dsh-context-show

Running long conversations with DeepSeek raises two constant questions: how much context window is left, and what did this turn actually cost. dsh-context-show answers both at a glance. A small ring-plus-percentage button sits in the conversation header; click it and a draggable, persistent panel slides open showing live usage ratio, per-model spending, and tool-level token attribution. Every number is pushed via session projection, so compaction shrinkage and snapshot changes refresh in real time without a page reload.

The panel handles peak and off-peak pricing natively. It automatically classifies each token bucket into peak hours (Mon-Fri 9:00-12:00 and 14:00-18:00 Beijing time) or off-peak (weekends and remaining hours) and multiplies by the matching rate. Default tables cover deepseek-v4-flash and deepseek-v4-pro across cache-hit, cache-miss, and output columns; you can switch to flat pricing or override supplier rates from the settings page at any time. Toggle between a compact view (usage and total cost only) and a detailed view (source breakdown, tool table, per-model line items), then drag the panel anywhere on screen. It stays put until you press Escape or click the button again.

This plugin is built for two audiences. First, developers who rely on long-context tasks such as code completion or agent orchestration and need a constant eye on remaining window space. Second, billing or ops teams that must break model-call costs down to individual tools, turns, and peak versus off-peak periods. No configuration is required out of the box; after installing and restarting the profile the ring appears in the header, and later price adjustments take effect immediately without another restart.