Record a workflow and edit resources(记录工作流并编辑资源(Record a workflow and edit resources))¶
This page provides instructions on the configuration and steps required to record a workflow.
Create a flow capture resource¶
- Open the Flow Capture application.
- Select + New Flow Capture in the top right.

- Choose a file name and location to save your resource in Compass.
- Select Save to create the new flow capture resource.
Select a system prompt template¶
After saving your resource, you will be redirected to the Record workflow page. Under What are you looking to document?, choose the template that best matches the documentation you want to generate.
The following templates are currently available:
- General documentation: Capture a workflow and translate it into Markdown documentation.
- Report a bug: Reproduce an issue you found in the platform to provide rich detail for product partners.
- Request a feature: Point out a shortcoming of the platform and suggest an improvement.
- Provide feedback: Describe product behavior, functionality, or usability feedback while you use the platform.
You can change the selected template later if you need to adjust the generation prompt for a different use case.
:::callout{theme="warning"} Before recording, ensure that you understand your organization's data handling policies. Flow Capture will record screenshots and optional audio; you must confirm that the content you record is appropriate for the intended use and is handled according to your security policies. :::
Prepare to record¶
- Select Record workflow to open the recorder acknowledgement.
- In the acknowledgement dialog, select the checkbox to confirm the required security settings, and then select Acknowledge.
Review recorder controls and options¶
- Screenshot counter: Displays the number of captured screenshots in the top left of the recorder overlay.
- Record audio: Enable this toggle to capture audio that will be transcribed and associated with the recording.
- Auto-screenshot: When this toggle is enabled, the recorder will automatically capture a screenshot on click while you navigate the platform.
- Size dropdown: Choose the capture size for screenshots. This dropdown provides the following options to scale the recording frame:
- Full size: Capture at the full resolution of the target page.
- Scaled: Scale the page to fit the recorder viewport.
- Custom: Specify custom dimensions for the capture frame.
- URL field: Use this to navigate directly to the page you want to capture with the recorder. The URL field only accepts same-domain URLs. The page you navigate to must be on the same Foundry enrollment that Flow Capture is running on.
Start recording¶
- Enable Record audio if you want spoken narration and transcription.
- Enable or disable Auto-screenshot depending on whether you want screenshots captured automatically on click.
- Select Record to begin capturing your session.
Navigate while recording¶
Navigate to the pages you want to document by entering a URL in the recorder URL field, or by browsing in the recorder iframe. Interact with the application as you normally would to demonstrate the workflow you want to document.
Capture screenshots manually¶
If Auto-screenshot is enabled, screenshots will be taken automatically for every click event, but you may take manual screenshots at any time. Use the Take screenshot option in the recorder controls to capture the current view, or use the keyboard shortcut Cmd+Shift+S (macOS) or Ctrl+Shift+S (Windows) to capture a screenshot.
Use recording controls during the session¶
- Select Pause to pause the capture.
- Select Record to resume after pausing.
- Select Discard and exit to cancel the session and remove any captured media.
- When you have finished capturing the workflow, select Done.
After recording¶
You can open the flow capture resource in edit mode to review and edit screenshots, transcriptions, and the generated context before generating the final documentation. If recorded, your audio will be transcribed automatically and associated with the captured screenshots.
Edit resources¶
Use the edit controls on a flow capture to refine screenshots and transcriptions, and define which artifacts are included in the LLM generation context. Edits are staged until you save them.
Add content to LLM context¶
You can add a snapshot or a transcription to the generation context to make it available to the LLM during generation. To do so, hover over the asset you want to add and select Add to context. You can also import images into Flow Capture by selecting Upload in the top right corner of the Assets page, and selecting the files you want to upload.
Crop images¶
Crop images to focus the reader's attention on relevant UI elements.
- Select the snapshot you want to edit.
- Select Crop image.
- Draw the crop region over the image by dragging the selection handles.
- Select Apply to accept the crop.
- When you have finished editing, select Save Edits, then select Done to return to edit mode.
Blur images¶
Blur sensitive information before you include screenshots in generated documentation or exports.
- Select the snapshot you want to edit.
- Select Blur Image.
- Draw a region to blur by dragging the selection handles.
- Select Apply to preview the blur.
- Select Save Edits, then select Done to return to edit mode.
Edit transcriptions¶
You can correct or improve transcriptions before generating documentation.
- Select the transcription text or the pencil icon to open the Edit transcription dialog.
- Edit the text in the Audio transcription field.
- To regenerate the transcription automatically, select Regenerate.
- Select Save and close the dialog to return to edit mode.
- Select Cancel to close the dialog without saving.
Delete or restore images and audio¶
You can remove images or audio files from a flow capture. Deleted items remain recoverable until you save your changes by selecting Save in the edit toolbar.
- Hover over the image or audio card to reveal the delete icon.
- Select the delete icon to mark the item as deleted.
- To restore a deleted item before saving your edits, select Restore snapshot.
- After selecting Save, deleted images and audio will be permanently removed.
