[FEATURE]: Add databricks serving endpoints as Provider and LLMs

[alaturqua]

Feature hasn't been suggested before.

• I have verified this feature I'm about to request hasn't been suggested before.

Describe the enhancement you want to request

Suggestion to add Databricks AI Gateway - Serving Endpoints as Provider and LLM.

https://docs.databricks.com/aws/en/ai-gateway/

[VJ-yadav]

Hey @kulvirgit @anandgupta42 : I've started working on this and have a working implementation on my fork:

Branch: feat/databricks-provider

What's implemented

• Auth plugin (src/altimate/plugin/databricks.ts) : PAT-based auth with host::token credential format, workspace host validation for AWS/Azure/GCP Databricks instances
• Custom loader in provider.ts — resolves workspace URL from PAT auth or env vars (DATABRICKS_HOST + DATABRICKS_TOKEN)
• 11 foundation models registered — Meta Llama 3.1 (405B/70B/8B), Claude Sonnet/Opus 4.6, GPT-5-4/Mini, Gemini 3.1 Pro, DBRX Instruct, Mixtral 8x7B
• Request transforms — normalizes max_completion_tokens ↔ max_tokens for Databricks OpenAI-compatible API
• 24 unit tests passing — host validation, PAT parsing, body transforms (E2E tests included but skip without credentials)
• TypeCheck passes (only pre-existing ClickHouse driver error)

Follows the same pattern as the Snowflake Cortex provider uses @ai-sdk/openai-compatible SDK since Databricks serving endpoints are OpenAI-compatible.

Next steps — E2E testing

I'm setting up a Databricks workspace to test against real serving endpoints. We use Databricks at my startup (StudentSucceed) so I'll be validating this end-to-end in our environment:

1. PAT auth flow — altimate-code providers login --provider databricks
2. Chat completions — run against foundation models via serving endpoints
3. Streaming — verify SSE streaming works correctly
4. Env var fallback — test DATABRICKS_HOST + DATABRICKS_TOKEN path
5. Screenshots of working CLI output + model responses

Will open the PR once E2E validation is complete. Happy to collaborate with @kulvirgit if there's existing work on this!

[anandgupta42]

Hey @VJ-yadav - Thanks for taking care of this. Please ensure you write some end-to-end tests in the same way we've done for Cortex. Also, update the documentation as well.

[anandgupta42]

Also, @VJ-yadav feel free to reach out to us in the slack community channel if you have any questions - https://bit.ly/agentic-data-engineering-slack

[VJ-yadav]

PR is up: #649

Unit tests passing (22/22), typecheck clean. Currently provisioning a Databricks workspace for E2E validation — will add screenshots of the provider login flow + chat completions against real serving endpoints once the workspace is ready.

@kulvirgit happy to incorporate any feedback or work alongside you on this.