dsh-plugin-memos-code-retrospect: AI Coding Conversation Retrospective Distillation Plugin

The DSH (DeepSeek Harness) plugin `dsh-plugin-memos-code-retrospect` is designed to address issues in AI coding tasks, such as forgetting historical context, repeatedly encountering the same pitfalls, and losing engineering constraints. It replicates mneme’s retrospective logic by extracting rejected solutions, pitfall records, and engineering constraints at the end of a conversation, tagging them, and writing them into the MemOS memory foundation; before reasoning about a coding task, it performs weighted recall and injects the relevant items into the context to guide the AI in avoiding past pitfalls. Core features include extracting entities from the LLM at the end of a conversation, weighted retrieval for coding tasks, and a robust v2 implementation that supports fallback retry and deduplication filtering. Users need to install the MemOS local plugin, configure `cordis.patch.yml`, and restart the server.