Using Vertex AI RAG to Call a Real-time API instead of a VASC Datastore

[ByronOfInnova]

I almost continued this post, but decided to start fresh.

https://www.googlecloudcommunity.com/gc/AI-ML/Struggling-to-build-a-simple-RAG-solution/m-p/705173

The solution to the post referenced creates a VASC datastore for the RAG call, but what I'd like to do is use parameters collected in an conversation to call an externally hosted service that returns information for my RAG (e.g., lookup an auto insurance policy based upon policy number).

The datastore approach to me implies that it's batch and static until the datastore is updated.

The API call implies a dynamic call as part of the conversation session.

I haven't found a good tutorial on doing this yet and I'm waiting for my Partner Advantage account to get updated with access to Qwiklabs so I can do the Vertex AI labs which I hope cover this use case.

Does anyone have experience implementing RAG like this and can you point me to examples/training?

[ByronOfInnova · Accepted solution]

I just watched this video with Kristopher Overholt.

Building and Deploying AI Agents with LangChain on Vertex AI
(URL Removed by Staff)
This video answered a lot of questions for me.  

Providing developers with a streamlined but open way of interacting using Reasoning Engine is very exciting.
The video was great and I'm looking forward to digging into the notebooks, etc.