dsh-feishu-integration

A DeepSeek Harness session finishes and the summary sits in a terminal or a browser tab. Wanting to follow up means switching back to that window. dsh-feishu-integration moves that loop into Feishu (or Lark): when a DSH turn completes, the text summary is pushed to a direct chat or group; when you reply to that summary in Feishu, the message is routed back to the exact DSH session that produced it. The plugin first acknowledges in-thread with the workspace path, session title, and session ID, then injects your text in queue mode, and the final answer posts back into the same Feishu thread. Beyond push-and-reply, when a DSH session pauses on ask_user_question the question and options are forwarded to your most recent Feishu thread—you answer on your phone with a number or a line of text and the session resumes, no terminal needed. The sender ships with exponential-backoff retry, DNS reachability gating, and idempotency UUIDs so a flaky network neither loses nor duplicates a summary; if every retry is exhausted the dead-letter is written locally for debugging. Routing uses a message_id to sessionId map plus an fsum- prefix to prevent loop-back.

Management lives in the DSH settings page: QR-code binding, binding status, reconnect, disconnect, and unbind are all in one panel. A fsum-admin CLI is also available for headless or scripted environments. Both the domestic Feishu domain and the international Lark domain are supported, and the plugin keeps the legacy local data layout so existing bindings and pinned sessions carry over during migration. By default takeoverInbound is off, so the plugin will not compete with other Feishu long-connection plugins in the same profile for event delivery; enable it only after the older plugin is confirmed disabled.

This plugin is for DeepSeek Harness users who work inside Feishu and want summary viewing, follow-up questions, and mid-session answers to stay in their chat window—especially engineers who need to pick up a conversation from a mobile device or who track AI sessions in a team group.