dsh-peak-valley-queue
Run the following command in DeepSeek Harness:
dsh plugin install EmptyCollin/dsh-peak-valley-queue
Paste the following prompt into your AI chat to install this plugin:
Run dsh plugin install EmptyCollin/dsh-peak-valley-queue inside DeepSeek Harness to install the plugin (repo: https://github.com/EmptyCollin/dsh-peak-valley-queue), then restart the service and refresh the browser page.
About this plugin
DeepSeek charges a premium for API calls made during peak windows compared to off-peak (valley) windows, and much of the typical workload—bulk rewriting, summarization batches, inspection checklists, recurring report generation—does not need an instant response. dsh-peak-valley-queue closes that timing gap with a single switch beside the composer: when toggled on, every outgoing message is intercepted into a FIFO task queue instead of being sent directly, and the queue auto-dispatches in order once a valley window opens and sufficient lead time remains before the next peak, all without manual intervention.
Core capabilities include fully configurable peak/valley schedules (overlapping ranges, weekend policies, timezone choice, with DeepSeek's official schedule as default); precise four-state completion detection (completed, aborted, failed, blocked) backed by redundant signals—a turn/end reason, agent idle status, a 10-second watchdog, and client-side polling—so no single missed event can wedge a task in the running state; automatic queue pause with a browser system notification (repeated every 30 minutes, plus an optional custom reminder command) on failure or interruption, resuming only when you explicitly click Continue to prevent silent infinite retries; per-session isolated, persistent queues written to the local DSH data directory and never touching workspace files; and integration with dsh-todo-enhancer, which flips a task to a pending-confirmation state and halts the queue when decision-list markers appear in a reply, waiting for you to resolve each item before proceeding.
This plugin is best suited to heavy DeepSeek users whose workflows skew toward batch, deferrable tasks and who want to anchor API spend to off-peak rates. If your use case demands real-time conversation, urgent delivery during peak windows, step-by-step interactive debugging where you need to see each model response before deciding the next move, or tasks that depend on immediate human confirmation, the queuing delay is counterproductive—simply leave the switch off to restore direct sending. The two modes coexist without conflict.
Use Cases
- Queue bulk rewriting and summarization to valley windows for lower API rates
- Auto-queue peak-hour inputs and dispatch them in order when valley windows open, no babysitting needed
- Execute multi-step batch tasks serially to avoid concurrency interference and keep context coherent
Best For
- Heavy DeepSeek API users whose workflows are dominated by batch, deferrable tasks
- Developers who want to pin API spend to off-peak rates without manually tracking schedules
- Batch-processing workflows that need visible pause and controlled retry on failure
Related Plugins
A method pack that makes AI coding agents plan against your real baseline, prove completion with fresh evidence, and reduce reworks and unsafe changes.
Turns the DeepSeek Harness session into a captain that builds a durable sub-agent team, splits goals into dependency-aware tasks, and coordinates work via direct messages and a live Web UI.
Gives coding agents design judgment, letting Claude Code, Cursor, and 70+ agents generate and iterate high-quality UI, presentations, and graphics right from the terminal.
Run the Pi ecosystem's plugins on DeepSeek Harness, unmodified, via a compatibility layer that implements Pi's public extension ABI on DSH's native services.