rooma

What problem does it solve?

rooma removes the “one-shot reply” limitation of typical chat integrations by letting an AI agent stay present in Feishu chats, listen to ambient context in real time, and take action when mentioned or when a clear task appears.

Core Features & Use Cases

• Live WS event streaming: streams inbound Feishu messages into the Claude Code conversation as JSON events with mention awareness.
• Chat discovery and context building: supports fetching chat lists, members, and room purpose (including pinned announcements) before responding.
• Action orchestration: sends messages and files, reacts with completion acknowledgements, and escalates to humans with targeted help requests.
• Session-aware broadcast & farewell: posts a “start working” broadcast to selected chats and automatically sends a farewell on session end (when supported by the hook).

Example use case: when a project question appears in a group where the bot is @-mentioned, rooma pulls recent context, consults the room’s stated purpose/announcement, then responds briefly while marking completion with a reaction if appropriate.

Quick Start

Tell the skill to start a Feishu collaboration session by saying “start feishu”, then wait for the monitor stream to come online and respond to the next @-mention event.