[feat] using tools + embeddings to handle RAG

[c-lamont]

This is mainly a question which could be a feature request.

Is it possible to provide some context with the users chat?

What I mean with that is; let's say they are looking at a recipe in my application, and they want to ask the question "Is this a healthy meal?". I would then like to provide the recipe, along with the question to the AI.

Thanks for any help.

[csells]

you bet. that's called RAG (Retrieval Augmented Generation) and you can do that via the messageSender method of the LlmChatView. There should be a demo of how that works. I'll put it onto the list.

[csells]

You can see the mechanics in the logging sample.

[c-lamont]

you bet. that's called RAG (Retrieval Augmented Generation) and you can do that via the messageSender method of the LlmChatView. There should be a demo of how that works. I'll put it onto the list.

So to be clear, I would include the recipe as an attachment, and every question the user would ask the attachment will be included in the sendMessageStream?

[csells]

You could but I wouldn't recommend it. The R in RAG means that you process each prompt that the user provides and search your app's data store to find the appropriate data. The data your search turns up is the data you send.

[c-lamont]

Okay, so when you say The data your search turns up is the data you send how exactly do I send this data? Do I append it in the prompt?

[csells]

You have your choice of updating the user's prompt and optionally adding an attachment.

[c-lamont]

Okay thanks for your help, and thanks for your work on this package, I think it could be a great addition to our app.

[dietrichson]

@csells how about function calls?

[csells]

how about function calls?

@dietrichson that's another feature request, but like a number of other features, outside the scope of v1. if you like, you can open another issue to track this for v2, but it's already on the v2 list.

[csells]

Okay thanks for your help, and thanks for your work on this package, I think it could be a great addition to our app.

Great! Happy to hear it.