dsh-token-attention

dsh-token-attention addresses the challenge that DeepSeek Harness users often face in tracking and optimizing token consumption and costs during daily tasks. By automatically logging token usage details for every conversation and integrating peak/off-peak pricing models, it provides clear insights into resource flows, helping avoid unexpected expenses.

The plugin's core capability lies in its multi-faceted management panel: featuring five views (Overview, Sessions, Tasks, Cost, and Settings), it aggregates data by day, week, or month, and includes a task-type recognition and advice engine. This engine offers specific optimization suggestions—such as execution windows and session switching—based on factors like context occupancy, hit rate, and cost.

It is especially suited for DSH users who need fine-grained control over token resources, such as developers, AI researchers, or teams frequently using DeepSeek models, enabling them to make data-driven decisions while pursuing efficiency.