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 #4821] [DOCS]: Question about agent long term memory behavior and documentation #3038

New Issue
Closed
opened 2026-02-22 18:32:22 -05:00 by yindo · 2 comments
yindo commented 2026-02-22 18:32:22 -05:00
Owner
Copy Link
Copy Source

Originally created by @HanJammer on GitHub (Jan 4, 2026).
Original GitHub issue: https://github.com/Mintplex-Labs/anything-llm/issues/4821

Description

Hi,
I noticed that AnythingLLM appears to use a file called agent_memory.txt as a kind of long-term memory when running agents (tested with GPT-4.1-mini).

What surprised me is that:

  • the model seems to write to this file autonomously,
  • later retrieves and cites it as a knowledge source via RAG,
  • yet there is no visible UI, setting, or documentation explaining:

When entries are written, what criteria decide persistence, how this file is indexed/retrieved, or how a user is supposed to audit, edit, or disable it? From the outside, this looks like an implicit, agent-level memory layer rather than simple conversation history, which has important implications for consistency, bias loops, and compliance.

Could someone point me to:

  • an official description of how agent_memory.txt is intended to work, or
  • the part of the codebase responsible for writing/retrieving it, or
  • recommended best practices for controlling or disabling this behavior?

Thanks - I just want to understand the design intent and guarantees here.

Originally created by @HanJammer on GitHub (Jan 4, 2026). Original GitHub issue: https://github.com/Mintplex-Labs/anything-llm/issues/4821 ### Description Hi, I noticed that AnythingLLM appears to use a file called agent_memory.txt as a kind of long-term memory when running agents (tested with GPT-4.1-mini). What surprised me is that: - the model seems to write to this file autonomously, - later retrieves and cites it as a knowledge source via RAG, - yet there is no visible UI, setting, or documentation explaining: When entries are written, what criteria decide persistence, how this file is indexed/retrieved, or how a user is supposed to audit, edit, or disable it? From the outside, this looks like an implicit, agent-level memory layer rather than simple conversation history, which has important implications for consistency, bias loops, and compliance. Could someone point me to: - an official description of how agent_memory.txt is intended to work, or - the part of the codebase responsible for writing/retrieving it, or - recommended best practices for controlling or disabling this behavior? Thanks - I just want to understand the design intent and guarantees here.
yindo added the documentation label 2026-02-22 18:32:22 -05:00
yindo closed this issue 2026-02-22 18:32:22 -05:00
yindo commented 2026-02-22 18:32:22 -05:00
Author
Owner
Copy Link
Copy Source

@shatfield4 commented on GitHub (Jan 6, 2026):

For agent-memory.txt it is not actually a file on disk. It's just a label the agent uses when it stores info directly into your workspace's vector database. This happens through the "RAG & long-term memory" skill that's on by default.

This skill does two things:

  • Search through documents in your workspace (including items it saved before)
  • Store new information when you ask it to

It's supposed to only save info when you explicitly say "remember this" or "save that to memory." Most models are pretty good about following this rule. But weaker models sometimes just decide to store things on their own.

The memories are storied directly in your workspace's vector database alongside your regular documents (but they do not show up in the file upload modal).

Unfortunately there's no audit log right now for when things get stored, and there's no separate UI to review just the agent memories vs your uploaded docs. If you're dealing with GDPR/HIPAA/etc I'd recommend just disabling the skill for now until we add better controls and logging.

@shatfield4 commented on GitHub (Jan 6, 2026): For `agent-memory.txt` it is not actually a file on disk. It's just a label the agent uses when it stores info directly into your workspace's vector database. This happens through the "RAG & long-term memory" skill that's on by default. This skill does two things: - Search through documents in your workspace (including items it saved before) - Store new information when you ask it to It's supposed to only save info when you explicitly say "remember this" or "save that to memory." Most models are pretty good about following this rule. But weaker models sometimes just decide to store things on their own. The memories are storied directly in your workspace's vector database alongside your regular documents (but they do not show up in the file upload modal). Unfortunately there's no audit log right now for when things get stored, and there's no separate UI to review just the agent memories vs your uploaded docs. If you're dealing with GDPR/HIPAA/etc I'd recommend just disabling the skill for now until we add better controls and logging.
yindo commented 2026-02-22 18:32:22 -05:00
Author
Owner
Copy Link
Copy Source

@timothycarambat commented on GitHub (Jan 9, 2026):

This issue should be closed in favor of a true memory system and UI to manage said memories per workspace & user.
https://github.com/Mintplex-Labs/anything-llm/issues/4846

@timothycarambat commented on GitHub (Jan 9, 2026): This issue should be closed in favor of a true memory system and UI to manage said memories per workspace & user. https://github.com/Mintplex-Labs/anything-llm/issues/4846
yindo changed title from [DOCS]: Question about agent long term memory behavior and documentation to [GH-ISSUE #4821] [DOCS]: Question about agent long term memory behavior and documentation 2026-06-05 14:49:57 -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 documentation
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#3038
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?