Hi everyone I have one doubt if we are using one NLP based search, do we need to do any extra step while loading the document into the typesense server. Thanks

Is NLP = Natural Language Search? If yes, then no additional step. It doesn't use embeddings, it just send your search to a LLM to generate a typesense query.

Thanks Alan I am feeding example intents to typesense And then using it to classify the user intent. When I am using NLP directly without embedding, I am not getting good results. What would be the best architecture for classification

Maybe the example might be messing it. You don't need a few-shot approach. It should work by just writing "give me cheap wines" and similar,since we already do the prompting on our side.

Thanks Alan So what do you suggest me to feed into the typsesense For example, I am 4 intents A,B,C,D with their respective fields I want user's query to be classifed into one of them, would simply running it NLP query works?

Vaibhav, could you expand on what you mean with "intent"? If you have 4 categories for example, and the llm query must add at least one of these categories, it's better to only have these 4 values as possible categories

Sorry for using the jargon, So lets take a case where there are 4 actions that can be done from frontend, User types her query, and we need to classify to her query belongs to which action. I am using NLP for it, what according to you is the best way for it. Thanks

I am using NLP for it, what according to you is the best way for it.

Could you share the documentation were you read it?

Typesense will add the parameters, choosing among all the available options. This will work for a field named "Category" for example, but if you want the user to click a button and it discover if the user wants to insert, delete, update and so on, you would need a hacky approach

Thanks Alan