model-usage-plugin

In the daily use of AI model inference, tracking token consumption and costs from API calls can be challenging, especially with high-frequency or multi-model scenarios, where cost management becomes critical. The model-usage-plugin offers an efficient solution for DeepSeek Harness users, enabling easy monitoring and optimization of resource usage.

The plugin's core capabilities revolve around comprehensive tracking and estimation. It provides detailed statistics for each model, distinguishing between cache hits, misses, and writes, and automatically estimates costs based on built-in pricing for popular models. For DeepSeek models, it supports flexible peak-valley pricing, allowing you to customize peak periods and prices to align with different billing strategies. Additionally, integrated currency conversion ensures costs are uniformly displayed in a target currency (like RMB), while real-time display of DeepSeek account balances makes resource management more intuitive.

The model-usage-plugin is ideal for developers, data scientists, and enterprise teams using the DeepSeek API. If you need to transparentize costs, control budgets, or maintain efficient resource monitoring during frequent AI model interactions, this plugin serves as a valuable tool. Through automated statistics and a user-friendly interface, it streamlines cost tracking, letting you focus more on core model inference and application development.