DSH_HOME single-instance lock plugin for DeepSeek Harness: aborts startup loudly when another dsh server uses the same data directory (prevents session-log corruption)
Pending human review — the facts below come from the source repository.
DSH_HOME single-instance lock plugin for DeepSeek Harness: aborts startup loudly when another dsh server uses the same data directory (prevents session-log corruption)
Users who want this capability in DSH; check the source and docs before installing.
dsh plugin --profile web add github:Tang-mm95/dsh-single-instance-guard No supported platforms were declared.
DSH_HOME data directory at startup and aborts loudly when another live dsh server already uses it. 中文说明 ## Why The JSONL session persistence backend documents "One live writer per session" and has no cross-process defense. Two dsh servers sharing one DSH_HOME (a desktop wrapper spawning its own server, or two dsh web processes) append batches with stale sequence cursors, corrupting session logs: history unavailable for session "…": Error: corrupt session log: seq gap in committed region … See this report for the full diagnosis, a read-only scanner, and a manual repair procedure. This plugin turns the silent corruption into a loud startup failure. ## Install After publishing to npm: bash dsh plugin --profile <profile> add dsh-single-instance-guard Manual (any dsh install): add the row to the profile's cordis.patch.yml before session-related bundles: yaml - insert: - id: single-instance-guard name: 'dsh-single-instance-guard' The guard is a profile bundle: its manifest declares the same patch, so the CLI installs it as a patch layer. ## How it works 1. Creates <DSH_HOME>/.dsh-server.lock atomically (O_EXCL), holding { pid, startedAt, hostname }. 2. On conflict, probes the holder's pid liveness: a live holder aborts startup with a clear bilingual error; a stale lock (dead pid or unparsable file) is removed and acquisition retried once. 3. On process exit the lock is removed — only if still owned by this process. DSH_HOME is resolved exactly like dsh itself: $DSH_HOME or ~/.dsh. Servers with different data roots never conflict. ## Known limits - The unlink+retry path has a tiny race when two processes discover the same stale lock simultaneously; the atomic O_EXCL write still admits exactly one winner, and the loser fails correctly on the retry. - The lock protects against concurrent servers sharing one DSH_HOME. It does not protect against two dsh instances deliberately pointed at the same session file via different roots (unsupported anyway). ## Development bash npm test # node --test ## License MITNot by stars — picked by hand, one at a time.
A visual settings dashboard for DeepSeek Harness: every registered settings namespace — including third-party plugin ones the official UI never covers — rendered as editable, schema-driven forms. Early, but the right direction: stop hand-editing YAML.
A DIY trajectory visualizer for DeepSeek Harness: watch the agent’s plan and execution unfold as a graph instead of scrolling raw logs. Early, but it fills a real gap — understanding at a glance what the agent is doing.
Dispatch DSH work from Claude Code / Codex: spin up in-host DSH agent sessions with per-tier presets, watch native subagent progress, and borrow a multimodal bridge that lends the text-only harness vision and image generation. An orchestration plugin that turns DSH into a backend for other coding agents.