CortexDB

CortexDB is a local-first AI memory and knowledge graph library implemented in pure Go, designed to provide a unified solution for managing vectors, search, retrieval-augmented generation (RAG), agent memory, and SPARQL knowledge graphs within a single SQLite file. It addresses the need for embeddable, developer-friendly infrastructure for AI and knowledge-driven applications, offering modular components for memory workflows, graph extraction, and structured data import.

The platform features a primary database facade that supports vectors, full-text search (FTS5), knowledge and memory recall, knowledge graph APIs, and a suite of tools for both in-process and external agent integration. Its architecture includes specialized packages: MemoryFlow for agent transcript ingest and recall, GraphFlow for corpus-to-graph extraction and visualization, and ImportFlow for one-pass import of structured data such as CSV or MySQL/PostgreSQL dumps with optional AI-assisted mapping and triple extraction. CortexDB also supports an embedded RDF knowledge graph with practical SPARQL subset, RDFS-lite inference, and SHACL-lite validation, enabling advanced querying and schema validation capabilities.

For interoperability and broader language support, CortexDB provides a gRPC sidecar server with bearer-token authentication, allowing clients in Rust, Python, and Node to access its full API through typed client libraries. The Rust client can auto-spawn the sidecar, while Python and Node integrations are supported via published packages. io format can be installed to give Python (Hermes) or Node (OpenClaw) agents durable memory and SPARQL knowledge graph capabilities. The system includes high-level APIs for saving, searching, and recalling knowledge and memory, as well as context management for agent workflows. A no-embedder mode is available, leveraging lexical retrieval and LLM-planned queries for flexible information access.

CortexDB is intended for developers building AI, agent, and knowledge graph solutions who require a local, embeddable, and language-agnostic backend. Its modular Go packages and SQLite-based storage make it suitable for both lightweight and advanced knowledge-driven applications. The tool is distributed as a native Go library, with additional support for Rust, Python, and Node through gRPC and client libraries.