yo-docs-db

Document knowledge base powered by jieba tokenization + SQLite inverted index. Supports mixed Chinese/English keyword search.

Install

npm install

Dependencies: @node-rs/jieba, Node.js 22+ (built-in node:sqlite).

Usage

import { YoDocsDB } from 'yo-docs-db';

const docs = new YoDocsDB({
  dbPath: './.yo_ddb/data/docs.db',
  docsDir: './.yo_ddb/docs',
});

// Write document (auto tokenize + build inverted index)
await docs.write({
  type: 'frontend',
  lang: 'vue',
  question: 'How to use Font Awesome in Vue',
  doc_name: 'fontawesome-usage',
  content: '# Font Awesome\n...',
});

// Keyword search (sorted by match count, returns document content)
const { results } = await docs.query({
  type: 'frontend',
  lang: 'vue',
  query: 'fontawesome icon',
  limit: 5,
});

// List documents (optional type/lang filter)
docs.list({ type: 'frontend', lang: 'vue' });

// Delete by ID (auto-cleanup orphaned keywords)
docs.delete({ id: 'uuid-xxx' });

// Tokenize text (expose underlying capability)
docs.tokenize('Vue Router how to configure');

// Release resources
docs.close();

API

| Method | Description | |--------|-------------| | write(params) | Write document + build index, returns { id, keywords } | | query(params) | Keyword search, sorted by match_count desc | | list(params?) | List documents, type/lang optional | | delete(params) | Delete by ID, auto-cleanup orphaned keywords | | getById(id) | Get document metadata by ID | | tokenize(text) | jieba tokenization (stopwords/single-char/dedup) | | close() | Close database connection |

Data Structure

SQLite tables:

• documents — document records (id, type, lang, question, doc_path)
• keywords — dictionary (deduplicated tokens)
• doc_keywords — inverted index (document↔keyword mapping)

Demo

node --experimental-strip-types demo.ts

Generates sample data in .yo_ddb/data/docs.db and .yo_ddb/docs/.