DSH-Self-Improved: DeepSeek Harness Local Long-Term Memory and Self-Evolution Plugin

`dsh-self-improved` is a long-term memory and self-evolution plugin for DeepSeek Harness, designed to address the pain point of manually maintaining context across multi-session conversations. This plugin automatically extracts key information from conversations through a four-layer architecture (conversation capture, memory extraction, scenario summarization, and user profiling), stores it in a local SQLite database (utilizing FTS5 and sqlite-vec for retrieval), and injects relevant memories before subsequent conversations. Its core capabilities include automatic memory extraction and integration, memory decay correction, distilling successful workflows into reusable `dsh-skill` skills, and a dynamic user profile that evolves with conversations. The plugin provides the `/memory` command for zero-cost memory queries, as well as a settings panel for users to manage, filter, and correct memories. Installation can be done via npm or G