Memories tab
The Memories tab inspects active memory, explores its graph, and performs maintenance actions.
Search
Use the search field to match text anywhere in each Markdown memory file. The search is case-insensitive and works together with the scope and type filters.
The search mode menu has two options:
- Text matches only shows memories whose file content or metadata directly matches the query.
- Text matches + related memories also shows connected dependency, duplicate, and semantic memories.
Clear the field to restore the full graph for the selected filters.
Scope chips
Filter the graph by memory source. Compare workspace vs global memory. Start with the current workspace only when the graph feels noisy.
Type chips
Filter the graph by memory type. Inspect rules, skills, or knowledge separately.
Semantic links toggle
Shows semantic graph edges. Turn off when the graph is visually noisy.
Refresh / rebuild
Reloads or rebuilds graph data. Use after edits, imports, archive actions, or config changes.
Memory preview
Reads selected memory content. Useful to audit what the agent will receive.
Sensitive local content may be visible in the browser. Treat the panel as open while previewing.
Edit memory
Opens the file in an editor and copies the path. Use for manual correction or review. The source of truth is the Markdown file.
Archive memory
Removes memory from active routing while preserving it under archive/. Use archive, not delete, for auditability.
Archiving changes routing immediately. Use archive, not manual deletion, so history is preserved.
증거 기반 메모리
The detail panel mirrors CLI/index metadata. It shows authority, revision, evidence_refs, derived_from, supersedes, superseded_by, valid_from, valid_until, and last_confirmed when present. These fields explain why a memory exists and which immutable trace supports it; the Web UI does not turn evidence into an instruction.
CLI equivalent
engram graph "<topic>"
engram quality-check
engram archive --reason "<why>" <id-or-file>