genAI context grounding

Help Studio
, ,
1

I have a number of long documents from each of these I need to ask specific questions, I don’t want to lump them all in the same storage bucket as I think that’ll introduce confusion as these are all long contracts and wordings are sort of similar. but I also don’t want to create a bucket per doc, please advise what is the best way handling this, thanks in advance

2

@Lynn_Song,

As per my understanding you can keep everything on single bucket and give the context in your prompt to the model that the result should be from which document. Make sure your documents have unique identification like SOW Contract or NOC Contract etc.

I haven’t tried this but I feel it’s feasible with the prompt.

3

Thanks for your help! In langchain unless I explicitly include an identifier during ingestion and indexing, the link between the original document and the retrieved data is lost. but I’m not sure how to include those when putting them in the bucket?

4

@Lynn_Song

  1. Wither go with creating separate
  2. Have all docs in one storage and then based on the query read the relavant do and use update context and update the singe point of context with relavant info and then allow the query…this might require you to update the context…based on document you are searching

Cheers

5

Thanks, where can I add metadata to each doc if I add them all in the same bucket? I don’t see anywhere either in the activities or in the ai trust center?

6

@Lynn_Song,

Let me try this and get back to you!