boundary

I would like to use [google's RAG](https://docs.cloud.google.com/vertex-ai/generative-ai/docs/model-reference/rag-output-explained) and [Grounding with google search](https://docs.cloud.google.com/vertex-ai/generative-ai/docs/model-reference/grounding) when calling Gemini 3 on vertex. Has anyone else done this with BAML, can you gice me some pointers?

I see a thread where <@99252724855496704> reverse engineered the google-ai api for google search. I guess this is different for the vertex api (joy). Also, I'm interested in capturing structured output about the RAG responses and Grounding / Search results too.

to capture any raw data you can use the Collector object

since that will give you all the http call results

I'm also not sure how to even use the _vertex_ API in BAML and leverage the RAG tool?

oh I see, so the RAG API is a unique vertex api that is not related to LLM generations

so unfortunately BAML does not support it. But you can run some python code to call that api, and inject the results into your BAML function