中文翻译¶
记录工作流并编辑资源(Record a workflow and edit resources)¶
本文档提供记录工作流所需的配置说明和操作步骤。
创建流程捕获资源(Flow Capture Resource)¶
- 打开流程捕获(Flow Capture)应用程序。
- 点击右上角的 + 新建流程捕获(+ New Flow Capture)。

- 选择文件名和保存位置,将资源保存在 Compass 中。
- 点击 保存(Save) 创建新的流程捕获资源。
选择系统提示模板(System Prompt Template)¶
保存资源后,您将被重定向至 记录工作流(Record workflow) 页面。在 您希望记录什么内容?(What are you looking to document?) 下,选择与您要生成的文档最匹配的模板。
目前提供以下模板:
- 通用文档(General documentation): 记录工作流并将其转换为 Markdown 文档。
- 报告缺陷(Report a bug): 重现您在平台中发现的问题,为产品合作伙伴提供详细信息。
- 请求功能(Request a feature): 指出平台的不足之处并提出改进建议。
- 提供反馈(Provide feedback): 在使用平台时描述产品行为、功能或可用性反馈。
如果您需要针对不同用例调整生成提示,可以稍后更改所选模板。
:::callout{theme="warning"} 在开始记录前,请确保您了解所在组织的数据处理政策。流程捕获会记录屏幕截图和可选的音频;您必须确认所记录的内容适合预期用途,并按照安全策略进行处理。 :::
准备记录¶
- 选择 记录工作流(Record workflow) 打开录制器确认对话框。
- 在确认对话框中,勾选复选框以确认所需的安全设置,然后选择 确认(Acknowledge)。
查看录制器控件和选项¶
- 截图计数器(Screenshot counter): 在录制器覆盖层的左上角显示已捕获的截图数量。
- 录制音频(Record audio): 启用此开关可捕获音频,音频将被转录并与录制内容关联。
- 自动截图(Auto-screenshot): 启用此开关后,录制器将在您浏览平台时自动在每次点击时捕获截图。
- 尺寸下拉菜单(Size dropdown): 选择截图的捕获尺寸。此下拉菜单提供以下选项来缩放录制帧:
- 全尺寸(Full size):以目标页面的完整分辨率捕获。
- 缩放(Scaled):缩放页面以适配录制器视口。
- 自定义(Custom):为捕获帧指定自定义尺寸。
- URL 字段(URL field): 使用此字段直接导航到您要捕获的页面。URL 字段仅接受同域 URL。您导航到的页面必须与流程捕获运行在同一个 Foundry 注册环境中。
开始录制¶
- 如果您需要语音旁白和转录,请启用 录制音频(Record audio)。
- 根据您是否希望在点击时自动捕获截图,启用或禁用 自动截图(Auto-screenshot)。
- 选择 录制(Record) 开始捕获您的会话。
录制过程中导航¶
通过在录制器 URL 字段中输入 URL,或在录制器 iframe 中浏览,导航到您要记录的页面。像平常一样与应用程序交互,以演示您要记录的工作流。
手动捕获截图¶
如果启用了 自动截图(Auto-screenshot),系统会在每次点击事件时自动截图,但您也可以随时手动截图。使用录制器控件中的 截图(Take screenshot) 选项捕获当前视图,或使用键盘快捷键 Cmd+Shift+S(macOS)或 Ctrl+Shift+S(Windows)进行截图。
在会话期间使用录制控件¶
- 选择 暂停(Pause) 暂停捕获。
- 选择 录制(Record) 在暂停后继续录制。
- 选择 放弃并退出(Discard and exit) 取消会话并删除所有已捕获的媒体。
- 完成工作流捕获后,选择 完成(Done)。
录制完成后¶
您可以打开流程捕获资源进入编辑模式,在生成最终文档前查看和编辑截图、转录文本以及生成的上下文。如果录制了音频,系统会自动转录并与捕获的截图关联。
编辑资源¶
使用流程捕获上的编辑控件来优化截图和转录文本,并定义哪些内容包含在 LLM 生成上下文中。编辑内容会暂存,直到您保存为止。
向 LLM 上下文添加内容¶
您可以将快照或转录文本添加到生成上下文中,使其在生成过程中可供 LLM 使用。将鼠标悬停在要添加的资源上,然后选择 添加到上下文(Add to context)。 您还可以通过选择 资源(Assets) 页面右上角的 上传(Upload),然后选择要上传的文件,将图像导入流程捕获。
裁剪图像¶
裁剪图像以将读者的注意力集中在相关的 UI 元素上。
- 选择要编辑的快照。
- 选择 裁剪图像(Crop image)。
- 通过拖动选择手柄在图像上绘制裁剪区域。
- 选择 应用(Apply) 确认裁剪。
- 完成编辑后,选择 保存编辑(Save Edits),然后选择 完成(Done) 返回编辑模式。
模糊图像¶
在将截图包含到生成的文档或导出内容之前,对敏感信息进行模糊处理。
- 选择要编辑的快照。
- 选择 模糊图像(Blur Image)。
- 通过拖动选择手柄绘制要模糊的区域。
- 选择 应用(Apply) 预览模糊效果。
- 选择 保存编辑(Save Edits),然后选择 完成(Done) 返回编辑模式。
编辑转录文本¶
您可以在生成文档之前更正或改进转录文本。
- 选择转录文本或铅笔图标,打开 编辑转录(Edit transcription) 对话框。
- 在 音频转录(Audio transcription) 字段中编辑文本。
- 要自动重新生成转录文本,请选择 重新生成(Regenerate)。
- 选择 保存(Save) 并关闭对话框,返回编辑模式。
- 选择 取消(Cancel) 关闭对话框而不保存。
删除或恢复图像和音频¶
您可以从流程捕获中移除图像或音频文件。已删除的项目在您通过编辑工具栏中的 保存(Save) 保存更改之前仍可恢复。
- 将鼠标悬停在图像或音频卡片上,显示删除图标。
- 选择删除图标,将项目标记为已删除。
- 要在保存编辑之前恢复已删除的项目,请选择 恢复快照(Restore snapshot)。
- 选择 保存(Save) 后,已删除的图像和音频将被永久移除。