Hey I m trying to figure out how to use

mmm not by default. You'd either have to stuff the prompt with the previous messages, write a custom selector that uses the previous messages, or use a data agent setup with the router engine as a tool

Hey, thanks. I guess stuffing the history to the prompt would be easiest, but what do you think would be better?

Probably best to use a chat engine / data agent that uses the query engine.

You could use an OpenAIAgent with the existing router engine

Or if you switch to a retriever router, a context chat engine might also work 👀

Great, thanks a lot for the advice 🙂

Hey Logan, I'm trying to understand where and how can I change the llama_index_starter_pack react demo to use Azure instead of OpenAI. Can you help me with that?

Yea, there's an azure guide here

https://gpt-index.readthedocs.io/en/stable/examples/customization/llms/AzureOpenAI.html

And a more detailed on here
https://gpt-index.readthedocs.io/en/stable/examples/llm/azure_openai.html

Thanks! is there a more updated fullstack demo by any chance? I'm really struggeling to make it fit to the current llamaindex version

ah yea, it's a little old 😅

What part was tripping you up?

haha yeah.. basically what I want to achieve is exactly the server of the demo, only updated and using Azure 🙂

Hmm, following that guide above should help no? Setup the llm and embed model, set the global service context, and off you go 🙏

Yeah I'm gonna try it this week. Definitely I will bug you about it 😆

You should come visit Israel, maybe do a webinar 😉