cortexdb

CortexDB

CortexDB is a pure-Go, single-file AI memory and knowledge graph library. It uses SQLite as the storage kernel and exposes vector search, lexical search, RAG knowledge storage, agent memory workflows, RDF/SPARQL/RDFS/SHACL knowledge graph features, corpus-to-graph workflows, and MCP-aligned tool APIs.

It is designed for local-first AI agents that need durable memory without running a separate vector database, graph database, or MCP service stack.

Architecture

pkg/cortexdb
  Core public DB facade: vectors, text search, knowledge, memory, KnowledgeMemory, KG, tools, MCP.

pkg/memoryflow
  Agent memory workflow: transcript ingest, recall, wake-up context, diary, promotion.

pkg/graphflow
  Corpus-to-graph workflow: extraction schema, build, analyze, report, export, HTML.

pkg/importflow
  External structured-data import (CSV / SQL dumps / live Postgres-MySQL) into RAG + knowledge-graph foundations, AI-assisted mapping optional.

pkg/connector
  Privacy gate over importflow: live Postgres/MySQL sources, PII classification + human-signed desensitization (masking, free-text redaction, reversible AES-256-GCM vault), and near-real-time CDC (polling / Postgres logical replication / MySQL binlog).

pkg/graph
  Low-level graph engine: property graph, RDF triples/quads, SPARQL, RDFS, SHACL.

pkg/core
  SQLite storage, embeddings, FTS5, vector indexes, chat/session primitives.

Use pkg/cortexdb first. Reach for pkg/memoryflow when building agent memory UX, pkg/graphflow when building graph extraction/report pipelines, and pkg/graph only when you need low-level RDF or property graph control.

Install

go get github.com/liliang-cn/cortexdb/v2

Quick Start

package main

import (
	"context"
	"fmt"
	"log"

	"github.com/liliang-cn/cortexdb/v2/pkg/cortexdb"
)

func main() {
	db, err := cortexdb.Open(cortexdb.DefaultConfig("KnowledgeMemory.db"))
	if err != nil {
		log.Fatal(err)
	}
	defer db.Close()

	ctx := context.Background()
	quick := db.Quick()

	_, _ = quick.Add(ctx, []float32{0.1, 0.2, 0.9}, "SQLite is a single-file database.")
	results, _ := quick.Search(ctx, []float32{0.1, 0.2, 0.8}, 1)

	if len(results) > 0 {
		fmt.Println(results[0].Content)
	}
}

Choose the Right Layer

Need vectors / collections / FTS5?      -> pkg/cortexdb / pkg/core
Need RAG knowledge storage/search?      -> pkg/cortexdb SaveKnowledge/SearchKnowledge
Need chat/session memory workflow?      -> pkg/memoryflow
Need RDF/SPARQL/RDFS/SHACL?             -> pkg/cortexdb knowledge graph APIs
Need corpus-to-graph/report/export?     -> pkg/graphflow
Need agent tools or MCP server?         -> db.GraphRAGTools() / db.NewMCPServer()
Need low-level graph control?           -> pkg/graph

High-Level Knowledge and Memory

_, _ = db.SaveKnowledge(ctx, cortexdb.KnowledgeSaveRequest{
	KnowledgeID: "apollo-plan",
	Title:       "Apollo launch plan",
	Content:     "Alice owns Apollo. Apollo ships on Friday.",
	ChunkSize:   24,
	Entities: []cortexdb.ToolEntityInput{
		{Name: "Alice", Type: "person", ChunkIDs: []string{"chunk:apollo-plan:000"}},
		{Name: "Apollo", Type: "project", ChunkIDs: []string{"chunk:apollo-plan:000"}},
	},
	Relations: []cortexdb.ToolRelationInput{
		{From: "Alice", To: "Apollo", Type: "owns"},
	},
})

resp, _ := db.SearchKnowledge(ctx, cortexdb.KnowledgeSearchRequest{
	Query:         "Who owns Apollo?",
	Keywords:      []string{"Apollo", "Alice", "owns"},
	RetrievalMode: cortexdb.RetrievalModeLexical,
	TopK:          3,
})
_ = resp.Context

Without an embedder, CortexDB uses lexical retrieval and planner-provided keywords. With an embedder, the same high-level APIs can use semantic or hybrid retrieval.

RAG benchmark coverage is available in pkg/cortexdb:

go test ./pkg/cortexdb -run '^$' -bench 'BenchmarkRAG' -benchmem

Reference run on Apple M2 Pro, -benchtime=3x: