When building agents, conversational models often face context window limitations and factual hallucinations. DeepSeek Harness (DSH) adopts the “everything-as-plugin” concept, using modular components to extend agents’ capability boundaries.

What Is It?

satan9394/dsh-rag is a DSH skill focused on Retrieval-Augmented Generation (RAG). It was inspired by the wshobson/agents project (38k★). The plugin aims to provide agents with core capabilities based on vector stores and document retrieval.

Core Capabilities

According to the project description, the plugin covers the following capabilities:

  • Vector stores and embeddings
  • Document pipelines
  • Retrieval strategies
  • Hallucination governance

Installation and Enablement

No official installation command is currently provided. Because fetching the README failed, the specific installation steps cannot be verified.

Typical Usage

The documentation does not provide specific usage examples.

Applicable Scenarios and Precautions

  • The plugin runs with the permissions of the current DSH process.
  • Check the source code and license before installing.
  • Because the information is incomplete, please confirm the feature details in the GitHub repository yourself.

Summary

This is a collection of RAG capabilities for DeepSeek Harness, designed to improve agents’ information accuracy and answer quality through Retrieval-Augmented Generation.

  • Catalog page: https://www.skillhub.cn/plugins/satan9394/dsh-rag
  • GitHub: https://github.com/satan9394/dsh-rag