dsh-cot-summerization: DeepSeek Harness Chain of Thought Summarization Plugin

Developed by MeowLynxSea, the DSH plugin `dsh-cot-summation` is designed to optimize reasoning display. Instead of directly displaying the raw Chain of Thought, it intercepts raw incremental data at the `llm/stream` layer and generates thumbnail summaries via a small model, which then flow into the UI as Reasoning blocks. This design ensures a clean interface while preserving the raw Chain of Thought within the model's visible history, avoiding interference with the Agent's multi-round reasoning. Core capabilities include streaming segmented summarization, dynamic chunking, near-duplicate elimination, and support for the Nekomimi boundary protocol. The plugin offers various error handling strategies and allows for the configuration of an independent small model channel. After installation, users can configure the enable status, summary style, and length limits via the Web Client. Although this incurs an additional cost for a small model call and requires the raw data to pass through the configured LLM channel, users can point to this...