codex2md.py

#!/usr/bin/env python3
"""把 Codex 的 JSONL 对话记录转成可读 Markdown（增量追加）"""
import json, sys, os, glob
from datetime import datetime

CODEX_SESSIONS_DIR = os.path.expanduser('~/.codex/sessions')
RECORDS_DIR = os.path.join(os.getcwd(), 'records')  # 写到运行时的当前目录


def get_text_from_content(content):
    if isinstance(content, str):
        return content
    if isinstance(content, list):
        parts = []
        for block in content:
            if isinstance(block, dict):
                t = block.get('type', '')
                if t in ('input_text', 'output_text', 'text'):
                    parts.append(block.get('text', ''))
        return '\n'.join(p for p in parts if p)
    return ''


def state_path(jsonl_path):
    basename = os.path.basename(jsonl_path)
    parts = basename.replace('.jsonl', '').split('-')
    uid = parts[6] if len(parts) > 6 else basename.replace('.jsonl', '')
    return os.path.join(RECORDS_DIR, f'.codex-state-{uid}')


def read_offset(jsonl_path):
    sp = state_path(jsonl_path)
    if os.path.exists(sp):
        try:
            with open(sp) as f:
                return int(f.read().strip().split('\n')[0])
        except Exception:
            pass
    return 0


def write_offset(jsonl_path, offset):
    os.makedirs(RECORDS_DIR, exist_ok=True)
    with open(state_path(jsonl_path), 'w') as f:
        f.write(str(offset) + '\n')


def out_path_for(jsonl_path):
    basename = os.path.basename(jsonl_path)
    parts = basename.replace('.jsonl', '').split('-')
    try:
        day = parts[3].split('T')[0]
        date = f'{parts[1]}-{parts[2]}-{day}'
        uid = parts[6]
    except:
        date = datetime.fromtimestamp(os.path.getmtime(jsonl_path)).strftime('%Y-%m-%d')
        uid = basename[:8]
    return os.path.join(RECORDS_DIR, f'codex-{date}-{uid}.md')


def convert(jsonl_path, out_path):
    offset = read_offset(jsonl_path)

    messages = []
    msg_start_offsets = []
    current_pos = offset

    with open(jsonl_path, 'rb') as f:
        f.seek(offset)
        for raw in f:
            line_start = current_pos
            current_pos += len(raw)
            try:
                obj = json.loads(raw)
            except Exception:
                continue

            role = None
            text = ''
            ts = obj.get('timestamp', '')

            t = obj.get('type', '')
            payload = obj.get('payload', {})

            if t == 'event_msg' and payload.get('type') == 'user_message':
                role = 'user'
                text = payload.get('message', '').strip()

            elif t == 'response_item':
                item_role = payload.get('role', '')
                if item_role == 'assistant':
                    role = 'assistant'
                    text = get_text_from_content(payload.get('content', ''))

            if role and text.strip():
                if ts:
                    try:
                        dt = datetime.fromisoformat(ts.replace('Z', '+00:00'))
                        ts = dt.astimezone().strftime('%H:%M:%S')
                    except:
                        ts = ''
                messages.append((role, ts, text.strip()))
                msg_start_offsets.append(line_start)

    new_offset = current_pos

    if messages and messages[-1][0] == 'user':
        new_offset = msg_start_offsets[-1]
        messages.pop()

    if not messages:
        return

    lines = []
    for role, ts, text in messages:
        prefix = f'❯ [{ts}] ' if role == 'user' else f'⏺ [{ts}] '
        if not ts:
            prefix = '❯ ' if role == 'user' else '⏺ '
        lines.append(prefix + text.replace('\n', '\n  '))
        lines.append('')

    os.makedirs(RECORDS_DIR, exist_ok=True)
    mode = 'a' if os.path.exists(out_path) else 'w'
    with open(out_path, mode) as f:
        f.write('\n'.join(lines) + '\n')

    write_offset(jsonl_path, new_offset)
    print(f'已追加 {len(messages)} 条到: {out_path}')


def catchup():
    if not os.path.exists(CODEX_SESSIONS_DIR):
        print('未找到 Codex sessions 目录')
        return
    files = glob.glob(os.path.join(CODEX_SESSIONS_DIR, '**/*.jsonl'), recursive=True)
    for jsonl in files:
        out = out_path_for(jsonl)
        offset = read_offset(jsonl)
        if offset < os.path.getsize(jsonl):
            convert(jsonl, out)


if __name__ == '__main__':
    if '--catchup' in sys.argv:
        catchup()
        sys.exit(0)

    args = [a for a in sys.argv[1:] if not a.startswith('--')]

    if args:
        jsonl = args[0]
        out = args[1] if len(args) > 1 else out_path_for(jsonl)
    else:
        files = glob.glob(os.path.join(CODEX_SESSIONS_DIR, '**/*.jsonl'), recursive=True)
        if not files:
            print('没有找到 Codex 对话文件')
            sys.exit(1)
        files.sort(key=os.path.getmtime, reverse=True)
        jsonl = files[0]
        out = out_path_for(jsonl)

    convert(jsonl, out)