Mintplex-Labs/anything-llm
1
0
Fork 0
mirror of https://github.com/Mintplex-Labs/anything-llm.git synced 2026-07-19 22:23:50 -04:00
Code Issues 328 Packages Projects Releases 10 Wiki Activity

[GH-ISSUE #4375] I doubt the accuracy of anythingllm. Why is it not accurate after my test 我怀疑anythingllm的准确度，我测试下来为什么不准确 #2792

New Issue
Closed
opened 2026-02-22 18:31:16 -05:00 by yindo · 1 comment
yindo commented 2026-02-22 18:31:16 -05:00
Owner
Copy Link
Copy Source

Originally created by @kenic123 on GitHub (Sep 10, 2025).
Original GitHub issue: https://github.com/Mintplex-Labs/anything-llm/issues/4375

This is a very discouraging issue. I have deployed anythin-llm. I pulled the code from the ComfyUI and comfyui-wanVideoWrapper repositories into the knowledge base and embedded it into the chat. During this process, I was very cautious and carefully monitored various error messages and resolved them one by one. I chose Deepseek's online API as the large model for chat. When everything was done, I started asking it questions, but its answers were very discouraging. They were completely ambiguous. I confirmed again that the mode of chat was query, and I was quite puzzled I tell it the name of a node to test its accuracy, such as "WANVIDIMAGE._"EMBEDS", its answer was completely self-fabricated. I then gave it the location of the file related to this concept and asked it what the output items of this node were, but it still gave an incorrect answer. So I'm doubting whether RAG technology is suitable for a knowledge base, or perhaps the current technology's accuracy has not yet reached that level. In short, I'm very frustrated. Who can tell me? Where did my train of thought go wrong?
这是一个很丧气issue，我部署好了anythin-llm, 我把ComfyUI和ComfyUI-WanVideoWrapper两个仓库的代码拉到知识库并嵌入到聊天中，这个过程我很小心谨慎的关注各种报错并一一解决，我选择了Deepseek在线的API作为chat的大模型，等这一切都完成了，我开始问它问题，但它的回答让我很丧气，完全是似是而非，我又确认了一下chat的模式是查询，我很奇怪：我告诉它一个节点的名字去测试它的准确性，比如“WANVIDIMAGE_EMBEDS”，它的回答完全是自我杜撰，我再把有关这个概念所在的文件的位置给它，再问它这个节点的输出项有哪些，它也回答不对，所以我在怀疑RAG技术是否适合做知识库，或者说目前技术的精准度尚未达到，总之我很丧气，谁能告诉我，我的思路错在哪里了？

Originally created by @kenic123 on GitHub (Sep 10, 2025). Original GitHub issue: https://github.com/Mintplex-Labs/anything-llm/issues/4375 This is a very discouraging issue. I have deployed anythin-llm. I pulled the code from the ComfyUI and comfyui-wanVideoWrapper repositories into the knowledge base and embedded it into the chat. During this process, I was very cautious and carefully monitored various error messages and resolved them one by one. I chose Deepseek's online API as the large model for chat. When everything was done, I started asking it questions, but its answers were very discouraging. They were completely ambiguous. I confirmed again that the mode of chat was query, and I was quite puzzled I tell it the name of a node to test its accuracy, such as "WANVIDIMAGE._"EMBEDS", its answer was completely self-fabricated. I then gave it the location of the file related to this concept and asked it what the output items of this node were, but it still gave an incorrect answer. So I'm doubting whether RAG technology is suitable for a knowledge base, or perhaps the current technology's accuracy has not yet reached that level. In short, I'm very frustrated. Who can tell me? Where did my train of thought go wrong? 这是一个很丧气issue，我部署好了anythin-llm, 我把ComfyUI和ComfyUI-WanVideoWrapper两个仓库的代码拉到知识库并嵌入到聊天中，这个过程我很小心谨慎的关注各种报错并一一解决，我选择了Deepseek在线的API作为chat的大模型，等这一切都完成了，我开始问它问题，但它的回答让我很丧气，完全是似是而非，我又确认了一下chat的模式是查询，我很奇怪：我告诉它一个节点的名字去测试它的准确性，比如“WANVIDIMAGE_EMBEDS”，它的回答完全是自我杜撰，我再把有关这个概念所在的文件的位置给它，再问它这个节点的输出项有哪些，它也回答不对，所以我在怀疑RAG技术是否适合做知识库，或者说目前技术的精准度尚未达到，总之我很丧气，谁能告诉我，我的思路错在哪里了？
yindo closed this issue 2026-02-22 18:31:16 -05:00
yindo commented 2026-02-22 18:31:17 -05:00
Author
Owner
Copy Link
Copy Source

@timothycarambat commented on GitHub (Sep 12, 2025):

This is how RAG works in AnythingLLM: https://docs.anythingllm.com/chatting-with-documents/rag-in-anythingllm
This is how to modify it to make it better for your use case: https://docs.anythingllm.com/chatting-with-documents/introduction#rag-settings

In all likelyhood, the default embedding model is likely not performing well on your, most likely, Chinese text documents - the default embedder model works best on English text, that is why we offer other models or the ability to use the one you want, like bge

@timothycarambat commented on GitHub (Sep 12, 2025): This is how RAG works in AnythingLLM: https://docs.anythingllm.com/chatting-with-documents/rag-in-anythingllm This is how to modify it to make it better for your use case: https://docs.anythingllm.com/chatting-with-documents/introduction#rag-settings In all likelyhood, the default embedding model is likely not performing well on your, most likely, Chinese text documents - the default embedder model works best on English text, that is why we offer other models or the ability to use the one you want, like bge
yindo changed title from I doubt the accuracy of anythingllm. Why is it not accurate after my test 我怀疑anythingllm的准确度，我测试下来为什么不准确 to [GH-ISSUE #4375] I doubt the accuracy of anythingllm. Why is it not accurate after my test 我怀疑anythingllm的准确度，我测试下来为什么不准确 2026-06-05 14:48:34 -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#2792
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?