[bot] Instrument RunwayML (56,353 weekly downloads)

[braintrust-bot]

Summary

Runway publishes an official Node.js/TypeScript SDK, @runwayml/sdk on npm (latest v4.17.0, GitHub: runwayml/sdk-node), providing task-based execution APIs for generative video and image models (Gen-4, Gen-4 Turbo, Gemini image models hosted on Runway's platform). This repository has zero instrumentation for any Runway SDK surface — no wrapper, no diagnostics channel, no plugin, no auto-instrumentation config, no e2e scenario. This is a clear media-generation execution gap of the same kind already tracked for other providers in this repo's history (e.g. OpenAI images.generate(), Google GenAI generateImages(), HuggingFace textToImage()).

What instrumentation is missing

The @runwayml/sdk npm package exposes these execution surfaces, none of which are instrumented:

SDK Method	Description
client.imageToVideo.create({ model, promptImage, promptText, ratio })	Generates video from a source image + prompt — returns an async task
client.textToImage.create({ model, promptText, ratio })	Generates an image from a text prompt — returns an async task
client.textToSpeech.create(...) / client.videoToVideo.create(...)	Additional generative task types (speech synthesis, video restyling)
task.waitForTaskOutput()	Convenience polling helper that awaits and returns the generation result, or throws TaskFailedError/TaskTimedOutError
client.tasks.retrieve(task_id)	Retrieves status/output of a previously started generation task

Every generation call returns a task record (id) that is polled to completion — a clear, well-bounded execution unit suitable for span-level tracing (start on create, end on task completion via waitForTaskOutput/retrieve).

No coverage in any instrumentation layer:

• No wrapper function (e.g. wrapRunwayML()) in js/src/wrappers/
• No diagnostics channels or plugin in js/src/instrumentation/plugins/
• No auto-instrumentation config in js/src/auto-instrumentations/configs/ (not listed in all.ts)
• No e2e test scenario under e2e/scenarios/
• Full repo grep for runway (case-insensitive) returns zero matches anywhere in the repository

Weekly downloads

Weekly downloads: 56,353 (as of 2026-08-31; window 2026-08-23 to 2026-08-29; https://api.npmjs.org/downloads/point/last-week/@runwayml/sdk)

Braintrust docs status

not_found. Runway / RunwayML does not appear anywhere on the Braintrust integrations index — https://www.braintrust.dev/docs/guides/tracing/integrations — which lists Tracing integrations, Agent frameworks, Eval frameworks, and Testing frameworks. There is no Runway entry in any category and no gateway/proxy workaround documented.

Upstream references

• @runwayml/sdk npm package: https://www.npmjs.com/package/@runwayml/sdk
• Official SDK GitHub: https://github.com/runwayml/sdk-node
• Runway API docs (SDKs): https://docs.dev.runwayml.com/api-details/sdks/
• Runway API product page: https://runwayml.com/api

Braintrust docs sources checked

• https://www.braintrust.dev/docs/guides/tracing/integrations

Local files inspected

• js/src/wrappers/*.ts — full directory listing — no runway.ts/runwayml.ts wrapper
• js/src/instrumentation/plugins/*.ts — full directory listing — no Runway channels or plugin file
• js/src/auto-instrumentations/configs/all.ts — full config group list (openai, anthropic, bedrock, ai-sdk, claude-agent-sdk, cloudflare-*, cursor-sdk, openai-agents, google-genai, huggingface(+transformers), langchain/langgraph, langsmith, openrouter(+agent), mistral, ollama, google-adk, cohere, groq, genkit, github-copilot, pi-coding-agent, strands-agent-sdk, flue, voyageai) — no Runway entry
• e2e/scenarios/ — no Runway test scenario
• Full repo grep for runway (case-insensitive) — zero matches