Introduction

The DeepSeek Harness (DSH) system extends the capability boundaries of agents through a plugin mechanism. When developing agents, you often encounter scenarios where the agent needs to understand local codebases or project documentation. Directly calling external LLM APIs for retrieval can introduce latency, cost, and privacy risks. A localized retrieval solution is a more practical choice.

Plugin Introduction

dsh-plugin-rag is a local knowledge base RAG plugin maintained by YYTbit. It solves the following problem: without external APIs or embedding models, it performs local TF-IDF keyword retrieval directly on project files, enabling agents to make decisions based on local data.

Core Features

  1. File Indexing: Supports indexing Markdown files in a project, various code files (such as .ts, .js, .py), configuration files (such as .json, .yaml), and documentation files.
  2. Keyword Search: Provides keyword retrieval capabilities to help agents quickly locate relevant context.
  3. Pure Local TF-IDF: Does not depend on external APIs or embedding models, has no heavy dependencies, and performs keyword matching entirely locally.

Installation and Enablement

To install this plugin in DSH, use the official installation command:

dsh plugin --profile your-profile add dsh-plugin-rag

Typical Usage

After installation, you can directly use the rag command in the command line:

  • Search related files:
rag "authentication" ./src
  • Display index statistics information:
rag --index ./src ./docs

Use Cases and Notes

This plugin is suitable for agent scenarios that need to search local codebases for documentation, code snippets, or configuration information. Before installing and using it, ensure that you have reviewed the plugin’s source code and understand its license (MIT). The plugin runs under the permissions of the current DSH process.