Back to search
Memory layer for LLM agents — persistent, temporally-aware recall with conflict detection. Rust core + MCP server and CLI.
Stars
0
Forks
0
Watchers
0
Open Issues
0
Overall repository health assessment
^1.0.0^5.0.0^3.24.0^3.0.0^22.0.0^17.3.1^5.7.0No contributors data available
feat: re-embed pipeline + model override support, BGE-large tested and rejected (80.0% vs 86.7% baseline)
f1f3579View on GitHubfeat: Phase 3B GLM-OCR pipeline, real scanned PDF ingest verified, 86.7% keyword
f606d0aView on GitHubfeat: Phase 3A artifact model, parser boundary, provenance wiring, engram store meeting-notes.txt works end to end
b98f5a5View on GitHubfeat: Phase 2C local embedding optimization, 48%72% keyword
b52d7b2View on GitHubfeat: Phase 1 complete recall fix, local embeddings, auto-embed on store
b28bec2View on GitHubfix: revert rrf_k to 15.0 — helped DevMem but regressed LongMemEval
bf87d21View on GitHubfeat: apply autoresearch winners + revert aggressive temporal filter
68b7c33View on GitHubLongMemEval full oracle (500q): 39.2% task-avg, 96.6% abstention
3d360b4View on GitHub