AI Agent Hub
Back to plugins
🧠

dsh-memory

Memory Updated 2026.08.25

Run the following command in DeepSeek Harness:

dsh plugin install joyiok/dsh-memory

Paste the following prompt into your AI chat to install this plugin:

Run dsh plugin install joyiok/dsh-memory in your terminal to install the plugin; source code is available at https://github.com/joyiok/dsh-memory . Restart dsh web after installation to activate it.

About this plugin

Every new DeepSeek Harness session starts from a blank slate. dsh-memory maintains a single shared JSON store under $DSH_HOME that all sessions-whether web, headless, or a different working directory-read and write to, giving your agent genuine persistent context across sessions.

The core surface is four agent tools: memory_add writes a tagged memory entry, memory_search queries by content or tags, memory_list browses by category, and memory_remove deletes by id. On a fresh session the agent proactively searches existing memories and stores new facts on its own, so you stop repeating the same instructions every time. A sidebar Memory panel on the web UI lets you add, search, and delete entries by hand without writing any tool call.

Under the hood, writes use a temp-file-then-rename atomic pattern that stays safe across concurrent processes, and reads check mtime to pick up external updates. If the JSON file is corrupt, the plugin renames it with a .corrupt- suffix and rebuilds rather than crashing.

Built for developers who run multiple DSH sessions and want their agent to actually remember project conventions and preferences. No extra configuration needed-just keep secrets and credentials out of your memory entries.

Use Cases

  • Sharing project conventions and coding preferences across multiple DSH sessions
  • Agent automatically recalls tech stack and style from prior sessions
  • Manually adding, searching, and deleting memory entries via the web sidebar

Best For

  • Developers running multiple concurrent DSH sessions
  • Teams wanting their agent to retain long-term project context
  • Users who prefer managing agent memory through the web UI