Skip to content

v1.0.0

Latest

Choose a tag to compare

@openai-sdks openai-sdks released this 21 Nov 03:46
daafe46

1.0.0 (2026-11-21)

⚠ BREAKING CHANGES

  • api: Ruby 1.0 requires Ruby 3.3.0 or later. See the Ruby SDK 1.0 migration guide for code and type changes.
    • Assistants and Threads. Removed client.beta.assistants, client.beta.threads, and their SDK types. Use Responses and Conversations.
    • Videos. Removed client.videos, its SDK types, the Sora 2 model constants, and /v1/videos as a Batch endpoint.
    • Retired Chat Completions models. Replace GPT_4_VISION_PREVIEW, GPT_3_5_TURBO_0613, and GPT_3_5_TURBO_16K_0613 with a supported model ID.
    • Responses compaction. GA output uses ResponseItem and beta output uses BetaResponseItem. Both can contain retained user, system, and developer messages alongside assistant output and compaction records. Update collection annotations and use ResponseFunctionToolCallItem or BetaResponseFunctionToolCallItem for compacted function calls.
    • Live Responses delegation. Use ToolChoice::Function and ToolChoice::Mcp in delegation and update configs. MCP choices require server_label and allow name to be omitted or set to nil.
    • Named graders. Pass a graders: hash keyed by grader name to MultiGrader, and use those names in calculate_output. Replace the MultiGrader::Graders type with MultiGrader::Grader.
    • Image sizes. Read size as a string in image responses and generation or edit stream events. Compare it with a string such as "1024x1024". The named size constants and Size.values use strings.

Features

Bug Fixes

  • api: decode assistant messages from responses.input_items.list as ResponseOutputMessage (#883) (6670181)
  • vertex: omit underlying exceptions from credential lookup and timeout errors (#1051) (6670181)

Chores

  • deps: bump ruby/setup-ruby from 1.342.0 to 1.343.0 (#1096) (b5e3080)
  • finalize Ruby 1.0 generation checkpoint (#1099) (681e6a9)

Documentation

  • api: clarify when to omit completion choice count (#1093) (f0f2c99)
  • explain nested request values and passing request models with **params.to_hash (6670181)