[bot] Azure AI Inference SDK (`@azure-rest/ai-inference`) not instrumented — no tracing for Azure AI Foundry model execution

[braintrust-bot]

Summary

Microsoft publishes a dedicated TypeScript SDK for Azure AI model inference: @azure-rest/ai-inference (part of the Azure SDK for JavaScript). It provides execution APIs for chat completions, streaming, and embeddings for models hosted on Azure AI Foundry (Phi, Mistral, Meta Llama, and others). This package uses a path-based REST client pattern that is architecturally distinct from the openai npm package and is not covered by any existing instrumentation. Users who call @azure-rest/ai-inference directly get no Braintrust spans.

What instrumentation is missing

The @azure-rest/ai-inference package exposes these execution surfaces, none of which are instrumented:

SDK Method	Description
client.path("/chat/completions").post({ body: { messages, ... } })	Chat completions for hosted models
.post({ body: { ..., stream: true } }).asNodeStream()	Streaming chat completions
client.path("/embeddings").post({ body: { input, model } })	Text embeddings

Users instantiate it as:

import ModelClient from "@azure-rest/ai-inference";
import { AzureKeyCredential } from "@azure/core-auth";

const client = ModelClient(endpoint, new AzureKeyCredential(apiKey));

const response = await client.path("/chat/completions").post({
  body: {
    messages: [{ role: "user", content: "Hello" }],
    model: "Phi-4",
  },
});

Auto-instrumentation gap:

The @azure-rest/ai-inference SDK uses a path-based REST client architecture (from @azure/core-rest-pipeline) that is fundamentally different from the openai package's class-method structure. None of the existing configs in js/src/auto-instrumentations/configs/ target any @azure-rest/* module paths. The existing openai instrumentation targets openai module file paths and does not fire for calls made through @azure-rest/ai-inference.

No coverage in any instrumentation layer:

• No auto-instrumentation config for @azure-rest/ai-inference in js/src/auto-instrumentations/configs/
• No wrapper function (e.g. wrapAzureAIInference())
• No channels or plugin for Azure AI Inference methods in js/src/instrumentation/plugins/
• No e2e test scenarios
• Grep for azure-rest, @azure, or ai-inference across js/src/ returns zero matches

Braintrust docs status

unclear — The Braintrust integrations page at https://www.braintrust.dev/docs/integrations lists "Azure AI Foundry" as a supported provider, but the JavaScript SDK documentation does not describe any direct instrumentation for the @azure-rest/ai-inference package.

Upstream references

• @azure-rest/ai-inference source: https://github.com/Azure/azure-sdk-for-js/tree/main/sdk/ai/ai-inference-rest
• npm install: npm install @azure-rest/ai-inference

Local files inspected

• js/src/auto-instrumentations/configs/ — full config list: anthropic.ts, openai.ts, ai-sdk.ts, groq.ts, mistral.ts, cohere.ts, huggingface.ts, google-genai.ts, google-adk.ts, genkit.ts, openai-agents.ts, openrouter.ts, openrouter-agent.ts, claude-agent-sdk.ts, cursor-sdk.ts, github-copilot.ts, openai-codex.ts, flue.ts — no Azure AI Inference config present
• js/src/auto-instrumentations/configs/all.ts — @azure-rest/ai-inference not listed among any instrumentation groups
• Full repo grep for azure-rest, @azure, ai-inference in js/src/ — zero matches
• e2e/scenarios/ — no Azure AI Inference test scenarios