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 #5294] [FEAT]: Exclude "@agent" which does not work, time to improve usage #5004

New Issue
Closed
opened 2026-06-05 14:51:26 -04:00 by yindo · 2 comments
yindo commented 2026-06-05 14:51:26 -04:00
Owner
Copy Link
Copy Source

Originally created by @computerworxdev on GitHub (Mar 28, 2026).
Original GitHub issue: https://github.com/Mintplex-Labs/anything-llm/issues/5294

What would you like to see?

Issue with using "@agent" in a prompt - to be honest not Inference apps use that approach looks like 5 years ago people used to do that. It's inconvenient , simply useless, not working with activated agents, and prompt and continues error about not being able to do something when prompted when enabled. Simply it's frustrating.
I deleted this app and reinstalled OSX system due to an update. I have not installed it back yet ! I'm waiting for the app to grow up.

However there are some feature I like and wanted to give a feedback for developers to improve it so it could be #1 app.

1 - Allow to create Folders to group Chats lets say per Category .
2 - get rid off useless "@agent" - AnythingLLM should be smart by now to detect enabled extensions/plugions and use them in a native language. It's even hard to speak via microphone @agent and it does not work, so typing and speaking is very inconvenient - also calling out agent fails. It should know when Agents are enabled and use them natively or say responses in human friendly way that it can not be done and make sure ability is enabled.

No one else is using chat communication in that old fashion difficult way. "@agent"

Please resolve this in coming updates so I can come back to it.

3 - you tube videos titles "search anything. online" is focused on authenticated site and search. Not searching anything. It should be implemented by default in the app and allow to search anything , this is a must now days for any AI Rag app !. Please improve that right now its so hard to use and search anything online using AnythingLLM

Originally created by @computerworxdev on GitHub (Mar 28, 2026). Original GitHub issue: https://github.com/Mintplex-Labs/anything-llm/issues/5294 ### What would you like to see? Issue with using "@agent" in a prompt - to be honest not Inference apps use that approach looks like 5 years ago people used to do that. It's inconvenient , simply useless, not working with activated agents, and prompt and continues error about not being able to do something when prompted when enabled. Simply it's frustrating. I deleted this app and reinstalled OSX system due to an update. I have not installed it back yet ! I'm waiting for the app to grow up. However there are some feature I like and wanted to give a feedback for developers to improve it so it could be #1 app. 1 - Allow to create Folders to group Chats lets say per Category . 2 - get rid off useless "@agent" - AnythingLLM should be smart by now to detect enabled extensions/plugions and use them in a native language. It's even hard to speak via microphone @agent and it does not work, so typing and speaking is very inconvenient - also calling out agent fails. It should know when Agents are enabled and use them natively or say responses in human friendly way that it can not be done and make sure ability is enabled. No one else is using chat communication in that old fashion difficult way. "@agent" Please resolve this in coming updates so I can come back to it. 3 - you tube videos titles "search anything. online" is focused on authenticated site and search. Not searching anything. It should be implemented by default in the app and allow to search anything , this is a must now days for any AI Rag app !. Please improve that right now its so hard to use and search anything online using AnythingLLM
yindo added the enhancementfeature request labels 2026-06-05 14:51:26 -04:00
yindo closed this issue 2026-06-05 14:51:26 -04:00
yindo commented 2026-06-05 14:51:27 -04:00
Author
Owner
Copy Link
Copy Source

@timothycarambat commented on GitHub (Mar 28, 2026):

get rid off useless @agent

Its not useless, in fact, it makes a lot of sense for people running super small models. Trying to use tools with super small models for every call can ruin the experience dramatically and for people who do that, they may prefer it be opt-in. That being said, what youre asking for already exists via the automatic mode. Using that chat mode would solve your issues entirely - especially the mic one!

If it is worth explaining why it even exists, it was because when AnythingLLM started LLama2 was barely a thing and tool calling didnt even yet exist! By the time it did, local models were still at best 8K, so bloating the window with tool calls where just asking simple QA prompts would spin the model off grabbing tools for no reason!

Luckily, this is no longer the major case so now we have automatic mode, which is about to also be the default for all new workspaces so that complaint should go away totally!

Note

Automatic mode is very recent and not in the desktop app yet but will be in the next patch sometime next week!

Allow to create Folders to group Chats lets say per Category .

