cc_export.py

#!/usr/bin/env python3
"""
Claude Code Transcript Exporter
用法: python3 cc_export.py <session.jsonl> [output.md]

将 Claude Code 的会话 JSONL 文件导出为带目录、可折叠的 Markdown 文档。
包含: 用户消息、助手回复、Thinking、工具调用(输入+输出)。

路径: ~/.claude/projects/<project-hash>/<session-id>.jsonl

例子：
# 导出当前会话
python3 /tmp/cc_export.py \
  ~/.claude/projects/<project>/<session>.jsonl \
  导出.md

# 默认输出到同名 .md
python3 /tmp/cc_export.py session.jsonl

"""

import json, sys, re, os

# ── 配置 ──────────────────────────────────────────────

TOOL_RESULT_MAX = 5000   # 工具输出截断长度
THINKING_COLLAPSED = True  # 默认折叠 Thinking

# ── 辅助函数 ──────────────────────────────────────────

def esc_md(s):
    """转义 Markdown 特殊字符"""
    return s.replace('|', '\\|').replace('<', '&lt;').replace('>', '&gt;')

def fmt_json(obj):
    """将 JSON 对象格式化为紧凑的可读字符串"""
    try:
        if isinstance(obj, str):
            obj = json.loads(obj)
        s = json.dumps(obj, indent=2, ensure_ascii=False)
        if len(s) > TOOL_RESULT_MAX:
            s = s[:TOOL_RESULT_MAX] + '\n\n... *(truncated)*'
        return s
    except:
        return str(obj)

def extract_content_blocks(content):
    """从助手消息的 content 数组中提取各类块"""
    blocks = []
    for item in (content if isinstance(content, list) else []):
        if not isinstance(item, dict):
            continue
        bt = item.get('type', '')
        if bt == 'thinking':
            blocks.append(('thinking', item.get('thinking', '')))
        elif bt == 'text':
            blocks.append(('text', item.get('text', '')))
        elif bt == 'tool_use':
            blocks.append(('tool_use', item))
    return blocks

# ── 主程序 ────────────────────────────────────────────

def main(input_file, output_file=None):
    if not os.path.exists(input_file):
        print(f"Error: {input_file} 不存在")
        sys.exit(1)

    if output_file is None:
        output_file = os.path.splitext(input_file)[0] + '.md'

    with open(input_file, 'r') as f:
        lines = f.readlines()

    out = []
    out.append("# Claude Code 会话 Transcript\n\n")
    out.append(f"**源文件**: `{input_file}`  \n")
    out.append(f"**记录数**: {len(lines)}\n\n")
    out.append("---\n\n")

    msg_idx = 0
    tool_idx = 0

    # 第一遍: 收集所有 tool_use → tool_result 配对
    tool_pool = {}  # tool_use_id -> (input_data, output_data)

    for line in lines:
        try:
            obj = json.loads(line)
        except:
            continue
        t = obj.get('type', '')

        # 收集工具结果
        if t == 'user' and obj.get('toolUseResult'):
            msg = obj.get('message', {})
            for block in (msg.get('content') if isinstance(msg.get('content'), list) else []):
                if isinstance(block, dict) and block.get('type') == 'tool_result':
                    tid = block.get('tool_use_id', '')
                    result = block.get('content', '')
                    if tid not in tool_pool:
                        tool_pool[tid] = [None, None]
                    tool_pool[tid][1] = result

        # 收集工具调用
        elif t == 'assistant':
            msg = obj.get('message', {})
            for block in (msg.get('content') if isinstance(msg.get('content'), list) else []):
                if isinstance(block, dict) and block.get('type') == 'tool_use':
                    tid = block.get('id', '')
                    if tid not in tool_pool:
                        tool_pool[tid] = [None, None]
                    tool_pool[tid][0] = block

    # 第二遍: 生成 Markdown
    for line in lines:
        try:
            obj = json.loads(line)
        except:
            continue
        t = obj.get('type', '')

        # ── 用户消息 ──
        if t == 'user' and not obj.get('toolUseResult'):
            msg = obj.get('message', {})
            content = msg.get('content', '')
            if isinstance(content, str) and content.strip():
                msg_idx += 1
                out.append(f"---\n\n## 👤 {msg_idx}\n\n{content}\n\n")

        elif t == 'queue-operation' and obj.get('operation') == 'enqueue' and msg_idx == 0:
            content = obj.get('content', '')
            if isinstance(content, str) and content.strip():
                msg_idx += 1
                out.append(f"---\n\n## 👤 {msg_idx}\n\n{content}\n\n")

        # ── 助手消息 ──
        elif t == 'assistant':
            msg = obj.get('message', {})
            blocks = extract_content_blocks(msg.get('content', []))
            if not blocks:
                continue

            out.append(f"### 🤖 {msg_idx}\n\n")

            for btype, bdata in blocks:
                if btype == 'thinking':
                    if THINKING_COLLAPSED:
                        out.append(f'<details>\n<summary>💭 Thinking ({len(bdata)} chars)</summary>\n\n')
                        out.append(bdata + '\n')
                        out.append('</details>\n\n')
                    else:
                        out.append(bdata + '\n\n')

                elif btype == 'text':
                    if bdata.strip():
                        out.append(bdata.strip() + '\n\n')

                elif btype == 'tool_use':
                    tool_idx += 1
                    name = bdata.get('name', '???')
                    inp = bdata.get('input', {})
                    tid = bdata.get('id', '')

                    # Input
                    inp_lines = []
                    for k, v in inp.items():
                        vs = str(v)
                        if len(vs) > 200:
                            vs = vs[:200] + '...'
                        inp_lines.append(f"  - **{k}**: `{esc_md(vs)}`")
                    inp_text = '\n'.join(inp_lines) if inp_lines else '  *(无参数)*'

                    # Output
                    result_text = ''
                    if tid in tool_pool and tool_pool[tid][1]:
                        result_text = fmt_json(tool_pool[tid][1])

                    out.append(f'''<details>
<summary>🔧 <b>{name}</b> #{tool_idx}</summary>

**Input:**
{inp_text}

''')
                    if result_text:
                        out.append(f'''**Output:**
```
{result_text}
```
''')
                    out.append('</details>\n\n')

    # ── 整理 ──
    text = re.sub(r'\n{4,}', '\n\n\n', ''.join(out))

    # Insert TOC after header
    toc = ["## 目录\n\n"]
    for i in range(1, msg_idx + 1):
        toc.append(f"- [👤 {i}](#{i})\n")
    toc_text = ''.join(toc)
    insert_pos = text.find('---\n\n')
    if insert_pos > 0:
        text = text[:insert_pos + 5] + '\n' + toc_text + '\n---\n\n' + text[insert_pos + 5:]

    with open(output_file, 'w') as f:
        f.write(text)

    # 统计
    tool_count = sum(1 for line in text.split('\n') if line.strip().startswith('<summary>🔧'))
    print(f"✅ {output_file}")
    print(f"   {len(text):,} bytes | {msg_idx} 轮对话 | {tool_count} 个工具调用")
    print(f"   工具输入+输出: {'✅' if any('**Output:**' in text for _ in [1]) else '⚠️'}")

if __name__ == '__main__':
    if len(sys.argv) < 2:
        print(__doc__)
        print("用法: python3 cc_export.py <session.jsonl> [output.md]")
        sys.exit(1)
    main(sys.argv[1], sys.argv[2] if len(sys.argv) > 2 else None)