[bot] RubyLLM speech synthesis (`RubyLLM.speak`) not instrumented

[braintrust-bot]

Summary

The ruby_llm gem provides RubyLLM.speak(input, ...) for text-to-speech synthesis using AI TTS models. This API is not instrumented by this SDK. The current RubyLLM integration only instruments RubyLLM::Chat — covering chat completions, tool execution, and streaming. speak is a distinct generative AI execution surface (it drives a TTS model to produce audio from text) and produces usage data (token counts via some providers), yet no span is created for it.

This is a direct parity gap with the OpenAI Audio integration gap (#146), which notes that TTS execution via client.audio.speech.create is also uninstrumented. Both use AI models to generate audio from text input.

What is missing

RubyLLM.speak(input, model:, voice:, ...) / RubyLLM::Speech.speak(input, ...) — invokes a TTS provider (OpenAI tts-1, tts-1-hd, or equivalent) and returns a RubyLLM::Speech object containing:

• data — raw audio bytes
• model — model ID used for synthesis
• voice — voice selected
• format — audio format (mp3, wav, flac, etc.)

What a span should capture

• Input: the text string passed to speak
• Metadata: model, voice, format, provider
• Output: audio format and size (not the raw binary bytes)
• Metrics: time_to_first_token (latency)

The implementation pattern already exists: RubyLLM::Embed, RubyLLM::Paint, and RubyLLM::Transcription are single-method execution surfaces (see issue #136). RubyLLM::Speech follows the same shape and would be instrumented with the same approach: a new SpeechPatcher wrapping the speak class method.

Braintrust docs status

not_found — The Braintrust RubyLLM integration docs at https://www.braintrust.dev/docs/integrations/sdk-integrations/ruby-llm document only chat completions, tool calls, token usage, and streaming. Speech synthesis is not mentioned.

Upstream sources

• RubyLLM Speech class: https://github.com/crmne/ruby_llm/blob/main/lib/ruby_llm/speech.rb
• RubyLLM module speak method: https://github.com/crmne/ruby_llm/blob/main/lib/ruby_llm.rb
• RubyGems (ruby_llm): https://rubygems.org/gems/ruby_llm — "Features chat (text, images, audio, PDFs), image generation, embeddings, tools (function calling)…" (audio includes TTS)

Local repo files inspected

• lib/braintrust/contrib/ruby_llm/patcher.rb — defines only ChatPatcher; no SpeechPatcher
• lib/braintrust/contrib/ruby_llm/instrumentation/chat.rb — wraps Chat#complete and Chat#execute_tool only
• lib/braintrust/contrib/ruby_llm/integration.rb — registers [ChatPatcher] only
• Appraisals — tests ruby_llm at versions 1.8, 1.9, and latest; speech.rb is present in all supported versions
• Grep for speak, Speech, tts, SpeechPatcher across lib/braintrust/ returns zero matches