AI Agent Hub
Back to plugins
🧠

dsh-persist

Memory Updated 2026.08.25

Run the following command in DeepSeek Harness:

dsh plugin install tluoluo/dsh-persist

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

To install the dsh-persist plugin in DeepSeek Harness, use the command dsh plugin install tluoluo/dsh-persist, with the source code available at https://github.com/tluoluo/dsh-persist.

About this plugin

DeepSeek Harness agents typically suffer from amnesia, losing all context as soon as a conversation ends. dsh-persist solves this by equipping agents with "persistent memory," constructing a layered file system to ensure key information is permanently saved rather than discarded.

Beyond basic persistence, dsh-persist offers powerful "selective injection" and semantic retrieval capabilities. It organizes memory into a smart multi-layer structure, from user personas to project experience and specific conversation details. Users can flexibly control which memories are active in the current session, defaulting to zero injection to keep the context clean. Additionally, it supports Vault semantic memory, allowing agents to "recall" information via vector search (with optional SiliconFlow integration) or keyword search, ensuring they always have the right historical context at hand.

This plugin is perfect for developers and researchers who require long-term context continuity, especially those managing multi-project workflows. If you want your DeepSeek agent to remember past conversations, understand project context, and automatically retrieve relevant materials for complex tasks, dsh-persist is an essential tool. It transforms agents from one-time helpers into intelligent partners with long-term memory.

Use Cases

  • Handling complex tasks requiring long-term context continuity
  • Managing multi-project development workflows in the same directory
  • Maintaining key information without loss during long conversations

Best For

  • Developers needing to build AI agents with long-term memory
  • Engineers wanting to share context across parallel projects
  • Researchers requiring reliable information retrieval and persistent storage