Overview
The Memory page is the central hub for managing your project's AI-generated knowledge. It has four tabs:
| Tab | Purpose |
|---|---|
| Brain | Browse and manage all extracted memory items |
| Sessions | View coding sessions and trigger enrichment |
| Review | Triage items flagged by staleness detection or low confidence |
| Hygiene | Run the AI-powered cleanup agent |
Brain tab
The Brain tab shows all memory items for the current project. Each item displays:
- Title -- A descriptive name for the memory
- Type -- Category (architecture, convention, implementation, decision, bug, todo)
- Confidence -- AI confidence score from 0 to 1
- Status -- Current state (draft, approved, deprecated, review, archived)
- Schema key -- Hierarchical location in the brain
- Source -- The session that generated this item
- Files -- Related file paths
Filtering
Use the filter controls at the top of the list:
- By type -- Architecture, Convention, Implementation, Decision, Bug, Todo
- By status -- Draft, Approved, Deprecated, Review, Archived
- By schema key -- Filter by brain hierarchy path (e.g.
root/architecture/auth)
Confidence scores
| Score | Meaning | Recommended action |
|---|---|---|
| 0.90 - 1.00 | High confidence | Likely accurate, safe to approve |
| 0.70 - 0.89 | Moderate confidence | Review briefly before approving |
| 0.50 - 0.69 | Lower confidence | Review carefully |
| Below 0.50 | Low confidence | Verify against source code |
Managing items
- Approve -- Include the item in the brain document served to AI tools
- Deprecate -- Hide the item from the brain (not permanently deleted)
- Edit -- Modify title, content, type, or schema key
- Delete -- Permanently remove the item
Bulk operations
Select multiple items with checkboxes, then use Approve All, Deprecate All, or Delete All.
Sessions tab
Browse coding sessions captured by the VS Code extension. Each session card shows source, status, event count, and extracted item count. Click a session to see its detail view with the pipeline timeline, events, and enrichment jobs.
For full details, see Sessions.
Review tab
The Review tab surfaces memory items that need human attention. Items appear here for two reasons:
Staleness detection
When new commits modify files referenced by existing memory items, those items are automatically flagged as review. If 2 or more of an item's referenced files were modified in a recent commit, the item may be outdated.
Low confidence
Items with a confidence score below 50% may contain inaccurate information and benefit from review.
Actions
| Action | Effect |
|---|---|
| Accept | Sets status to active and resolves associated problem docs |
| Edit | Opens an inline editor -- saving also accepts the item |
| Archive | Sets status to archived and resolves problem docs |
| Boost | (Low Confidence only) Sets confidence to 80% |
Select multiple items with checkboxes for Accept All or Archive All bulk operations.
Hygiene tab
The Hygiene tab runs an AI-powered cleanup agent that proposes safe edits: renaming titles, retagging, merging duplicates, deprecating superseded items, and flagging quality issues. The agent runs in two phases -- analyze first, then review and selectively apply changes. No modifications are made without your approval.
For details, see Hygiene Agent Guide.
Next steps
- Sessions -- Session lifecycle and enrichment triggers
- Operations -- System monitoring and operational tools
- Enrichment Guide -- How the enrichment pipeline works
- Brain Assembly Guide -- How items are assembled into the brain