Mintplex-Labs/anything-llm
1
0
Fork 0
mirror of https://github.com/Mintplex-Labs/anything-llm.git synced 2026-07-19 14:13:35 -04:00
Code Issues 328 Packages Projects Releases 10 Wiki Activity

[GH-ISSUE #4556] Poor results with embedding of documents #2895

New Issue
Closed
opened 2026-02-22 18:31:43 -05:00 by yindo · 1 comment
yindo commented 2026-02-22 18:31:43 -05:00
Owner
Copy Link
Copy Source

Originally created by @thomasbrunner-spec on GitHub (Oct 17, 2025).
Original GitHub issue: https://github.com/Mintplex-Labs/anything-llm/issues/4556

Dear Community,

I use AnythingLLM with ollam and the following models on a Mac:
gpt-oss120:b
Qwen3-Embedding-8B:F16

Chunking 1000, overlap 100

When I upload a document (PDF) into my workspace (example 10 pages), it seem that significant information is missing. Acutally it is a contract to be analysed and it contains regulations for termination. If I prompt to list the regulations during the contract analysis, the answer is that this information can not be found in the document.

Did anybody make similar experiences? How can I fix or improve this issue?

Thanks

Thomas

Originally created by @thomasbrunner-spec on GitHub (Oct 17, 2025). Original GitHub issue: https://github.com/Mintplex-Labs/anything-llm/issues/4556 Dear Community, I use AnythingLLM with ollam and the following models on a Mac: gpt-oss120:b Qwen3-Embedding-8B:F16 Chunking 1000, overlap 100 When I upload a document (PDF) into my workspace (example 10 pages), it seem that significant information is missing. Acutally it is a contract to be analysed and it contains regulations for termination. If I prompt to list the regulations during the contract analysis, the answer is that this information can not be found in the document. Did anybody make similar experiences? How can I fix or improve this issue? Thanks Thomas
yindo closed this issue 2026-02-22 18:31:43 -05:00
yindo commented 2026-02-22 18:31:43 -05:00
Author
Owner
Copy Link
Copy Source

@timothycarambat commented on GitHub (Oct 17, 2025):

Embedding uses chunks of the information for semantic retrieval. The downside of this is the semantics do not perform well when you a comprehensive question. For example, regulations may be 3-5 chunks in the entire set. So in this case you might want to increase the # of chunks, or enable reranking, or both to improve retrieval for documents too large to be fully injected.

@timothycarambat commented on GitHub (Oct 17, 2025): Embedding uses chunks of the information for semantic retrieval. The downside of this is the semantics do not perform well when you a _comprehensive_ question. For example, regulations may be 3-5 chunks in the entire set. So in this case you might want to increase the[ # of chunks](https://docs.anythingllm.com/chatting-with-documents/introduction#vector-database-settings--max-context-snippets), or [enable reranking](https://docs.anythingllm.com/chatting-with-documents/introduction#vector-database-settings--search-preference-reranking), or both to improve retrieval for documents too large to be fully injected.
yindo changed title from Poor results with embedding of documents to [GH-ISSUE #4556] Poor results with embedding of documents 2026-06-05 14:49:08 -04:00
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
Desktop
Docker
Integration Request
Integration Request
OS: Linux
OS: Mobile
OS: Windows
UI/UX
blocked
bug
bug
core-team-only
documentation
duplicate
embed-widget
enhancement
feature request
github_actions
good first issue
investigating
needs info / can't replicate
possible bug
pull-request
Mirrored from GitHub Pull Request
 question
stage: specifications
wontfix
No Label
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
yindo
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: Mintplex-Labs/anything-llm#2895
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?