Hacktoberfest 2026:维护者为十月标记出来的 issue,仍然开放、适合新手。 浏览 Hacktoberfest issue

[bot] Google GenAI Live API not instrumented — zero visibility for real-time multimodal sessions

未关闭
#1,732 1 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看

维护者通常 1 天内回复

还没有人认领这个 Issue。

评估

难度
5/5
预计耗时
一周以上
新手友好度
45/100
Issue 类型
功能
描述清晰度
基本清楚
活跃度
冷清
技术栈
typescript
领域
ai, observability

调研方向

首先阅读 js/src/wrappers/google-genai.ts、js/src/auto-instrumentations/configs/google-genai.ts、js/src/instrumentation/plugins/google-genai-channels.ts 和 js/src/instrumentation/plugins/google-genai-plugin.ts,以比较现有的模型 instrumentation。然后检查 js/src/vendor-sdk-types/google-genai.ts 和 e2e/scenarios/google-genai-instrumentation/ 目录。完成的标准是:Live API 各表面都具备针对会话活动、响应、使用量元数据和工具调用的 instrumentation 覆盖及端到端场景。

由索引模型根据 Issue 内容生成。

描述

never-stale

Summary

The @google/genai SDK provides a GA Live API (client.live.connect()) for low-latency, real-time bidirectional audio, video, and text interaction with Gemini models. This repo has zero instrumentation for any Live API surface — no channels, no plugin handlers, no wrapper proxies, and no auto-instrumentation configs. Users building real-time voice or multimodal AI applications with the Google GenAI Live API get no Braintrust spans.

What instrumentation is missing

No coverage in any layer:

  • Wrapper (js/src/wrappers/google-genai.ts): Only intercepts instance.models (line 74–76). No proxy for the live resource.
  • Auto-instrumentation config (js/src/auto-instrumentations/configs/google-genai.ts): Only defines configs for generateContentInternal and generateContentStreamInternal. No config for Live API methods.
  • Channels (js/src/instrumentation/plugins/google-genai-channels.ts): Only two channels defined (generateContent, generateContentStream). No Live API channels.
  • Plugin (js/src/instrumentation/plugins/google-genai-plugin.ts): No handler for Live API calls.
  • Vendor types (js/src/vendor-sdk-types/google-genai.ts): GoogleGenAIClient only declares models property. No live property.

A grep for live.connect, sendRealtimeInput, and sendClientContent across js/src/ returns zero matches.

Key upstream API surfaces with no Braintrust tracing:

SDK Method Description
client.live.connect({ model, config }) Creates a WebSocket session for real-time interaction. Config includes system instructions, tools, voice, response modalities, and speech config.
session.sendClientContent() Sends text content in the session.
session.sendRealtimeInput() Sends real-time audio/video data.
session.sendToolResponse() Sends tool call results back to the model.

The Live API session emits server events including model responses with usage metadata and tool call requests. These contain the same kind of data (content, token counts, tool calls) that is already captured for models.generateContent() in this repo.

Braintrust docs status

not_found — The Braintrust Gemini integration page at https://www.braintrust.dev/docs/integrations/ai-providers/gemini documents generateContent and generateContentStream as supported methods. The Live API is not mentioned.

Upstream reference

  • Google GenAI Live API docs: https://ai.google.dev/gemini-api/docs/live
  • SDK method: client.live.connect({ model, config }) — returns a WebSocket session
  • Capabilities: real-time audio/video/text interaction, tool calling, 70+ language support, interruption handling
  • Session config: systemInstruction, tools, responseModalities, speechConfig, realtimeInputConfig
  • This is a production API documented on the official Google AI for Developers site.
  • The @google/genai SDK (latest v1.48.0+) provides full TypeScript support for the Live API.

Precedent

  • This is the Google equivalent of the OpenAI Realtime API gap.
  • Both represent a new category of AI API (WebSocket-based real-time sessions) that the current instrumentation architecture does not cover.

Local files inspected

  • js/src/wrappers/google-genai.ts — only wraps models property
  • js/src/auto-instrumentations/configs/google-genai.ts — no Live configs
  • js/src/instrumentation/plugins/google-genai-channels.ts — no Live channels
  • js/src/instrumentation/plugins/google-genai-plugin.ts — no Live handlers
  • js/src/vendor-sdk-types/google-genai.ts — no live property on client type
  • e2e/scenarios/google-genai-instrumentation/ — no Live test scenarios
主要语言
TypeScript
星标
29
派生
16
平均合并
1 天 19 小时
30 天内合并 PR
69

环境准备

  • 提供 Dockerfile 或 Docker Compose 文件
  • 没有 Pull Request 模板
  • 没有贡献指南

从这里开始

  1. 先读完整个 Issue,再读项目的贡献指南。
  2. 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
  3. Fork 仓库,在一个分支上完成修改。
  4. 提交 Pull Request,并在描述里引用这个 Issue 编号。

braintrustdata/braintrust-sdk-javascript 的其他 Issue

查看 braintrustdata/braintrust-sdk-javascript 的全部 Issue

相似的 Issue

更多 TypeScript Issue

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。