Introduction

The design philosophy of DeepSeek Harness (DSH) is “everything is a plugin.” In local development scenarios, agents often lack the ability to understand codebase structure and cannot perform efficient code retrieval locally. The dsh-code-intel plugin addresses this issue through symbol-aware indexing and a hybrid search mechanism.

Plugin Positioning

  • Name: dsh-code-intel
  • Author: lonelymoon87
  • Category: Memory
  • Core features: symbol-aware code search, code outline parsing, background indexing, SQLite persistence, file system watching, hybrid search (lexical and semantic), multi-language support (TS, JS, Python, Go, Rust, Java)

Installation

Before installing, make sure the DSH version is >=0.1.0-rc.6 <0.2.0 and the Node.js version is ^22.19 || >=24.

dsh plugin --profile web add https://github.com/lonelymoon87/dsh-code-intel/releases/download/v0.1.3/dsh-code-intel-0.1.3.tgz

Usage Workflow

This tool uses an asynchronous indexing mechanism. On first use, you must interact using a Job ID.

  1. When code_search or directory code_outline is called for the first time, the tool returns an indexing result and a job id.
  2. Use the built-in jobs tool to check or wait for the task to complete.
  3. After the task completes, call the code tool again to obtain the search or outline results.

Configuration

By default, the plugin uses lexical mode and does not require network requests. To enable hybrid search mode, configure an OpenAI-compatible embedding endpoint.

Lexical mode configuration (default):

- id: code-intel
  name: dsh-code-intel
  config:
    indexDir: .dsh/code-index
    include: [.ts, .tsx, .js, .jsx, .py, .go, .rs, .java]
    exclude: [.dsh, .git, node_modules, dist, build, coverage, vendor]
    maxFileSize: 1000000
    maxChunkChars: 12000
    maxResults: 20
    watch: true
    embedding: false

Hybrid mode configuration:

    embedding:
      provider: openai-compatible
      endpoint: https://embedding.example/v1/embeddings
      model: your-embedding-model
      credentialRef: EMBEDDING_API_KEY
      batchSize: 32

Note: indexDir and excluded entries must be relative to the workspace. Only extensions supported by the listed parsers are accepted.

Technical Details and Limitations

  • Cache location: ~/.dsh/code-index/ (workspace-relative path).
  • Node version: Node 22+ is required because the built-in SQLite module is experimental.
  • Credential handling: Credentials in the configuration are references, not actual keys in YAML. The plugin resolves the references at runtime through ctx.credentials.
  • Data privacy: No telemetry, and no custom persistent session events are registered.
  • API limitations: DeepSeek’s official API does not expose a public embedding endpoint, so the plugin does not fabricate endpoints or silently label lexical results as semantic results.

Ecosystem Information

  • Directory page: https://www.skillhub.cn/plugins/lonelymoon87/dsh-code-intel
  • Source code: https://github.com/lonelymoon87/dsh-code-intel
  • License: MIT