LibreChat icon indicating copy to clipboard operation
LibreChat copied to clipboard

Published 20 hours ago • verified publisher icon danny-avila

Enhancement: Drag or paste documents to provide context

Open eburnette opened this issue 1 year ago • 4 comments
trafficstars

Contact Details

Redacted

What features would you like to see added?

Note this is different from RAG and file upload. I just want it to work like Anthropic's claude site where I can drag files (pdf, txt, etc.) into the chat and have it included in the context. In addition, if I paste a large amount of text in there, it should treat it like I dragged a text file in (see the screenshot).

More details

Both claude and chatgpt let you drag files into the chat window to provide additional context. It doesn't perform embedding or vector search on them, it just provides additional context. The experience should be identical to adding images.

Which components are impacted by your request?

General

Pictures

image

Code of Conduct

  • [X] I agree to follow this project's Code of Conduct

eburnette avatar Dec 10 '23 20:12 eburnette

I've got the same needs.

UPman24 avatar Dec 18 '23 01:12 UPman24

I've got the same needs.

Actually, dragging and dropping isn't necessary. As long as I can upload files like PDFs, Word docs, and PowerPoints, that works for me too.

UPman24 avatar Dec 18 '23 01:12 UPman24

Would love to see RAG options included.

ewebgh33 avatar Jan 10 '24 07:01 ewebgh33

Would love to see RAG options included.

RAG is planned as well as the concern of this issue (adding files to the context, not vectorizing them)

danny-avila avatar Jan 10 '24 13:01 danny-avila

@danny-avila, I'm trying on adding vector database plugins, like connect Pinecone index. Any suggestions? Is this feature on your roadmap by any chance ?

iiitmahesh avatar Mar 11 '24 16:03 iiitmahesh

@danny-avila, I'm trying on adding vector database plugins, like connect Pinecone index. Any suggestions? Is this feature on your roadmap by any chance ?

Just released RAG API to the project: https://github.com/danny-avila/rag_api

I will close this issue today as I'll add an option to load the full context of the document

danny-avila avatar Mar 21 '24 12:03 danny-avila

Closed by https://github.com/danny-avila/LibreChat/pull/2169

  • Adds experimental new variable RAG_USE_FULL_CONTEXT which will retrieve the full document and add it to the prompt instead of using semantic retrieval.
RAG_USE_FULL_CONTEXT=true

There is still a frontend option yet to be implemented to allow toggling this behavior from a user level instead of from an environment variable.

danny-avila avatar Mar 23 '24 21:03 danny-avila