dsh-cost-tracker

When using DeepSeek Harness for large language model interactions, users often struggle with tracking token costs accurately. Pricing variations across different models, changes in peak and off-peak rates, and the complexities of caching mechanisms make precise expense accounting challenging. The dsh-cost-tracker plugin emerges to address this issue, offering a transparent and configurable cost management solution.

Its core capability lies in granular token cost tracking. It automatically calculates the cost of each finalized LLM call based on a user-configurable pricing table, covering cache-hit and cache-miss inputs, output tokens, and optional peak-window rates. Simultaneously, the plugin publishes a per-session cost projection, providing a live readout of total costs, peak/off-peak breakdowns, and per-model details for intuitive expense monitoring. Additionally, it intelligently flags calls within configured peak windows and surfaces unconfigured models to prevent hidden costs, ensuring billing accuracy.

With cross-session billing persistence, dsh-cost-tracker ensures continuity in cost data, making it suitable for developers, data scientists, or enterprise administrators who need precise oversight of API usage costs. Whether optimizing budgets for personal projects or managing large-scale model calls for teams, this plugin helps you make informed decisions and effectively control expenses.