Capturas de pantalla del proyecto
Resumen
ReMe operates on the principle of 'Memory as File,' converting raw agent dialogs and external resources into structured Markdown files with frontmatter and wikilinks. It continuously indexes, links, and consolidates these files into a self-evolving knowledge base. The system follows a capture, index, consolidate, and recall loop, allowing both human users and AI agents to read and write memory directly. It supports progressive hybrid search combining BM25, wikilinks, and optional embeddings.
Características principales
- Memory as File: Markdown files with frontmatter and wikilinks
- Self-evolving knowledge base via Auto Memory, Auto Resource, and Auto Dream
- Progressive hybrid search combining BM25, wikilinks, and embeddings
- Agent-friendly integration through SKILL.md, CLI, HTTP API, MCP server, and SDK
- Local-first architecture
Requisitos, instalación y guía rápida
Uso
Compatibilidad de modelos y casos de uso
Not stated in the repository metadata
Notas sobre la licencia y los riesgos
The project is open source under the Apache License 2.0.
Editorial verification 2026-08-09: repository URL, owner, description, license and repository statistics were reviewed. License metadata: Apache-2.0. README was fetched for the channel draft; re-check repository dependencies, releases and model terms before production use.
Lanzamiento y mantenimiento
The repository mentions previous versions (0.3.x, 0.2.x, MemoryScope) and notes recent additions such as Cookbooks for Daily Paper and Auto Fin, as well as an experience-driven enhancement method available on arXiv. A related paper was accepted to Findings of ACL 2026.