Google Docs-Like Editor for PDF/DOCX with LLM Integration for Recommendations

Community
1

I am in the process of developing a web application akin to Google Docs with some a specific use case. The main feature I’m aiming to implement is the ability for users to make edits and annotations similar to what Google Docs offers. However, there’s a specific functionality I need to integrate: the ability to ingest text data into a local large language model (LLM) to generate and suggest edits (similar to the commenting feature in Google Docs).

So basically looking for:

  • Editing Capability: Robust text editing features for PDF and DOCX files, including annotations and layout adjustments.
  • Integration with LLM: Seamless integration to send text to and from a local LLM for real-time editing suggestions.
  • Backend Compatibility: Effective backend integration for document management and processing to handle interactions between the document editor and the LLM.

Are there any known open source/free libraries or tools that support both document editing and that may be easy to integrate with backend services?

So far I have only found tools doing one or the other, primarily for viewing/editing in the application/browser (ONLYOFFICE, WebODF) or just the editing programmatically but nothing to view in (MuPDF)

Thanks!

2

If I understand your question correctly then while this is not an answer because it is not free it might meet the other requirements

As for something that does this I think that most programmers that have the knowledge of how to do this would just do it using two different APIs and do the connection work possibly with a UI hook in the edit application. That is where I would start looking.

3

Sigh, I had a feeling that was the case, Appreciate you

4

Bit of a shameless plug but this is why we built agentoffice.dev. It is the only way (to my knowledge) for LLMs to make edits into DOCX or PDF files. You just have to add an ‘edit_document’ tool your agent then you can use whatever document renderer you like in the frontend. Syncfusion etc.