knowledge-base

Personal Knowledge Base

A local knowledge base based on SQLite + FTS5, helping users save, organize, and retrieve knowledge snippets. Data is fully localized.

Use Cases

The user says "help me remember this" or "save this knowledge point"
The user says "have I saved anything about XX before" or "search my knowledge base"
The user says "organize the Python knowledge I've saved"
The user wants to accumulate knowledge cards during research/learning

Execution Method

Database Structure

The knowledge base is stored in ~/Documents/xiaodazi/knowledge.db：

CREATE TABLE knowledge (
    id INTEGER PRIMARY KEY AUTOINCREMENT,
    title TEXT NOT NULL,
    content TEXT NOT NULL,
    tags TEXT DEFAULT '',           -- 逗号分隔的标签
    source TEXT DEFAULT '',         -- 来源 URL 或文件
    created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
    updated_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP
);

CREATE VIRTUAL TABLE knowledge_fts USING fts5(
    title, content, tags,
    content='knowledge',
    content_rowid='id'
);

Core Operations

Save Knowledge：

import sqlite3

db = sqlite3.connect("~/Documents/xiaodazi/knowledge.db")
db.execute(
    "INSERT INTO knowledge (title, content, tags, source) VALUES (?, ?, ?, ?)",
    ["Python 装饰器原理", "装饰器是一个接受函数并返回函数的高阶函数...", "python,编程", ""]
)
db.execute(
    "INSERT INTO knowledge_fts (rowid, title, content, tags) VALUES (last_insert_rowid(), ?, ?, ?)",
    ["Python 装饰器原理", "装饰器是一个接受函数并返回函数的高阶函数...", "python,编程"]
)
db.commit()

Search Knowledge：

results = db.execute("""
    SELECT k.id, k.title, snippet(knowledge_fts, 1, '**', '**', '...', 30) as excerpt,
           k.tags, k.created_at
    FROM knowledge_fts
    JOIN knowledge k ON knowledge_fts.rowid = k.id
    WHERE knowledge_fts MATCH ?
    ORDER BY rank
    LIMIT 10
""", ["Python 装饰器"]).fetchall()

Browse by Tag：

results = db.execute("""
    SELECT title, substr(content, 1, 100) as preview, created_at
    FROM knowledge
    WHERE tags LIKE '%python%'
    ORDER BY updated_at DESC
""").fetchall()

Interaction Flow

用户：帮我记住——Python 的 GIL 是全局解释器锁，同一时刻只有一个线程执行字节码
→ 保存到知识库，自动提取标签 [python, 并发]
→ 回复：已保存 ✅「Python GIL」— 标签：python, 并发

用户：我之前存过什么关于 Python 的知识？
→ 搜索知识库
→ 找到 3 条记录：
  1. Python 装饰器原理（2/20）
  2. Python GIL（2/26）
  3. Python 异步编程（2/22）

Output Specifications

When saving, confirm the title and tags so the user knows how to find it again
Search results are sorted by relevance and show a summary preview
Tags are automatically extracted, and users can also specify them manually
Supports saving directly from the output of other Skills (such as the conclusions of in-depth research)