[GH-ISSUE #3305] Is it a json for each file? #2125

Closed
opened 2026-02-22 18:28:14 -05:00 by yindo · 2 comments
Owner

Originally created by @wangyujiewj on GitHub (Feb 21, 2025).
Original GitHub issue: https://github.com/Mintplex-Labs/anything-llm/issues/3305

Sorry to bother you. I uploaded the file which contains many rules and regulations in AnythingLLM for Retrieval-Augmented Generation, but the model answered poorly. Therefore, I found the json file that the uploaded file was converted to, and I found that all the rules and regulations were converted into one json block. I have set the text block size to 512 in the settings, but AnythingLLM is not generating many json blocks based on this text length. I would like to ask if the text splitting works after the generation of the json file? Does splitting the file into multiple files for uploading help the RAG effect more?

Originally created by @wangyujiewj on GitHub (Feb 21, 2025). Original GitHub issue: https://github.com/Mintplex-Labs/anything-llm/issues/3305 Sorry to bother you. I uploaded the file which contains many rules and regulations in AnythingLLM for Retrieval-Augmented Generation, but the model answered poorly. Therefore, I found the json file that the uploaded file was converted to, and I found that all the rules and regulations were converted into one json block. I have set the text block size to 512 in the settings, but AnythingLLM is not generating many json blocks based on this text length. I would like to ask if the text splitting works after the generation of the json file? Does splitting the file into multiple files for uploading help the RAG effect more?
yindo closed this issue 2026-02-22 18:28:14 -05:00
Author
Owner

@timothycarambat commented on GitHub (Feb 21, 2025):

RAG != full text comprehension. In general LLMs do not need whitespace for text-understanding, so while it may look messy in the text, that is usually okay.

https://docs.anythingllm.com/llm-not-using-my-docs

@timothycarambat commented on GitHub (Feb 21, 2025): RAG != full text comprehension. In general LLMs do not need whitespace for text-understanding, so while it may look messy in the text, that is usually okay. https://docs.anythingllm.com/llm-not-using-my-docs
Author
Owner

@wangyujiewj commented on GitHub (Feb 21, 2025):

Thanks for your answer. I also want to know whether document splitting is done after json file generation. Does generating one or more json blocks have no effect on vector splitting and retrieval?

@wangyujiewj commented on GitHub (Feb 21, 2025): Thanks for your answer. I also want to know whether document splitting is done after json file generation. Does generating one or more json blocks have no effect on vector splitting and retrieval?
yindo changed title from Is it a json for each file? to [GH-ISSUE #3305] Is it a json for each file? 2026-06-05 14:44:41 -04:00
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: Mintplex-Labs/anything-llm#2125