[bot] Instrument aws-sdk-polly (3,227 weekly downloads)

[braintrust-bot]

Summary

The Braintrust Rust SDK has no instrumentation for aws-sdk-polly, AWS's official Rust SDK crate for Amazon Polly, AWS's text-to-speech (speech synthesis) generative-audio service. Braintrust's own OpenAI integration already documents tracing OpenAI's Speech (text-to-speech) API as a first-class audio-generation capability, but no Braintrust SDK instruments Polly, the AWS equivalent hosted service.

What is missing

aws_sdk_polly::Client exposes these execution-oriented operations (per https://docs.rs/aws-sdk-polly/latest/aws_sdk_polly/client/struct.Client.html):

• synthesize_speech — the core generation call: converts input text or SSML into a synthesized audio stream using a specified voice/engine/output format. Direct analog of OpenAI's audio.speech.create.
• start_speech_synthesis_task — asynchronous speech generation for long-form text, writing output audio to S3.
• start_speech_synthesis_stream — bidirectional streaming synthesis session (text/control events in, audio events out) for low-latency generative-audio use cases (e.g. voice agents).
• get_speech_synthesis_task / list_speech_synthesis_tasks — status/result retrieval for async synthesis jobs.

None of this is wired into the SDK:

1. No usage extraction path — extract_openai_usage()/extract_anthropic_usage() in src/extractors.rs parse a JSON usage object with token counts. Polly's SynthesizeSpeechOutput returns a raw audio Blob (e.g. MP3/PCM/OGG bytes) plus a request_characters header — no token-usage concept at all — so neither extractor applies, and there is no character-based usage/cost extractor for Polly's per-character billing model.
2. No stream/span wrapper — BraintrustStream/wrap_stream_with_span() (src/stream.rs) require an async Stream<Item = Result<Value, E>> deserializing OpenAI Chat-Completions-shaped JSON chunks. Polly's start_speech_synthesis_stream returns a bidirectional audio-event stream with no JSON/choices structure, structurally incompatible with the existing helper.
3. No audio-generation span type — there is no equivalent of a "speech synthesis span" (input text, voice/engine parameters, output audio reference, character count) anywhere in src/types.rs.

Library significance

• Official AWS-maintained crate, part of the aws-sdk-rust monorepo (https://github.com/awslabs/aws-sdk-rust), published by the "AWS SDK Rust Bot"
• Actively released: latest version 1.113.0
• API docs: https://docs.rs/aws-sdk-polly/latest/aws_sdk_polly/
• Crate: https://crates.io/crates/aws-sdk-polly

Weekly downloads

Weekly downloads: 3,227 (as of 2026-08-12; https://crates.io/api/v1/crates/aws-sdk-polly/downloads)

Methodology: exact sum of the crates.io downloads API's daily version_downloads entries (summed across all versions publishing downloads on a given day) plus meta.extra_downloads entries, for the complete 7-day window 2026-08-06 through 2026-08-12 inclusive (checked on 2026-08-17; more recent days were excluded due to crates.io's daily-aggregation lag). Per-day combined totals: 08-06: 502; 08-07: 506; 08-08: 328; 08-09: 173; 08-10: 382; 08-11: 579; 08-12: 757. Sum = 3,227. 90-day recent_downloads reported directly by crates.io: 51,327 (consistent with this weekly figure, ≈3.9K/week average).

Braintrust docs status

supported for the underlying capability class (text-to-speech), but not_found for Polly/AWS specifically. Checked https://www.braintrust.dev/docs/integrations/ai-providers/openai directly on 2026-08-17: the Python span-coverage table lists "Audio APIs (audio captured as an attachment)" as covering "Transcription, Translation, Speech" — confirming Braintrust already treats speech synthesis (OpenAI's audio.speech.create) as a traced, in-scope generative-audio capability. Neither Amazon Polly nor any AWS speech-synthesis service is mentioned on https://www.braintrust.dev/docs/integrations/ai-providers (AWS Bedrock is the only AWS entry listed) or https://www.braintrust.dev/docs/integrations/ai-providers/bedrock.

Relationship to existing issues

• Distinct from [bot] Instrument whisper-rs (58,308 weekly downloads) #84 (whisper-rs): that issue covers local speech-to-text (transcription) execution; this issue covers AWS's hosted text-to-speech (generation) service — the inverse audio-generation direction.
• Distinct from [bot] OpenAI Chat Completions audio output transcript silently dropped from span output #72/[bot] ChatMessage and StreamDelta drop OpenAI audio output (gpt-4o-audio models) #61 (OpenAI Chat Completions audio output / gpt-4o-audio): those cover audio generated as a side-channel of OpenAI's chat/completions responses; Polly is a dedicated, standalone speech-synthesis service with its own request/response shape entirely unrelated to OpenAI's schema.
• Searched this repository's open issues for "polly", "speech synthesis", "text-to-speech", "synthesize_speech" — no existing issue covers this crate.

Upstream sources

• aws-sdk-polly on crates.io: https://crates.io/crates/aws-sdk-polly
• API docs (Client — synthesize_speech, start_speech_synthesis_stream, start_speech_synthesis_task): https://docs.rs/aws-sdk-polly/latest/aws_sdk_polly/client/struct.Client.html
• aws-sdk-rust GitHub monorepo: https://github.com/awslabs/aws-sdk-rust
• crates.io downloads API (raw data used for the weekly-downloads calculation): https://crates.io/api/v1/crates/aws-sdk-polly/downloads

Braintrust docs sources

• OpenAI integration docs (confirms Speech/TTS as a traced capability class): https://www.braintrust.dev/docs/integrations/ai-providers/openai
• AI providers index: https://www.braintrust.dev/docs/integrations/ai-providers
• AWS Bedrock integration docs: https://www.braintrust.dev/docs/integrations/ai-providers/bedrock

Local repo files inspected

• src/extractors.rs — extract_openai_usage() and extract_anthropic_usage() both operate on token-usage JSON; no character-based usage extractor for Polly's Blob-typed audio response
• src/stream.rs — BraintrustStream::aggregate()/wrap_stream_with_span() require a Stream<Item = Result<Value, E>> deserializing OpenAI-shaped chunks; incompatible with Polly's bidirectional audio-event stream
• src/types.rs — no speech-synthesis/audio-generation span fields (voice, engine, output format, character count)
• src/lib.rs — public API exports; no Polly/AWS speech references
• Cargo.toml — no aws-sdk-polly dependency
• Full local codebase search for polly, synthesize_speech, SpeechSynthesis — zero results