232 lines
4.9 KiB
Markdown
232 lines
4.9 KiB
Markdown
# AI Chat Converter - AI 助手完全指南
|
||
|
||
> **把这个文档发给 AI 助手，它就知道怎么帮你用这个工具了**
|
||
|
||
---
|
||
|
||
## 🎯 这是什么？
|
||
|
||
一个把 Codex、Kiro、Gemini、Claude 的聊天记录全部存到一个 SQLite 数据库的工具。
|
||
|
||
**数据库位置**: `项目目录/output/chat_history.db`
|
||
|
||
---
|
||
|
||
## 🚀 怎么启动？
|
||
|
||
### 方式一：双击启动（推荐）
|
||
```bash
|
||
./start.sh # Linux/macOS
|
||
start.bat # Windows（双击）
|
||
```
|
||
|
||
### 方式二：命令行
|
||
```bash
|
||
cd ai-chat-converter
|
||
python src/main.py --watch # 持续监控（推荐）
|
||
python src/main.py # 同步一次就退出
|
||
```
|
||
|
||
### 方式三：后台运行
|
||
```bash
|
||
nohup ./start.sh > /dev/null 2>&1 &
|
||
```
|
||
|
||
---
|
||
|
||
## 📊 数据库长啥样？
|
||
|
||
### 主表：sessions
|
||
|
||
| 字段 | 说明 | 例子 |
|
||
|------|------|------|
|
||
| file_path | 主键，文件路径 | `/home/user/.codex/sessions/xxx.jsonl` |
|
||
| session_id | 会话ID | `019b2164-168c-7133-9b1f-5d24fea1d3e1` |
|
||
| source | 来源 | `codex` / `kiro` / `gemini` / `claude` |
|
||
| cwd | 工作目录 | `/home/user/projects/myapp` |
|
||
| messages | 消息内容（JSON） | `[{"time":"...", "role":"user", "content":"..."}]` |
|
||
| start_time | 开始时间 | `2025-12-18T10:30:00` |
|
||
| token_count | Token 数量 | `1234` |
|
||
|
||
---
|
||
|
||
## 🔍 常用查询（直接复制用）
|
||
|
||
### 1. 看看有多少数据
|
||
|
||
```sql
|
||
SELECT source, COUNT(*) as 会话数, SUM(token_count) as Token总数
|
||
FROM sessions
|
||
GROUP BY source;
|
||
```
|
||
|
||
### 2. 最近的 10 个会话
|
||
|
||
```sql
|
||
SELECT session_id, source, cwd, start_time, token_count
|
||
FROM sessions
|
||
ORDER BY start_time DESC
|
||
LIMIT 10;
|
||
```
|
||
|
||
### 3. 搜索包含某个词的对话
|
||
|
||
```sql
|
||
SELECT session_id, source, cwd, start_time
|
||
FROM sessions
|
||
WHERE messages LIKE '%要搜索的词%'
|
||
ORDER BY start_time DESC
|
||
LIMIT 20;
|
||
```
|
||
|
||
### 4. 查某个项目的所有对话
|
||
|
||
```sql
|
||
SELECT session_id, source, start_time, token_count
|
||
FROM sessions
|
||
WHERE cwd LIKE '%项目名%'
|
||
ORDER BY start_time;
|
||
```
|
||
|
||
### 5. 看某个会话的完整内容
|
||
|
||
```sql
|
||
SELECT messages FROM sessions WHERE session_id = '会话ID';
|
||
```
|
||
|
||
### 6. 统计每天用了多少 Token
|
||
|
||
```sql
|
||
SELECT
|
||
date(start_time) as 日期,
|
||
SUM(token_count) as Token数
|
||
FROM sessions
|
||
GROUP BY 日期
|
||
ORDER BY 日期 DESC
|
||
LIMIT 7;
|
||
```
|
||
|
||
### 7. 统计每个来源的 Token
|
||
|
||
```sql
|
||
SELECT source, SUM(token_count) as tokens
|
||
FROM sessions
|
||
GROUP BY source
|
||
ORDER BY tokens DESC;
|
||
```
|
||
|
||
---
|
||
|
||
