dsh-phoenix
Graceful restart + client auto-reconnect + cross-restart goal continuation for DeepSeek Harness (dsh). After a plugin update dsh-phoenix restarts dsh at a safe idle boundary, keeps the browser page connected, and resumes a running goal so a task can keep evolving across restarts. 适合需要在插件更新后重启DSH、自动重连页面并恢复未完目标任务的用户。
同名パッケージの別リポジトリ
インストール
$
npx -p @deepseek-ai/dsh dsh plugin --profile web add github:StvLi/dsh-phoenixドキュメント
README 全文を読む ↗⚙️ Configuration
All knobs are environment variables with safe defaults — no configuration file required.
| Env | Default | Purpose |
|---|---|---|
DSH_PHOENIX_UNIT | dsh-web | systemd user unit to restart |
DSH_PHOENIX_DELAY | 8 | seconds between stop and start |
DSH_PHOENIX_ARMING_MS | 5000 | ignore signals for N ms after load (prevents self-trigger) |
DSH_PHOENIX_DEBOUNCE_MS | 3000 | collapse burst signals into one reboot |
DSH_PHOENIX_DEFER_POLL_MS | 3000 | idle re-check interval while deferring |
DSH_PHOENIX_DEFER_SOFT_MS | 300000 | soft safety deadline — log a warning at this point, keep waiting |
DSH_PHOENIX_DEFER_HARD_MS | 900000 | hard safety deadline — force the restart (if policy is auto) |
DSH_PHOENIX_DEFER_POLICY | auto | auto (force at hard deadline) or wait (never force; may defer indefinitely) |
DSH_PHOENIX_HEALTH_MS | 4000 | browser heartbeat interval |
DSH_PHOENIX_RESTART_CMD | (empty) | custom restart command override for non-systemd deployments (see the Requirements warning) |
DSH_PHOENIX_REARM_MS | 8000 | boot delay before the lifecycle recovery/resume check |
DSH_PHOENIX_MAX_RESUME_ATTEMPTS | 1 | resume attempts per restart (1 = one resume per restart; >1 retries; reaching it ends the continuation) |
DSH_PHOENIX_STATE_FILE | (empty) | path to the durable lifecycle checkpoint; empty disables the lifecycle |