Learning Timeline
Key Insights
Debugging with Console View
Instead of relying solely on the final output, click the 'Console' tab during execution. This allows you to inspect the raw input and output of every single node individually to verify data is passing through correctly.
Multi-Modal Model Selection
The 'Generate' node isn't limited to text. You can swap the model to 'Imagen' for image generation, 'Audio LM' for text-to-speech, 'Veo' for video creation, or 'Lyria' for music generation depending on your workflow needs.
Prompts
Archive Data Scraper Prompt
Target:
Google Opal (Gemini 2.5 Flash)
Go to this archive URL and then scrape the first 10 results. Parse each result into a table with the columns title, authors, and archive link.
Article Link Extractor Prompt
Target:
Google Opal (Gemini 2.5 Flash)
Go to this URL and scrape the links to the first five articles.
Markdown Summarizer Prompt
Target:
Google Opal (Gemini 2.5 Flash)
Go to each link and generate a one paragraph summary of the article in the following markdown format.
[Insert Sample Markdown Paragraph Here: Make the first sentence bold using **text**, include a few sentences of summary, and end with a [Read More](URL) link].
Output all the summaries separated by line breaks.
Step by Step
Creating a Web Scraper to Spreadsheet Workflow
- Click the 'Create new' button to initialize a blank workspace.
- Locate the node selection menu and click 'Generate' to create an AI processing node.
- Click the model dropdown within the Generate node and select 'Gemini 2.5 Flash'.
- Click into the 'Prompt' text field within the node.
- Enter the specific scraping instructions (e.g., 'Go to this archive URL... parse each result into a table').
- Hover over the right edge of the Generate node until a connection point appears.
- Click and drag a connection line ('noodle') out from the Generate node.
- Select the 'Save' node from the resulting menu to create a spreadsheet destination.
- Click the title field (default 'Untitled Opal app') at the top of the screen and rename it to 'Archive Scraper'.
- Click the 'Start' button to execute the workflow.
- Wait for the process to complete, then click the Google Sheets link provided in the final node output to view the data.
Building a Dynamic AI Newsletter Workflow
- Click 'Create new' to start a fresh workflow.
- Select the 'User Input' node from the menu to allow dynamic URL entry.
- Type 'Enter URL to scrape articles' into the input label field of the User Input node.
- Drag a connection line from the User Input node and select a 'Generate' block.
- In the Generate node, ensure the input is set to receive data from the User Input node.
- Enter the prompt to extract links: 'Go to this URL and scrape the links to the first five articles.'
- Drag a connection line from the first Generate node and create a second 'Generate' node.
- Select 'Gemini 2.5 Flash' as the model for the second Generate node.
- Enter the summarization prompt instructing the AI to visit each link and format the output in Markdown.
- Drag a connection line from the second Generate node and select the 'Output' node to finalize the data flow.
- Click the 'Preview' tab at the top of the interface.
- Click 'Start'.
- Paste the target website URL (e.g., TechCrunch) into the prompt field when requested.
- Click 'Run' to generate the newsletter content.
