Expand description
§MindCore
A pluggable, feature-gated memory engine for AI agent applications.
MindCore provides persistent storage, keyword search (FTS5), vector search, hybrid retrieval (RRF), graph relationships, memory consolidation, cognitive decay modeling, and token-budget-aware context assembly.
§Quick Start
use mindcore::prelude::*;
// Define your memory type
// (implement MemoryRecord for your struct)
// Build the engine
// let engine = MemoryEngine::<MyMemory>::builder()
// .database("memory.db")
// .build()?;§Feature Flags
| Feature | Description |
|---|---|
fts5 (default) | FTS5 keyword search with Porter stemming |
vector-search | Candle embeddings + hybrid RRF search |
reranking | Cross-encoder reranking (post-RRF) |
graph-memory | SQLite relationship tables + CTE traversal |
temporal | Bi-temporal validity tracking |
consolidation | Hash/similarity/LLM-assisted dedup pipeline |
activation-model | ACT-R cognitive decay model |
two-tier | Global + project memory with promotion |
encryption | SQLCipher encryption at rest |
mcp-server | MCP server interface |
full | All features except encryption and mcp-server |