases, release notes and advisories…

Other targets from 3.0.76Latest4.0.108
  1. 3.0.76yours
  2. 58 releases
  3. 3.0.136target
0
changes to make
0
advisories fixed
14
new
40
fixes

Release by release

  1. 3.0.136

    3 Nov 20262 fixes
    • fixed

      openai: Preserve valid JSON string tool arguments and serialize BigInt and cycles in tool calls and results, retaining malformed argument replay fallbacks

    • fixed

      Forward languages and keywords transcription hints as repeated multipart fields instead of silently dropping them.

  2. 3.0.135nothing new or fixed in the note

  3. 3.0.134

    27 Oct 20262 fixes
    • fixed

      openai: Report unfinished web searches as tool errors

    • fixed

      Report failed and incomplete web searches as tool errors instead of successful tool results in OpenAI Responses API calls.

  4. 3.0.133

    27 Oct 20261 new · 1 fix
    • new

      openai: Add GPT-6.1 Terra model support

    • fixed

      Fix gpt-transcribe REST transcription. Requests with OpenAI provider options now use response_format: 'json' for gpt-transcribe and dated gpt-4o-mini-transcribe snapshots, which reject verbose_json. durationInSeconds is read from usage.seconds when a duration-billed response has no top-level duration.

  5. 3.0.132

    26 Oct 20261 new
    • new

      Add azure.tools.toolSearch and support azure provider options on responses function tools

  6. 3.0.131

    21 Oct 20261 new · 1 fix
    • new

      openai: Support 12h prompt cache writes and cache prewarming

    • fixed

      Defer validation schema construction until first use to reduce provider import time.

  7. 3.0.130nothing new or fixed in the note

  8. 3.0.129

    15 Oct 20261 fix
    • fixed

      Restore raw OpenAI custom-tool input deltas for consumers that forward them to Responses API clients. Mark streamed input with inputEncoding: 'text' so UI clients can display it progressively without JSON parsing. Preserve the encoding when partial UI messages are saved and resumed. Final provider tool-call inputs remain JSON-encoded.

32 internal lines (docs, tests, tooling) left out.