PlutoKeating/dsh-lark-bot

dsh-lark-bot: A bot that bridges DeepSeek Harness (dsh) into Feishu/Lark, including complete project workspace management, parallel tasks, multi-role Agent, session archiving and safety net protection (Feishu can still save itself after dsh crashes)

dsh-lark-bot brings a local Harness agent into Feishu or Lark private chats, groups, and topics. It streams bounded progress cards, keeps tool details and raw reasoning local, isolates project worktrees, supports parallel jobs and named roles, archives sessions, manages provider settings, and can request tool or plan approvals through interactive cards. A persistent job ledger distinguishes queued, running, completed, failed, and interrupted work, while its guardian service keeps a safe control path available when the main DSH profile crashes.

Messaging & Communication ★ 25 updated 2026-08-25 ✅ runtime-tested
View on GitHub ↗

Install

npx dsh-lark-bot@latest setup --profile dsh-lark

Verified as the README's only official installation command. Then start with dsh --profile dsh-lark and scan the terminal QR code in Feishu or Lark. Do not download third-party Windows executables presented under this project's name.

Compatibility

DeepSeek Harness with a configured provider key, Node >=22.19, and a Feishu or Lark account. Linux, macOS, and Windows are supported through an outbound WebSocket; no public IP, domain, or tunnel is required.

Details

Recent updates

The v0.7.0-era README documents guardian safe mode, multi-bot handoffs, per-scope roles and concurrency, durable job recovery, Web session projection, provider management, plan gates, notifications, archiving, and bilingual Card JSON 2.0 interactions.

FAQ

What is the official dsh-lark-bot install command?
Run npx dsh-lark-bot@latest setup --profile dsh-lark. The project explicitly warns that it does not distribute Windows exe installers.
Do I need a public webhook server?
No. The bridge uses an outbound Feishu/Lark WebSocket and the README says no public IP, domain, server, or tunnel is required.
What happens to an in-flight task after a crash?
Queued jobs are restored; a running task becomes interrupted with its checkpoint retained and must be explicitly retried from /jobs to avoid repeating side effects.

Alternatives

amlyczz/dsh-lark-link · xmanrui/dsh-feishu · xmanrui/dsh-im

More plugins in Messaging & Communication

Browse more in Messaging & Communication

Guides for Messaging & Communication plugins