Card

Use me when your pipeline produces text, not numbers. While most plots turn data into charts, I turn structured JSON output from your pipeline nodes into a clean, readable card. If your pipeline runs an LLM, a summarization node, or an extraction node and you want to display the result — product names, AI summaries, sentiment scores, recommendations — I'll lay it out in a titled card with a header, body, and footer. No axes, no scales, just the information, formatted.

Overview

The Card plot displays structured JSON data produced by your pipeline as a formatted card layout. Unlike chart-based plots that visualize numerical or categorical data, the Card is designed specifically for text-rich, structured outputs — the kind of data that AI and data transformation nodes produce. Each section of the card (title, description, content, footer) maps to a key in the JSON output of the connected pipeline node.

Best used for:

• Showing LLM-generated summaries, analyses, or recommendations
• Displaying extracted entities or structured fields from unstructured text
• Presenting model predictions alongside their explanations
• Surfacing key facts from document processing or scraping pipelines
• Building lightweight report cards from automated pipeline runs
• Reviewing output from classification, scoring, or tagging nodes

Common Use Cases

AI-Generated Product Summaries

Connect a Card to an LLM node that summarizes product reviews or descriptions. Map product_name to Title, summary to Content, and sentiment to Footer. The card renders each pipeline run's output as a readable summary card on your dashboard.

Document Extraction Results

After running an extraction node on contracts, resumes, or forms, display the extracted fields as a card. Map the document title or filename to Title, the main extracted text to Content, and a confidence or status field to Footer.

Classification and Scoring Outputs

Show the result of a classification pipeline: the input record's identifier as Title, the model's label or class as Description, the reasoning or probability breakdown as Content, and a recommendation as Footer.

Structured Report Outputs

After a multi-step pipeline that analyzes data and produces a JSON summary, use the Card to surface the key findings without needing to build a separate reporting view.

LLM Chain Outputs

When chaining multiple LLM calls (e.g., extract then summarize then recommend), display the final structured output as a card to present the result end-to-end in one visualization.

Options

All Card fields are optional text inputs where you enter the JSON key name from your pipeline node's output. The card is populated by looking up that key in each row of the pipeline's output data.

Card Layout

Title Field

Optional — JSON key to use as the card header title.

Enter the name of the key whose value should appear as the large, prominent title at the top of the card. This is typically a short identifier — a product name, document title, record ID, or category label.

Example: If your pipeline outputs {"product_name": "Wireless Headphones X3", ...}, enter product_name.

Description Field

Optional — JSON key to use as the card header description or subtitle.

Enter the name of the key whose value should appear as a smaller subtitle beneath the title. Use this for a brief secondary label — a category, date, source, or short tagline.

Example: If your pipeline outputs {"category": "Consumer Electronics", ...}, enter category.

Content Field

Optional — JSON key to use as the main card body content.

Enter the name of the key whose value should appear as the primary body text of the card. This is the main readable output — a summary paragraph, extracted passage, model explanation, or recommendation.

Example: If your pipeline outputs {"summary": "Premium over-ear headphones with active noise cancellation...", ...}, enter summary.

Footer Field

Optional — JSON key to use as the card footer text.

Enter the name of the key whose value should appear at the bottom of the card, typically in a smaller, secondary style. Use this for metadata — a score, sentiment label, timestamp, confidence value, or status indicator.

Example: If your pipeline outputs {"sentiment": "Positive (0.87)", ...}, enter sentiment.

Tips

1. The Card reads from pipeline node output, not raw dataset columns: This plot is designed for the output of AI or transformation nodes that produce structured JSON, not for columnar tabular data. Make sure your pipeline node is connected and has run successfully before configuring the Card.

2. You only need to fill in the fields you want to show: All four layout fields are optional. If your output only has a title and body, fill in Title Field and Content Field and leave the rest blank. The card will render with only the populated sections.

3. Key names are case-sensitive: Enter the JSON key exactly as it appears in your pipeline output. If the output has "Product_Name" with capital letters, entering product_name (lowercase) will not match — double-check casing.

4. Use the Content field for long text: The Content field is the body of the card and has the most vertical space. Put your LLM-generated summaries, extracted paragraphs, or explanations here. Keep the Title and Description fields short (ideally under 100 characters).

5. The Footer is ideal for scores and status: Numerical scores, sentiment labels, confidence percentages, and processing timestamps all work well as footer values. They provide context for the content without cluttering the main body.

6. Combine with other plots for full pipeline visibility: Use the Card alongside a Table plot (to see all rows of output) or a Number plot (to display aggregate scores like average sentiment) to build a complete view of your pipeline's results.

7. If the card appears empty: Verify that your pipeline has run and produced output data. Then check that the key names you entered in the field inputs exactly match the keys in the JSON output — even a minor typo will result in a blank field.