Origin: Local AI Knowledge Daemon with Git Versioning

What changed

The Origin daemon, built in Rust, is presented as a local-first solution for managing AI-generated knowledge. It aims to prevent the loss of valuable information that typically occurs when AI conversations end. Origin achieves this by transforming documents, notes, and past AI interactions into a source-backed knowledge base that evolves with the user's work. The system distinguishes between three core components: Sources, which provide traceable material; Memories, which capture atomic decisions and lessons learned with provenance; and Pages, which compile current knowledge from Sources and Memories into reusable, source-cited Markdown.

Origin extends the LLM-wiki concept by incorporating a memory lifecycle and a knowledge graph. It supports Git-versioned memories and distilled wiki pages, allowing for durable storage and retrieval of information. The system's knowledge graph is built upon typed Entities, observations, and directed relations extracted from Memories. Relations are defined with a seeded vocabulary (e.g., uses, part_of, contradicts) and can store confidence, explanations, and their source Memory. Entity linking and resolution aim to reuse existing nodes, preventing redundancy. Correction is handled without erasure, keeping related claims, corrections, and explicit supersessions inspectable alongside original Sources and Memory history.

Retrieval in Origin is a local hybrid pipeline combining multiple stages. It includes a full-text index (SQLite FTS5) for exact wording, FastEmbed for semantic similarity using quantized embeddings indexed by libSQL cosine DiskANN, and a weighted RRF fusion of lexical and semantic rank lists. An optional cross-encoder reranking stage can further refine results. The system also incorporates a graph-memory stream that uses eligible entity links as a third RRF signal, all while respecting defined Space boundaries.

Two distinct lifecycles are managed: Atomic Memory (CAPTURE -> CLASSIFY -> ENRICH -> LINK -> RECONCILE) and maintained Pages. The Atomic Memory process focuses on capturing complete, self-contained ideas following the Zettelkasten principle. Model-backed stages are optional and depend on configured models, with a reconcile pass off by default. The system supports AI clients such as Claude Code, Cursor, and Codex, and any MCP client, through a bundled daemon and CLI.

Why it matters for builders

Origin provides AI builders with a robust framework for managing the lifecycle of knowledge generated through AI interactions. By offering Git-versioned memories and source-backed pages, it enables the creation of persistent, evolving knowledge bases. This is crucial for projects that require continuous learning, iterative development, and reliable access to past decisions and insights. The local-first approach ensures data privacy and control.

Practical impact

Builders can integrate Origin into their workflows to maintain a structured and traceable repository of AI-generated information. This includes setting up the daemon via its desktop app (available for macOS and Windows) or headless runtime, and configuring AI clients like Claude Code, Cursor, or Codex to connect to the knowledge base. The setup guide, accessible via a provided URL, offers client-specific instructions for seamless integration. Developers can explore the system's capabilities for managing sources, memories, and pages, and experiment with its hybrid retrieval pipeline for enhanced information access.

Caveats and source limits

The provided source information indicates that Origin is a fresh release, with the latest version noted as v0.7.0. While macOS and Windows have desktop builds, a Linux desktop build is not yet available. macOS Intel also currently lacks a supported complete-runtime install. The Windows x64 installer is not yet signed, requiring users to bypass a SmartScreen warning. The source does not provide details on pricing, specific performance benchmarks, or extensive user adoption metrics, limiting a deeper analysis of its market impact or scalability beyond the described technical features.