Repository navigation
docs(memory): list all nine knowledge categories - #6864
Conversation
The guide said six, but VALID_CATEGORIES has nine. It now also explains that profile, system and permission can't be created from a chat turn. Closes amd#3211
|
Verdict: Approve The memory guide said there were six knowledge categories; there are nine. This PR fixes the count and adds Real-world evidenceN/A: docs-only change, with no runnable surface to exercise. The verdict rests on checking the docs against the memory store code. 🔍 Technical detailsIssues found 🟢 One sentence slightly overstates what the dashboard can write ( Checked
Strengths
|
Summary
Corrects the memory guide's category count from six to nine and adds the three missing categories:
profile,systemandpermission.Why
The Knowledge Categories section says there are six categories, but
VALID_CATEGORIESinsrc/gaia/agents/base/memory_store.pyhas nine, andgaia memory statusreports the others. The same guide already lists all nine in its Quick Start, so it contradicts itself.I split the list because the missing three work differently:
store()rejects them from a chat turn unless the caller passesallow_privileged=True.Linked issue
Closes #3211
Changes
profile,systemandpermission, and says where they come from, based on the comment above_PRIVILEGED_CATEGORIES.profileexample follows the template inbootstrap.py, and thepermissionexample comes from the comment onVALID_CATEGORIES.Test plan
VALID_CATEGORIES: both have the same 9 names.docs/guides/memory.mdxagainst open PRs fix(memory): forget every copy of a memory, and don't re-save it on a forget turn #6858 and fix(memory): teach the next chat what worked after a failed Python import #6828, which also edit it: they change other sections.Evidence
N/A: this is a docs-only change. It doesn't touch the Agent UI, MCP tools, CLI behavior or the HTTP API.
Checklist
Closes #N/Fixes #N/Refs #N).python util/lint.py --all,pytest tests/unit/). Not run: this PR only changes one.mdxfile and no code.