dsh-model-quota-usage

When you juggle DeepSeek, Kimi, GLM, Claude, and other Harness providers every day, account balances and token usage are scattered across multiple vendor dashboards. dsh-model-quota-usage consolidates your real-time DeepSeek Open Platform balance and daily token consumption across all providers and models into a single draggable, collapsible floating panel, eliminating the need to tab back and forth.

The top section shows your DeepSeek balance; the body lists input, cache-read, cache-write, and output tokens per provider and per model, with a combined total. The token formula is explicit: Total = Input + Cache-read + Cache-write + Output. Reasoning tokens appear separately in model detail and are not added into the total again, preventing double-counting on providers that already include them in output. The panel is edge-anchored, remembers its position after you drag it, and snaps back to the default bottom-right corner with a double-click on the title bar. Local statistics refresh every 60 seconds; all data stays on your machine with zero telemetry.

Built for engineers who rotate across multiple providers in production and want persistent, at-a-glance cost awareness, as well as developers who prefer lightweight, local-only monitoring over yet another browser tab.