[bot] OpenAI Videos API (Sora video generation) not instrumented

[braintrust-bot]

Summary

The OpenAI Ruby SDK exposes a stable videos resource for Sora video generation (client.videos.create, edit, extend_, remix). None of these generative execution methods are instrumented by the Braintrust Ruby SDK. This is the same class of gap as the already-filed Images API issue (#147), but for video generation.

What is missing

The official openai gem provides OpenAI::Resources::Videos with four generative execution methods:

• client.videos.create(prompt:, model:, seconds:, size:, ...) — Initiates a video generation job from a text prompt. Uses Sora-2 or Sora-2-pro models. Returns OpenAI::Models::Video with a job ID and status.
• client.videos.edit(video_id:, prompt:, ...) — Edits an existing video using a text prompt.
• client.videos.extend_(video_id:, ...) — Extends a completed video with additional generated segments.
• client.videos.remix(video_id:, prompt:, ...) — Remixes a completed video using a new prompt.

None of these are patched. The existing openai patcher (lib/braintrust/contrib/openai/patcher.rb) registers ChatPatcher, ResponsesPatcher, and ModerationsPatcher only.

What a span should capture

Video generation spans (client.videos.create) should capture:

• Input: the text prompt
• Metadata: model (sora-2, sora-2-pro), seconds (clip duration), size (resolution), provider (openai), endpoint (/v1/videos/generations)
• Output: the returned Video object's id and status
• Metrics: latency of the create call

Video editing spans (edit, remix) follow the same pattern with an additional video_id input.

Note: The Videos API is async/job-based — create returns immediately with a job ID; video generation completes asynchronously. Instrumentation of create captures the initiation; retrieve polls are CRUD and out of scope.

Braintrust docs status

not_found — The Braintrust OpenAI integration docs at https://www.braintrust.dev/docs/integrations/ai-providers/openai and https://www.braintrust.dev/docs/guides/tracing do not mention video generation or the Sora API. All documented examples cover chat completions and audio.

Upstream sources

• Official OpenAI Ruby SDK Videos resource: https://github.com/openai/openai-ruby/blob/main/lib/openai/resources/videos.rb — exposes create, edit, extend_, remix, retrieve, list, delete, download_content
• OpenAI Sora-2 model docs: https://developers.openai.com/api/docs/models/sora-2
• OpenAI API reference — Video generation: https://platform.openai.com/docs/api-reference/video-generation

Local repo files inspected

• lib/braintrust/contrib/openai/patcher.rb — defines ChatPatcher, ResponsesPatcher, ModerationsPatcher; no VideosPatcher
• lib/braintrust/contrib/openai/instrumentation/ — contains chat.rb, responses.rb, moderations.rb, common.rb; no videos.rb
• Grep for video, sora, Videos across lib/braintrust/ returns zero matches
• Issue [BOT ISSUE] OpenAI Images API not instrumented (openai + ruby-openai gems) #147 (OpenAI Images API not instrumented) demonstrates the equivalent existing gap for images — the same pattern applies here for video generation