dsp-deepseek-peak

DeepSeek API bills at a higher rate during peak hours (Beijing time 09:00–12:00 and 14:00–18:00) compared to off-peak. It is easy to sit down, fire off a long DeepSeek V4 Pro task during a work break, and only notice the extra cost on the next billing cycle. dsp-deepseek-peak solves exactly that: it pins a compact peak or off-peak badge to the status bar beneath the DSH Web input box, with a green dot for off-peak and an orange dot for peak, plus a live Beijing clock that refreshes every 30 seconds so you always know which rate tier you are in.

When three conditions line up, the plugin intercepts your submission and asks for a second confirmation before the first step runs. You are in a peak window, the composer model is DeepSeek V4 Pro, and this is a fresh turn you are submitting. You can choose to proceed, suppress reminders for the rest of the Beijing day (persists across page refreshes and DSH restarts via localStorage, auto-expires at midnight), or cancel the turn outright. Every code path is fail-open: a missing dialog component, a broken session, or a failed model read will simply let the task through rather than blocking your workflow.

This plugin suits anyone who regularly submits DeepSeek tasks from the DSH Web console, especially those who lean on V4 Pro. It runs entirely on your machine: pure ESM on the host side, a standard ModuleLoader bundle on the client side, zero third-party runtime dependencies, and no data ever leaves the local loopback. If you prefer to tweak things, the peak windows, trigger model, dialog copy, and refresh interval are each a one-line edit in the source.