[bot] Add instrumentation wrapper for async-openai client

[braintrust-bot]

Summary

The Braintrust Rust SDK has no wrapper for async-openai, the dominant Rust client for the OpenAI API. Every other Braintrust SDK provides an OpenAI client wrapper (wrapOpenAI in TypeScript, wrap_openai in Python, traceopenai.NewMiddleware in Go, BraintrustOpenAI.wrapOpenAI in Java, BraintrustOpenAI.WrapOpenAI in .NET, Braintrust.instrument!(:openai) in Ruby), but the Rust SDK has no equivalent.

What is missing

A wrap_openai-style function that takes an async-openai Client and returns a traced version that automatically creates spans for:

• Chat completions — client.chat().create() and client.chat().create_stream()
• Responses API — client.responses().create() and streaming variant
• Embeddings — client.embeddings().create()

The wrapper should capture inputs, outputs, model name, token usage, latency, and tool calls in Braintrust spans, matching the behavior of the OpenAI wrappers in other Braintrust SDKs.

The SDK already has building blocks that could support this:

• extract_openai_usage() in src/extractors.rs parses OpenAI usage fields
• wrap_stream_with_span() in src/stream.rs attaches spans to async streams
• SpanBuilder / SpanHandle in src/span.rs for manual span creation

But there is no integration that wires these together around async-openai client calls.

Library significance

• 4.3M+ total downloads on crates.io, ~2M recent downloads
• 106 published versions, latest 0.35.0 (April 2026)
• By far the most adopted Rust OpenAI client
• Covers: chat completions, responses API, embeddings, audio, images, assistants, fine-tuning, streaming, and more
• Official docs: https://docs.rs/async-openai
• GitHub: https://github.com/64bit/async-openai
• Crate: https://crates.io/crates/async-openai

Braintrust docs status

supported (in other languages) — Braintrust documents OpenAI client wrapping for TypeScript, Python, Ruby, Go, Java, and .NET on the Trace LLM calls page and the OpenAI integration page. Rust is not listed as a supported language for OpenAI wrapping.

Upstream sources

• async-openai docs: https://docs.rs/async-openai
• async-openai GitHub: https://github.com/64bit/async-openai
• async-openai on crates.io: https://crates.io/crates/async-openai
• Braintrust Trace LLM calls docs: https://www.braintrust.dev/docs/instrument/trace-llm-calls
• Braintrust OpenAI integration docs: https://www.braintrust.dev/docs/integrations/ai-providers/openai

Local files inspected

• src/extractors.rs — extract_openai_usage() parses OpenAI usage fields but is not wired to any client wrapper
• src/stream.rs — wrap_stream_with_span() exists but is a generic stream helper, not an async-openai integration
• src/span.rs — span creation infrastructure exists
• src/lib.rs — public API exports; no async-openai references
• Cargo.toml — no async-openai dependency
• Full codebase search for "async-openai", "async_openai", "openai::Client" — zero results