Function process_level_async

mainprogress/determine_toc_levels.py:321–407

处理除第一张以外的其他图片，使用首图的 CSV 结果作为 Few-shot 上下文。

(semaphore: asyncio.Semaphore, img_file: Path, csv_file: Path, output_path: Path)

            content_list.extend([
                {"type": "text", "text": "参考示例（第一页图片及其正确的 CSV 格式层级分析结果）："},
                {"type": "image_url", "image_url": {"url": FIRST_PAGE_EXAMPLE["image_base64"]}},
                {"type": "text", "text": f"参考结果 (CSV 格式):\n{FIRST_PAGE_EXAMPLE['result_csv_str']}"},
                {"type": "text", "text": "---\n请严格参照上述示例的 CSV 格式和层级判断标准，分析以下当前页图片："}
            ])
        else:
            write_log(f"警告：未找到首图示例，将无参考处理 {img_file.name}")
        
        # 添加当前页图片和提示词
        content_list.extend([
            {"type": "image_url", "image_url": {"url": get_encoded_image(img_file)}},
            {"type": "text", "text": f"{PROMPT_TEXT}\n\n当前页提取的原始 CSV 数据如下：\n{csv_content}"}
        ])

        messages = [{"role": "user", "content": content_list}]

        try:
            for attempt in range(MAX_RETRIES):
                try:
                    response = await client.chat.completions.create(
                        model=LLM_CONFIG["model"],
                        messages=messages,
                        extra_body={"enable_thinking": False},
                        timeout=REQUEST_TIMEOUT,
                        temperature=0,
                    )

                    content = response.choices[0].message.content.strip()

                    # 本地解析 CSV 转为 JSON
                    try:
                        parsed_data = parse_csv_response(content, img_file.name)
                    except Exception as parse_err:
                        write_log(f"第 {attempt+1} 次尝试解析 CSV 失败：{parse_err}")
                        if attempt == MAX_RETRIES - 1:
                            return False
                        await asyncio.sleep(2 ** attempt)
                        continue

                    if parsed_data:
                        sorted_data = sorted(parsed_data, key=lambda x: x['number'])

                        with open(output_file, 'w', encoding='utf-8') as f:
                            json.dump(sorted_data, f, ensure_ascii=False, indent=2)

                        print(f"已判断层级：{img_file.name}")
                        return True
                    else:
                        write_log(f"解析结果为空：{img_file.name}")
                        if attempt == MAX_RETRIES - 1:
                            return False
                        await asyncio.sleep(2 ** attempt)

                except (APIError, Timeout) as e:
                    write_log(f"第 {attempt+1} 次 API 请求失败 ({type(e).__name__}): {str(e)}")
                    if attempt == MAX_RETRIES - 1:
                        return False
                    await asyncio.sleep(2 ** attempt)
                except Exception as e: