dsh-convo-cost

When you drive conversations through the DeepSeek API, cost is often invisible—tokens fly, money drains, and you only find out how much a session actually cost by digging through invoices after the fact. dsh-convo-cost puts a real-time cost line right below the conversation input bar, updating instantly as each assistant/message usage event lands, so you always know where you stand: ¥x.xx · xx tok.

Hover over the status line for a full breakdown: call count, input tokens (with cache-hit split), output tokens (with reasoning split), the active pricing mode (base or peak/valley), and the model name. Pricing is strictly aligned to the official DeepSeek rate card (CNY per million tokens) and auto-switches to peak pricing during Beijing-time peak windows effective 2026-08-17. Unknown models are billed at ¥0 to avoid inflating estimates. The entire price table lives in a single config block in lib/index.js—update it once after an official price change and you are done.

It is built for developers who call the DeepSeek API frequently and want to keep a live eye on per-conversation spend, especially those juggling multiple models or operating in cost-sensitive peak/valley environments. No extra SDK or dashboard needed: install the plugin, restart dsh web, and the cost line appears.