I don't see how this benefit extends beyond what workspaces already do? Can you explain some more so I can understand what a UI/UX for that looks and feels like?

you tube videos titles "search anything. online" is focused on authenticated site and search. Not searching anything. It should be implemented by default in the app and allow to search anything , this is a must now days for any AI Rag app !. Please improve that right now its so hard to use and search anything online using AnythingLLM

If I understand the question you are saying how the web-search skill should be enabled by default so that you can get web resources without having to toggle it on? I have considered this for sure, web-scraping is default but obviously that cannot "search" like web-search can - which can be annoying I understand.

That being said, it is not a toggle in tools - so it's easier to flip on

Image

This is all great feedback and clearly thought out so help me understand the bits I am missing so that we can incorporate this - I just dont want to make assumptions on what youre looking for so we are always building the right stuff.

<!-- gh-comment-id:4149180953 --> @timothycarambat commented on GitHub (Mar 28, 2026): > get rid off useless `@agent` Its not useless, in fact, it makes a lot of sense for people running super small models. Trying to use tools with super small models for every call can ruin the experience dramatically and for people who do that, they may prefer it be opt-in. That being said, [what youre asking for already exists via the automatic mode.](https://docs.anythingllm.com/features/chat-modes#available-chat-modes) Using that chat mode would solve your issues entirely - especially the mic one! If it is worth explaining _why_ it even exists, it was because when AnythingLLM started LLama2 was barely a thing and tool calling didnt even yet exist! By the time it did, local models were still at best 8K, so bloating the window with tool calls where just asking simple QA prompts would spin the model off grabbing tools for no reason! Luckily, this is no longer the major case so now we have automatic mode, which is about to also be the default for all new workspaces so that complaint should go away totally! > [!NOTE] > Automatic mode is very recent and not in the desktop app yet but will be in the next patch sometime next week! > Allow to create Folders to group Chats lets say per Category . I don't see how this benefit extends beyond what workspaces already do? Can you explain some more so I can understand what a UI/UX for that looks and feels like? > you tube videos titles "search anything. online" is focused on authenticated site and search. Not searching anything. It should be implemented by default in the app and allow to search anything , this is a must now days for any AI Rag app !. Please improve that right now its so hard to use and search anything online using AnythingLLM If I understand the question you are saying how the `web-search` skill should be enabled by default so that you can get web resources without having to toggle it on? I have considered this for sure, web-scraping is default but obviously that cannot "search" like web-search can - which can be annoying I understand. That being said, it is not a toggle in tools - so it's easier to flip on <img width="575" height="514" alt="Image" src="https://github.com/user-attachments/assets/060b2e21-d107-4989-94c5-d83bcb2026d2" /> This is all great feedback and clearly thought out so help me understand the bits I am missing so that we can incorporate this - I just dont want to make assumptions on what youre looking for so we are always building the right stuff.
yindo commented 2026-06-05 14:51:27 -04:00
Author
Owner
Copy Link
Copy Source

@computerworxdev commented on GitHub (Mar 28, 2026):

Thank you so much for your detailed response, and I appreciate your time to answer my concerns.

I am so happy that automatic mode is coming to Desktop app. I did not realize that for super small model that was an issue and thanks for including option to switch it to automatic mode since small models are not so good at anything people use tools models that are 9B and up

Regarding search I know about toggle switch and it failed in many cases to work. However you think in a near future web scraping can have a better implementation to scrape other sites automatically? a lot of services already that that.

Reading folder grouping I assume Workspaces supposed to do that?
example of what I meant

  • Web Development (Grouped folder)
    - landing page domain1 (chat window)
    - backend domain2 (chat window)
  • Online Research (Grouped folder) is that a Workspace?
    - Html5 research chat
    - node js deployment chat
  • Casual problem (grouped folder)
    - how to fix a deck (chat)
    - most popular web frameworks (chat)

Thank you again

<!-- gh-comment-id:4149263204 --> @computerworxdev commented on GitHub (Mar 28, 2026): Thank you so much for your detailed response, and I appreciate your time to answer my concerns. I am so happy that automatic mode is coming to Desktop app. I did not realize that for super small model that was an issue and thanks for including option to switch it to automatic mode since small models are not so good at anything people use tools models that are 9B and up Regarding search I know about toggle switch and it failed in many cases to work. However you think in a near future web scraping can have a better implementation to scrape other sites automatically? a lot of services already that that. Reading folder grouping I assume Workspaces supposed to do that? example of what I meant - Web Development (Grouped folder) - landing page domain1 (chat window) - backend domain2 (chat window) - Online Research (Grouped folder) is that a Workspace? - Html5 research chat - node js deployment chat - Casual problem (grouped folder) - how to fix a deck (chat) - most popular web frameworks (chat) Thank you again
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 enhancement feature request
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#5004
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?