Create API from visual RAG

lji

I'm using visual recipes to create a simple RAG system. Can I create API from my RAG set up?

Operating system used: Linux

DarinB

Hi @lji,

There may be a couple of ways to accomplish what you want.

One way, and maybe the easiest), is to use the augmented LLM that you've already created in your simple RAG setup. Here's a tutorial called Perform Completion Queries on LLMs. This example shows you how to find and use your augmented LLM programmatically with Python.

Another way to go about this could be accomplish using the Dataiku Answers API. Dataiku Answers is a chat UI built by Dataiku for scalable conversational applications. This approach requires a few more steps than previous one:
1. Set up your RAG (which you've already done).
2. Install Dataiku Answers (I believe you need Dataiku 12.5 or later) via the plugin store.
3. Configure your Dataiku Answers chat application (documentation).
4. Enable and use the Dataiku Answers API (documentation).
5. Create and expose a Python function that sends queries to your chat application as an API service (documentation).

Using Dataiku Answers and the Dataiku Answers API offer a couple of advantages: you get a webapp without having to write any code and it includes a conversation history dataset (you'd have to build this yourself).