Files
val-blog/scripts/memory_fts5.py
T

264 lines
6.9 KiB
Python
Executable File

#!/usr/bin/env python3
"""
FTS5 记忆索引系统
用于全文搜索 memory/*.md 文件
"""
import sqlite3
import os
import re
from pathlib import Path
from datetime import datetime
# 配置
MEMORY_DIR = Path.home() / ".openclaw" / "workspace" / "memory"
DB_PATH = MEMORY_DIR / ".fts5" / "memory.db"
def init_db():
"""初始化 FTS5 数据库"""
DB_PATH.parent.mkdir(parents=True, exist_ok=True)
conn = sqlite3.connect(DB_PATH)
cursor = conn.cursor()
# 创建文件表
cursor.execute("""
CREATE TABLE IF NOT EXISTS files (
id INTEGER PRIMARY KEY,
path TEXT UNIQUE NOT NULL,
modified_at TIMESTAMP,
indexed_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP
)
""")
# 创建 FTS5 虚拟表用于全文搜索
cursor.execute("""
CREATE VIRTUAL TABLE IF NOT EXISTS memory_index USING fts5(
content,
file_id,
chunk_index,
tokenize='porter unicode61'
)
""")
# 创建 chunks 表存储分块信息
cursor.execute("""
CREATE TABLE IF NOT EXISTS chunks (
id INTEGER PRIMARY KEY,
file_id INTEGER,
chunk_index INTEGER,
content TEXT,
start_line INTEGER,
end_line INTEGER,
FOREIGN KEY (file_id) REFERENCES files(id)
)
""")
conn.commit()
conn.close()
print(f"✓ 数据库初始化完成: {DB_PATH}")
def chunk_content(content, chunk_size=500, overlap=100):
"""
将内容分块,带重叠以保持上下文
Args:
content: 文件内容
chunk_size: 每块字符数
overlap: 重叠字符数
Returns:
list of (chunk_text, start_line, end_line)
"""
lines = content.split('\n')
chunks = []
current_chunk = []
current_size = 0
start_line = 0
for i, line in enumerate(lines):
line_with_newline = line + '\n'
if current_size + len(line_with_newline) > chunk_size and current_chunk:
# 保存当前块
chunk_text = ''.join(current_chunk)
chunks.append((chunk_text, start_line + 1, i))
# 重叠部分
overlap_text = chunk_text[-overlap:] if len(chunk_text) > overlap else chunk_text
current_chunk = [overlap_text]
current_size = len(overlap_text)
start_line = i - len(overlap_text.split('\n')) + 1
current_chunk.append(line_with_newline)
current_size += len(line_with_newline)
# 最后一块
if current_chunk:
chunk_text = ''.join(current_chunk)
chunks.append((chunk_text, start_line + 1, len(lines)))
return chunks
def index_file(file_path, conn=None):
"""索引单个文件"""
close_conn = False
if conn is None:
conn = sqlite3.connect(DB_PATH)
close_conn = True
cursor = conn.cursor()
# 读取文件
try:
with open(file_path, 'r', encoding='utf-8') as f:
content = f.read()
except Exception as e:
print(f"✗ 读取失败 {file_path}: {e}")
return
# 获取文件信息
stat = os.stat(file_path)
modified_at = datetime.fromtimestamp(stat.st_mtime)
rel_path = str(Path(file_path).relative_to(MEMORY_DIR.parent))
# 插入或更新文件记录
cursor.execute("""
INSERT OR REPLACE INTO files (path, modified_at, indexed_at)
VALUES (?, ?, CURRENT_TIMESTAMP)
""", (rel_path, modified_at))
file_id = cursor.lastrowid
# 删除旧索引
cursor.execute("DELETE FROM memory_index WHERE file_id = ?", (file_id,))
cursor.execute("DELETE FROM chunks WHERE file_id = ?", (file_id,))
# 分块并索引
chunks = chunk_content(content)
for chunk_idx, (chunk_text, start_line, end_line) in enumerate(chunks):
# 插入 chunks 表
cursor.execute("""
INSERT INTO chunks (file_id, chunk_index, content, start_line, end_line)
VALUES (?, ?, ?, ?, ?)
""", (file_id, chunk_idx, chunk_text, start_line, end_line))
# 插入 FTS5 索引
cursor.execute("""
INSERT INTO memory_index (content, file_id, chunk_index)
VALUES (?, ?, ?)
""", (chunk_text, file_id, chunk_idx))
conn.commit()
if close_conn:
conn.close()
print(f"✓ 已索引: {rel_path} ({len(chunks)} 块)")
def index_all():
"""索引所有 memory 文件"""
init_db()
conn = sqlite3.connect(DB_PATH)
# 查找所有 .md 文件
md_files = list(MEMORY_DIR.glob("*.md"))
print(f"\n开始索引 {len(md_files)} 个文件...")
for file_path in md_files:
if file_path.name.startswith('.'):
continue
index_file(file_path, conn)
conn.close()
print(f"\n✓ 索引完成")
def search(query, limit=10):
"""
全文搜索
Args:
query: 搜索关键词
limit: 返回结果数量
Returns:
list of dict with path, snippet, rank
"""
conn = sqlite3.connect(DB_PATH)
cursor = conn.cursor()
# 使用 FTS5 的 snippet 功能生成摘要
cursor.execute("""
SELECT
f.path,
snippet(memory_index, 0, '[', ']', '...', 32) as snippet,
c.start_line,
c.end_line,
rank
FROM memory_index
JOIN files f ON memory_index.file_id = f.id
JOIN chunks c ON memory_index.file_id = c.file_id
AND memory_index.chunk_index = c.chunk_index
WHERE memory_index MATCH ?
ORDER BY rank
LIMIT ?
""", (query, limit))
results = []
for row in cursor.fetchall():
results.append({
'path': row[0],
'snippet': row[1],
'start_line': row[2],
'end_line': row[3],
'rank': row[4]
})
conn.close()
return results
def print_search_results(query, results):
"""打印搜索结果"""
print(f"\n🔍 搜索: '{query}'")
print(f"找到 {len(results)} 条结果:\n")
for i, r in enumerate(results, 1):
print(f"{i}. {r['path']} (行 {r['start_line']}-{r['end_line']})")
print(f" {r['snippet']}")
print()
if __name__ == "__main__":
import sys
if len(sys.argv) < 2:
print("用法:")
print(f" {sys.argv[0]} index # 索引所有文件")
print(f" {sys.argv[0]} search <query> # 搜索")
sys.exit(1)
command = sys.argv[1]
if command == "index":
index_all()
elif command == "search":
if len(sys.argv) < 3:
print("错误: 需要提供搜索关键词")
sys.exit(1)
query = sys.argv[2]
results = search(query)
print_search_results(query, results)
else:
print(f"未知命令: {command}")
sys.exit(1)