Back to Vault
Agent Memory
Tier 3
semantic-memory (RecursiveIntell)
Summary
Local-first hybrid semantic search backed by authoritative SQLite state plus a vector sidecar: facts, chunked documents, conversation messages, and episodes searched via BM25 (FTS5) and vector retrieval fused with Reciprocal Rank Fusion, with explainable search.
Key Takeaways
- Hybrid BM25 + vector with rank fusion beats either alone
- Keep an authoritative relational store; treat vectors as a sidecar index
- Explainable retrieval helps debug why a memory surfaced
Reliability Note
Open-source implementation; retrieval patterns are broadly applicable.
Flows informed by this source
2Memory & Context
Memory Recall: Ranking and Relevance
Make recall surface the right memories: hybrid lexical+semantic search, rank fusion, and explainable retrieval.
01Build the lexical layer
02Add the semantic layer
03Fuse rankings with boosts
+1 more steps to Done
Best forproduction-grade builds with strict verification
4 steps90-120 minutesAdvanced
Memory & Context
Vector vs Markdown Memory: Make the Decision
Choose your memory backend on evidence: when plain files win, when vectors earn their complexity, and how to test it on your own data.
01Profile the workload honestly
02Benchmark both on your data
03Decide and set the tripwires
Best foryour first pass at this - no prior setup assumed
3 steps45-60 minutesBeginner