## 💻 命令行用法
|
||
|
||
| 命令 | 干啥的 |
|
||
|------|--------|
|
||
| `python src/main.py` | 同步一次 |
|
||
| `python src/main.py -w` | 持续监控（推荐） |
|
||
| `python src/main.py --stats` | 看统计信息 |
|
||
| `python src/main.py --search "关键词"` | 搜索 |
|
||
| `python src/main.py --export json` | 导出 JSON |
|
||
| `python src/main.py --export csv` | 导出 CSV |
|
||
| `python src/main.py --prune` | 清理已删除文件的记录 |
|
||
|
||
---
|
||
|
||
## 🐍 用 Python 查询
|
||
|
||
```python
|
||
import sqlite3
|
||
import json
|
||
|
||
# 连接数据库
|
||
db = sqlite3.connect('output/chat_history.db')
|
||
|
||
# 查所有 Codex 会话
|
||
for row in db.execute("SELECT session_id, cwd, token_count FROM sessions WHERE source='codex'"):
|
||
print(f"{row[0]}: {row[2]} tokens - {row[1]}")
|
||
|
||
# 搜索包含 "python" 的对话
|
||
for row in db.execute("SELECT session_id, source FROM sessions WHERE messages LIKE '%python%'"):
|
||
print(f"[{row[1]}] {row[0]}")
|
||
|
||
# 获取某个会话的消息
|
||
row = db.execute("SELECT messages FROM sessions WHERE session_id=?", ('会话ID',)).fetchone()
|
||
if row:
|
||
messages = json.loads(row[0])
|
||
for msg in messages:
|
||
print(f"{msg['role']}: {msg['content'][:100]}...")
|
||
```
|
||
|
||
---
|
||
|
||
## 📁 文件在哪？
|
||
|
||
```
|
||
ai-chat-converter/
|
||
├── start.sh ← 双击这个启动
|
||
├── output/
|
||
│ ├── chat_history.db ← 数据库在这
|
||
│ └── logs/ ← 日志在这
|
||
└── src/
|
||
└── main.py ← 主程序
|
||
```
|
||
|
||
---
|
||
|
||
## ❓ AI 助手任务示例
|
||
|
||
当用户说这些话时，你应该这样做：
|
||
|
||
| 用户说 | 你做 |
|
||
|--------|------|
|
||
| "帮我查最近的对话" | 执行最近会话 SQL |
|
||
| "搜索关于 Python 的讨论" | 用 `--search` 或 SQL 搜索 |
|
||
| "这个月用了多少 Token" | 执行 Token 统计 SQL |
|
||
| "导出所有 Codex 记录" | `python src/main.py --export json --source codex` |
|
||
| "启动监控" | `./start.sh` 或 `python src/main.py -w` |
|
||
| "数据库在哪" | `output/chat_history.db` |
|
||
|
||
---
|
||
|
||
## 🔧 出问题了？
|
||
|
||
### 问题：找不到数据库
|
||
```bash
|
||
# 先运行一次同步
|
||
python src/main.py
|
||
```
|
||
|
||
### 问题：依赖没装
|
||
```bash
|
||
pip install -r requirements.txt
|
||
```
|
||
|
||
### 问题：权限不够
|
||
```bash
|
||
chmod +x start.sh
|
||
```
|
||
|
||
---
|
||
|
||
## 📊 消息格式
|
||
|
||
数据库里的 `messages` 字段是 JSON 数组：
|
||
|
||
```json
|
||
[
|
||
{
|
||
"time": "2025-12-18T10:30:00",
|
||
"role": "user",
|
||
"content": "帮我写个 Python 脚本"
|
||
},
|
||
{
|
||
"time": "2025-12-18T10:30:05",
|
||
"role": "ai",
|
||
"content": "好的，这是一个简单的脚本..."
|
||
}
|
||
]
|
||
```
|
||
|
||
- `role`: `user`（用户）或 `ai`（AI 回复）
|
||
- `time`: ISO 格式时间
|
||
- `content`: 消息内容
|