docindex: A Local Document Semantic Index Plugin for DeepSeek Harness

johnxu22786/docindex

Data & KnowledgeVerified
Listed on 2026-08-20
Page last updated 2026-09-30

A DSH plugin that builds a local semantic index over workspace documents, enabling hybrid retrieval with line-numbered citations and incremental updates for natural language or keyword queries.

docindex is a document indexing plugin for DeepSeek Harness that turns workspace files like Markdown, PDF, and DOCX into a searchable local knowledge base. It combines SQLite FTS5 lexical retrieval with local embedding semantic retrieval, fusing results via RRF, with CJK tokenization, incremental updates, and line-numbered citations. Models can query documents through four tools, or you can use the standalone CLI. Ideal for DSH users who need to quickly locate document content.

How to Install

install
dsh plugin --profile web add github:johnxu22786/docindex
Category
Data & Knowledge
Platform
DSH Plugin
Author
johnxu22786
Distribution
Plugin

docindex Key Features

Local semantic indexHybrid retrieval & rankingLine-numbered citationsIncremental updates & watch
Listed on dsh-plugin.org

docindex Repo Summary

What Does It Do?

docindex is a DSH plugin for DeepSeek Harness that turns a workspace into a searchable local knowledge base by indexing Markdown, plain text, PDF, DOCX, PPTX, and XLSX files. It solves the problem of quickly locating content in documents, letting you query with natural language or keywords and get matched documents, snippets with exact line numbers, and relevance scores. Core capabilities include hybrid retrieval (SQLite FTS5 lexical + local embedding semantic), incremental updates, and model-facing tools.

Core Features

  • Hybrid retrieval: combines SQLite FTS5 (with CJK n-gram tokenizer) and a local embedder, fused via Reciprocal Rank Fusion (RRF).
  • Incremental updates: only changed files are re-indexed, with optional file watcher and full rebuild fallback.
  • Line-numbered citations: results include exact line numbers for precise reference.
  • Model tools: provides doc_scan, doc_query, doc_reindex, and doc_stats for agent use.
  • Standalone CLI: offers docindex commands for scanning, querying, rebuilding, and stats.

How to Use This Plugin?

After enabling it in DSH, the model can query documents via tools, e.g., call doc_query for "RLHF and llama.cpp" to get hits with path, line, snippet, and score. Alternatively, use the standalone CLI: run docindex scan to index the current directory, docindex query "keyword" to search, and docindex reindex --full for a full rebuild. Configuration options include root directories, database path, embedding provider (default local, optional transformers), and exclusion rules.

This page is an independent rewrite of the plugin's official README — for authoritative documentation and the latest changes, refer to the source: johnxu22786/docindex. The plugin is third-party code that runs on your machine once installed; inclusion does not imply endorsement — please review the source before installing.

Install, update, and uninstall this plugin in the Plugin Market of DeepSeek Harness's DSH Plugin Hub

DSH Plugin FAQ