AI Agent Hub
Back to plugins
⚙️

deepseek-peak-blocker

Workflow Updated 2026.08.26

Run the following command in DeepSeek Harness:

dsh plugin install zisekongling/deepseek-peak-blocker

Paste the following prompt into your AI chat to install this plugin:

Run dsh plugin install zisekongling/deepseek-peak-blocker in the DeepSeek Harness terminal to install the plugin (source: https://github.com/zisekongling/deepseek-peak-blocker), then restart DSH to activate the bundle.

About this plugin

DeepSeek's official API queues heavily during weekday peak windows (09:00–12:00 and 14:00–18:00 CST). A single stalled request can drag down an entire agent workflow. peak-blocker solves this by inserting a friendly interception at the request level: new calls heading to api.deepseek.com trigger a modal that lets you confirm, cancel, or exempt the current session for this peak cycle. In-flight requests are never interrupted, and third-party proxies, Ollama, and custom providers pass through untouched—only the official billing channel is gated.

Beyond interception, the plugin handles notifications and observability. The instant the schedule flips from off-peak to peak, a non-blocking banner slides in (auto-dismissing after 10 seconds) to alert you that an active conversation is now subject to slower responses. A persistent status panel in the bottom-right corner shows a red/green peak indicator, a live countdown to the next phase, and the count of active DeepSeek requests. The panel is draggable, collapsible, resizable from 60% to 160%, and its preferences persist across page reloads via localStorage. The /bypass command grants a one-shot exemption for the current session, and /status echoes all four real-time metrics on demand.

If you run agent workflows, batch evaluations, or long conversations through DeepSeek's direct billing endpoint, peak-blocker turns peak-hour risk into a visible, controllable, and optionally exemptible decision point. No architecture changes are needed, weekends are fully exempt, and everything works out of the box with sensible defaults.

Use Cases

  • Prevent agent batch requests from stalling in peak-hour queues during workdays
  • Monitor live DeepSeek API load and phase countdown in the middle of a long conversation
  • Bypass the gate with a one-shot /bypass token when an urgent task must go through

Best For

  • Developers who run agent workflows on DeepSeek's official API every day
  • Researchers who rely on long batch runs or evaluations and need stable response times
  • Teams that want to minimize wasted peak-hour queuing and fine-tune API billing efficiency