dsh-memory-palace

When using AI assistants, we often face the problem of losing memory after a session ends: every time a new conversation starts, the assistant has to learn your project context and preferences from scratch, which significantly hampers efficiency. The dsh-memory-palace plugin is designed to solve this pain point by bringing WorkBuddy's file-based memory system into DeepSeek Harness, providing a human-readable and directly editable Markdown memory library for AI.

The core capability of this plugin is enabling persistent, cross-session memory. It stores all information in simple Markdown files (like MEMORY.md and daily logs), allowing you to view and modify them with any text editor. It supports a dual-layer memory structure—user-level preferences and project-level conventions—and automatically injects relevant memory into each conversation turn, ensuring contextual continuity. Additionally, the plugin offers intelligent automatic recording, log migration, and memory management tools, and can even seamlessly bridge data from existing WorkBuddy projects, making memory maintenance effortless and efficient.

dsh-memory-palace is ideal for users who want their AI assistant to truly remember work details and reduce repetitive communication. Whether you're a developer needing to maintain project context or a researcher hoping for the AI to continuously accumulate domain knowledge, it helps you build a more focused and coherent smart assistant experience. The built-in graphical settings panel also lets you easily adjust memory behavior without any technical background.