[bot] Google GenAI `models.generateVideos()` (Veo) not instrumented

[braintrust-bot]

Summary

The @google/genai SDK exposes ai.models.generateVideos() for AI-powered video generation using Veo models. This is a distinct generative execution surface — separate from generateContent — and is completely absent from this repo's instrumentation. Users generating video with Veo through the JS SDK get no Braintrust spans.

What's missing

• No channel definitions for generateVideos or any video generation method
• js/src/auto-instrumentations/configs/google-genai.ts only instruments generateContentInternal, generateContentStreamInternal, embedContent/embedContentInternal, and BaseInteractions.create — no video generation entries
• No plugin handler for generateVideos
• No wrapper support: js/src/wrappers/google-genai.ts only intercepts the models property for generateContent/generateContentStream
• js/src/vendor-sdk-types/google-genai.ts — GoogleGenAIModels interface does not declare generateVideos
• e2e/scenarios/google-genai-instrumentation/ — no video generation test scenarios

Upstream SDK surface

ai.models.generateVideos(params) — available in @google/genai SDK (current versions):

ai.models.generateVideos({
  model: "veo-2.0-generate-001",  // or veo-3.0, etc.
  prompt: "...",
  config: {
    aspectRatio: "16:9",
    numberOfVideos: 1,
    durationSeconds: 5,
    // ...
  }
}): Promise<GenerateVideosOperation>

Returns a GenerateVideosOperation (long-running operation). The operation must be polled to completion. This is architecturally distinct from generateContent: it is not streaming, has no token counts, and models are Veo-series (not Gemini text/multimodal models).

Supported GA models include veo-2.0-generate-001 and veo-3.0-generate-001. This is a stable, production API documented on the official Google AI for Developers site.

Relation to existing gaps

This is the Google GenAI equivalent of the OpenAI images.generate() gap (#1628) — a dedicated generative media API that is distinct from the text/chat generation path. Google GenAI image generation (models.generateImages() / Imagen) is tracked in #1673; video generation via Veo is a separate API surface.

Braintrust docs status

not_found — The Braintrust Gemini integration page documents generateContent and generateContentStream as supported methods. generateVideos / Veo is not mentioned.

Upstream sources

• Google AI Veo video generation docs: https://ai.google.dev/gemini-api/docs/video
• @google/genai npm package: https://www.npmjs.com/package/@google/genai
• @google/genai GitHub: https://github.com/googleapis/js-genai
• Veo models: veo-2.0-generate-001, veo-3.0-generate-001
• Long-running operation pattern (returns GenerateVideosOperation, poll with operations.get())

Braintrust docs sources

• https://www.braintrust.dev/docs/guides/tracing
• https://www.braintrust.dev/docs/integrations/ai-providers/gemini

Local repo files inspected

• js/src/auto-instrumentations/configs/google-genai.ts — no generateVideos entry
• js/src/instrumentation/plugins/google-genai-channels.ts — no video generation channels
• js/src/instrumentation/plugins/google-genai-plugin.ts — no video generation handlers
• js/src/wrappers/google-genai.ts — only wraps models for generateContent/generateContentStream
• js/src/vendor-sdk-types/google-genai.ts — GoogleGenAIModels missing generateVideos
• e2e/scenarios/google-genai-instrumentation/ — no video generation scenarios