In the DeepSeek Harness (DSH) plugin ecosystem, memory modules are key components for building long-term memory agents. Traditional memory solutions often rely on LLM calls to retrieve or associate information, leading to high costs or high latency. Here, we introduce a plugin focused on knowledge graphs and memory management, which provides a way to build memory without relying on LLMs.

Plugin Positioning

This plugin is named satan9394/dsh-knowledge-brain and belongs to the memory category. It is maintained by developer satan9394. It mainly addresses issues related to how agents handle complex knowledge associations, long-term memory storage, and cross-session memory isolation.

Core Features

Based on verified facts, the plugin has the following capabilities:

  • Synthetic answers with citations: When generating synthetic answers, it can automatically link and cite specific content from the knowledge base, improving answer traceability.
  • Self-linking knowledge graph with zero LLM calls: It builds a knowledge graph in which nodes automatically form connections, and this process does not require LLM API calls.
  • Gap analysis: It can analyze the differences between current knowledge and the target state.
  • 24/7 Dream Cycles: It keeps the memory system running continuously.
  • Company Brain login isolation: It provides data isolation across different login environments.

Installation and Enablement

No official installation command is provided in the reference materials. Please visit the plugin directory for specific installation instructions.

Use Cases and Considerations

  • Use cases: Suitable for enterprise-grade or complex agent scenarios that require long-term memory, knowledge association analysis, and data isolation.
  • Note: This plugin runs with the permissions of the current DSH process. Before installing it, inspect the source code to ensure it meets security requirements and verify the license information.

Summary

This plugin builds knowledge graphs with zero LLM calls, making it suitable for development scenarios that require a stable memory layer. For more information, refer to the plugin directory or the source code repository.