dsh-simple-wiki-memory

When using AI assistants, long-term memory management often faces a dilemma: stuffing all memories into prompts leads to excessive token consumption, while heavy systems like RAG or vector libraries are overly complex. The DSWM plugin addresses this by offering a super-simplified Wiki memory solution built on DeepSeek Harness's native dsh-agent-instructions mechanism. It uses pure Markdown files and a git repository for memory management, automatically loading a lightweight index and reading topic files on demand to minimize token usage without sacrificing memory depth. Core capabilities include real-time writing, confirmation workflows, regular organization, and automatic backups, all governed by simple text rules without any runtime LLM calls or external dependencies. If you're a DeepSeek Harness user tired of complex and costly memory management, DSWM is the ideal choice. It suits developers or AI enthusiasts who prefer a lightweight, transparent, and customizable memory system, allowing you to easily control and share long-term memories across harnesses while focusing on core tasks rather than underlying complexities.