Perceive Beating AI Newsflash, Kimi has started the internal testing of the new Lark AI colleague, Mira. It resides directly within Lark, without the need to switch apps or open new windows. Once added to a project group, members can assign tasks to it just like they would to a real colleague using @ mentions. In private chats, Mira can engage in Q&A, assist with writing, and manage todos.
Mira will also continuously organize messages in the background, conduct retrospectives, and perform regular check-ins without the need to wait for someone to ask. It can follow up on projects, remind about tasks, and upon receiving user feedback, it can review code and documentation. If it identifies a code bug, Mira can even fix the code by itself and submit a PR.
Each group and each private chat is treated as a separate space with default contextual isolation. To allow Mira to read messages from other groups, separate authorization is required, and it will only communicate within its own space. Organization admins can also globally configure Skills, MPC, and overall context.
The Beta version currently only supports Lark and utilizes the model quota of Kimi Code members. Official recommendation is to use K3, as native API Key is not supported at the moment. Due to Mira's proactive nature, it may continue to consume model quota even without user queries.
