funes-memory
收藏资源简介:
funes memory store 是一个基于 funes 框架的记忆存储数据集,专为代理(agent)会话设计。该数据集将代理会话内容进行分块处理,生成嵌入向量,并以 Lance 表的形式存储,形成一个衍生索引。索引中包含带有确切来源的逐字段落,而非原始转录本,确保了数据的可追溯性。数据集规模为 22,597 个分块,属于中等规模(10K 到 100K 之间),嵌入模型未知,最后更新日期为 2026 年 7 月 20 日。该数据集适用于代理记忆管理、信息检索和对话分析等任务,用户或代理可以直接通过 funes recall 命令从存储中回忆信息,无需构建本地索引,提高了使用便捷性。数据集由 funes push 工具发布并定期更新,以保持内容的最新性。
funes memory store is a memory storage dataset based on the funes framework, specifically designed for agent sessions. The dataset chunks agent session content, generates embedding vectors, and stores them in Lance table format, forming a derived index. The index includes verbatim paragraphs with exact sources, rather than raw transcripts, ensuring data traceability. The dataset size is 22,597 chunks, which is medium-sized (between 10K and 100K), with an unknown embedding model, and the last update date is July 20, 2026. It is suitable for tasks such as agent memory management, information retrieval, and dialogue analysis. Users or agents can directly recall information from storage via the funes recall command without building local indexes, enhancing usability. The dataset is published and regularly updated by the funes push tool to maintain content freshness.




