[合作讨论] 聊天人格学习如何扩展到多身份、多关系和受控回复

Hello, we are developing Yance: a locally prioritized personal relationship and communication intelligence brain powered by multiple MCP connectors. It is not another chat record exporter or single-platform bot. Connectors like WeChat, Little Red Book, Lazada, DingTalk, FeiShu, etc. are only responsible for providing messages within the authorized scope, and clearly identify "who is the user and who is the other party"; Yance is responsible for linking multiple identities of the user in different scenarios, as well as multiple accounts, platform identities, and relationships of the same person, and learning communication characteristics of both parties from long-term chats. In the early system, only recommended responses are provided, and users select, modify, and confirm them. After accumulating enough feedback, users can enable limited, auditable, and suspended automatic responses based on identity, account, object, and scenario. In the long term, we will also explore controllable communication signals (including the so-called PUA mode), implied intentions, opportunity recognition, and multi-party relationship strategies, but all judgments must retain evidence, counter-evidence, alternative interpretations, and uncertainty.