[bot] Instrument AssemblyAI Python SDK (585,463 weekly downloads)

Start by reviewing existing providers under py/src/braintrust/integrations/ and wrappers under py/src/braintrust/wrappers/, then inspect py/pyproject.toml and py/noxfile.py for integration registration and test patterns. Trace the listed AssemblyAI entry points across prerecorded.v2, sync.v1, dictation.v1, and streaming.v3; done means these surfaces are instrumented and covered by the repository's integration test session.

<!-- library-gap-audit: assemblyai-python-sdk -->

Summary

assemblyai is AssemblyAI's official Python SDK. This repo already instruments two adjacent speech/audio execution surfaces — ElevenLabs (TTS generation) and Deepgram (STT transcription, tracked separately as an open gap) — but assemblyai itself has zero instrumentation.

• Current version 1.5.5, actively released
• Not present in py/pyproject.toml [tool.braintrust.matrix], no noxfile.py session, no directory under py/src/braintrust/integrations/ or py/src/braintrust/wrappers/
• Repo-wide grep for assemblyai under py/ returns zero matches

What is missing

The SDK's core execution entrypoints, none of which currently produce a Braintrust span:

• Transcriber.transcribe() / .transcribe_async() / AsyncTranscriber (prerecorded.v2) — the polled-job transcription API for long-form/URL audio, with audio-intelligence features (speaker labels, chapters, etc.)
• SyncTranscriber (sync.v1) — low-latency single-request transcription for short clips
• DictationTranscriber (dictation.v1) — short dictated-note transcription with optional LLM-based rewriting of the transcript, a generative post-processing step distinct from raw transcription
• RealTimeTranscriber / AsyncRealTimeTranscriber (streaming.v3) — live websocket-based streaming transcription (.connect(), .stream(), .disconnect())

None of these call sites are covered by any existing instrumentation in this repo — they do not go through openai, anthropic, or any other already-instrumented client.

Weekly downloads

Weekly downloads: 585,463 (as of 2026-09-21; https://pypistats.org/api/packages/assemblyai/recent)

Braintrust docs status

not_found — checked https://www.braintrust.dev/docs/integrations/ai-providers (lists ElevenLabs as the only speech-related provider; no AssemblyAI) and https://www.braintrust.dev/docs/guides/traces/integrations (no mention of AssemblyAI or transcription SDKs generally).

Upstream sources

• GitHub: https://github.com/AssemblyAI/assemblyai-python-sdk
• PyPI: https://pypi.org/project/assemblyai/ (v1.5.5)
• Docs: https://www.assemblyai.com/docs

Local repo files inspected

• py/src/braintrust/integrations/ — no assemblyai/ directory
• py/src/braintrust/wrappers/ — no AssemblyAI wrapper
• py/pyproject.toml [tool.braintrust.matrix] — no assemblyai entry
• py/noxfile.py — no test_assemblyai session
• Repo-wide case-insensitive grep for assemblyai under py/ — zero matches