[bot] Missing instrumentation for Databricks SDK for Java (Model Serving execution)

[braintrust-bot]

Summary

The Braintrust Java SDK does not instrument the official Databricks SDK for Java (com.databricks:databricks-sdk-java). Databricks is already listed by Braintrust as a supported "cloud platform" for model access, but this repo has no instrumentation for the SDK's ServingEndpointsAPI, which is the execution surface Java applications use to invoke Databricks Model Serving / Foundation Model API endpoints (chat completions, text completions, and embeddings hosted or proxied by Databricks).

What is missing

The com.databricks:databricks-sdk-java SDK provides this execution-oriented surface, not instrumented anywhere in this repo:

API	Class	Description
query()	ServingEndpointsAPI / ServingEndpointsService	Invokes a Databricks Model Serving endpoint — covers chat-completions-style requests, legacy completions, and embeddings, including Databricks Foundation Model APIs (e.g. served Llama/DBRX/GTE models) and external-model endpoints proxying OpenAI/Anthropic/etc.
Streaming query responses	ServingEndpointsAPI	Server-sent-event style incremental chat responses from a serving endpoint

This is a genuine model-execution surface (request/response over a served model), distinct from the admin/CRUD operations the same SDK also exposes for creating or managing endpoints (out of scope by this audit's own criteria).

Why existing instrumentation does not cover this

• None of the six existing instrumentation modules (openai_2_15_0, anthropic_2_2_0, genai_1_18_0, langchain_1_8_0, springai_1_0_0, aws_bedrock_2_30_0) reference Databricks classes or the com.databricks package.
• The Databricks SDK uses its own HTTP client/transport layer, entirely separate from the OpenAI/Anthropic/Google/AWS clients this repo currently wraps, so no existing module incidentally covers it.

Braintrust docs status

• Databricks is listed as a supported cloud platform at https://www.braintrust.dev/docs/integrations/ai-providers (models accessible through it), but there is no mention of Java SDK-level instrumentation, or of ServingEndpointsAPI/model-serving execution tracing, anywhere in Braintrust's docs: unclear (provider is nominally supported at the proxy/config level, but code-level Java instrumentation of the official SDK's execution calls is not_found)

Upstream sources

• Serving endpoints query API reference: https://docs.databricks.com/api/workspace/servingendpoints/query
• Query a chat model guide: https://docs.databricks.com/aws/en/machine-learning/model-serving/query-chat-models
• Java SDK repository: https://github.com/databricks/databricks-sdk-java
• Java SDK docs: https://docs.databricks.com/aws/en/dev-tools/sdk-java
• Javadoc for ServingEndpointsAPI: https://javadoc.io/doc/com.databricks/databricks-sdk-java/latest/com/databricks/sdk/service/serving/ServingEndpointsAPI.html
• Maven Central: com.databricks:databricks-sdk-java — actively released

Local repo files inspected

• settings.gradle (lines 28–33) — lists all instrumentation modules; no Databricks module present
• braintrust-sdk/instrumentation/ — contains openai_2_15_0, anthropic_2_2_0, genai_1_18_0, langchain_1_8_0, springai_1_0_0, aws_bedrock_2_30_0; no Databricks directory
• Full-repo grep for databricks (case-insensitive) — zero matches