Memory research

Consolidation should leave receipts.

Memory stores get messy over time. Consolidation helps, but opaque rewrites are dangerous. This track makes every split, merge, and update explain what changed and which source memories it came from.

See all memory features How memory works Issue #561
This is one research track, not the whole Remnic memory system. Remnic also includes extraction, recall injection, hybrid search, entity tracking, lifecycle hygiene, trust zones, procedural memory, local LLM routing, versioning, retention tiers, disclosure controls, project scoping, importers, benchmarks, recall X-ray, and more. These pages explain focused memory tracks inside that broader system.
What changes

The product idea in plain terms.

Name the operation

SPLIT, MERGE, and UPDATE make the transformation explicit instead of hiding it inside a summary.

Point to evidence

Derived memories retain links to source versions, so a consolidated fact can explain what it rests on.

Undo bad calls

Because source snapshots remain addressable, bad consolidation can be reversed with a dry-run preview.

Already in Remnic
  • Fuzzy duplicate scanning and semantic consolidation.
  • Page versioning for overwritten memories.
  • Doctor checks for memory-store integrity.
Landed in this track
  • PR #574 added operator vocabulary and derived_from / derived_via fields.
  • PR #624 and PR #632 made consolidation writes and prompts operator-aware.
  • PR #634 added doctor integrity checks and PR #637 added consolidation undo.
References
Related Remnic pages