dsh-project-wiki

When using the DeepSeek Harness AI Agent for software development, a persistent challenge is preserving and inheriting project knowledge. Insights like architectural decisions, design constraints, and debugging conclusions often remain trapped in individual sessions, making them hard for new sessions to access. Reloading entire chat histories is costly and can introduce outdated information. DSH Project Wiki addresses this by providing an auditable, workspace-local Wiki system where such knowledge is systematically captured.

At its core, the plugin stores knowledge as Markdown pages within the workspace, ensuring transparency and version control. Agents can propose new knowledge, but it must pass user review before becoming an official fact. Features include full-text search via SQLite FTS5, bidirectional linking with a knowledge graph, and staleness detection that flags pages needing updates when source files change. This approach keeps the knowledge base both useful and trustworthy.

DSH Project Wiki is ideal for development teams and individual developers using DeepSeek Harness who need to maintain a long-term knowledge repository. It benefits those managing project documentation, design rationales, or coding standards by localizing knowledge integration into the development workflow, enhancing collaboration, and reducing the risk of knowledge loss.