diff --git a/README.md b/README.md
index 7bb8043..b414ef0 100644
--- a/README.md
+++ b/README.md
@@ -82,7 +82,7 @@ pip install numpy pandas opencv-python pillow openai
 3. 复制 `example_config.json` 为 `poster_gen_config.json`，并根据你的 API Key、模型端点、文件路径、图片根目录等进行修改。确保 `prompts_dir` 下的 `Style/` 和 `Demand/` 目录包含选题生成 AI 可能输出的文件名 (带 `.txt` 后缀)。确保 `Refer/` 目录下包含所有需要的参考文件。
 4. **运行完整流程**: 
    ```bash
-   python main.py
+   python main.py --config poster_gen_config.json
    ```
 5. **分阶段运行**: (参考 `examples/` 目录下的脚本，**注意需要传递 run_id**)
    ```bash
@@ -235,4 +235,39 @@ finally:
 
 这个模式允许你灵活地利用流式输出，同时在需要时也能访问到最终的完整文本。参考 `examples/test_stream.py` 获取可运行的示例（该示例也包含了拼接逻辑，只是默认注释了最后的打印）。
 
-**注意：** 如果你只需要最终的完整结果而不需要流式处理，可以直接调用 `ai_agent.work(...)` 方法，它会内部处理好拼接并直接返回结果字符串。
\ No newline at end of file
+**注意：** 如果你只需要最终的完整结果而不需要流式处理，可以直接调用 `ai_agent.work(...)` 方法，它会内部处理好拼接并直接返回结果字符串。
+
+### Planned Refactoring: Decoupling Generation and Output Handling
+
+To enhance the flexibility and extensibility of this tool, we are planning a refactoring effort to separate the core content/image generation logic from the output handling (currently, saving results to the local filesystem).
+
+**Motivation:**
+
+*   **API Integration:** Allow generated results (topics, text content, image URLs/data) to be easily returned via an API endpoint instead of only being saved locally.
+*   **Alternative Storage:** Enable saving results to different backends like databases, cloud storage (e.g., S3, OSS), etc.
+*   **Modularity:** Improve code structure by separating concerns.
+
+**Approach:**
+
+1.  **Modify Core Functions:** Functions responsible for generating topics, content, and posters (primarily in `utils/tweet_generator.py` and potentially `core` modules) will be updated to **return** the generated data (e.g., Python dictionaries, lists, PIL Image objects, or image bytes) rather than directly writing files to the `./result` directory.
+2.  **Introduce Output Handlers:** An "Output Handler" pattern will be implemented.
+    *   An abstract base class or interface (`OutputHandler`) will define methods for processing different types of results (topics, content, configurations, images).
+    *   An initial concrete implementation (`FileSystemOutputHandler`) will replicate the current behavior of saving all results to the `./result/{run_id}/...` directory structure.
+3.  **Update Main Workflow:** The main script (`main.py`) will be modified to:
+    *   Instantiate a specific `OutputHandler` (initially the `FileSystemOutputHandler`).
+    *   Call the generation functions to get the data.
+    *   Pass the returned data to the `OutputHandler` instance for processing (e.g., saving).
+
+**Future Possibilities:**
+
+This refactoring will make it straightforward to add new output handlers in the future, such as:
+
+*   `ApiOutputHandler`: Formats results for API responses.
+*   `DatabaseOutputHandler`: Stores results in a database.
+*   `CloudStorageOutputHandler`: Uploads results (especially images) to cloud storage and potentially stores metadata elsewhere.
+
+**(Note:** This refactoring is in progress. The current documentation reflects the existing file-saving behavior, but expect changes in how results are handled internally.)
+
+### 配置文件说明 (Configuration)
+
+主配置文件为 `poster_gen_config.json` (可以复制 `example_config.json` 并修改)。主要包含以下部分：
\ No newline at end of file