Skip to main content
Adaline datasets are structured as tables where columns represent the variables in your prompt, rows represent individual test cases, and cells hold the value for each variable in a given test case. Since Adaline supports multimodal prompts, each cell in a dataset can hold its own modality — text, image, or PDF — independently of any other cell in the same row or column. This means a single row can mix a text question in one column, an image in another, and a PDF in a third — matching exactly how your multimodal prompt consumes different input types. Multimodal cells in a dataset

​
Supported modalities

ModalityInput methodsSupported formats
TextType or paste directly into the cellPlain text
ImagePaste a publicly accessible URL or a base64-encoded stringPNG, JPEG/JPG, GIF, WebP
PDFPaste a publicly accessible URL or a base64-encoded stringPDF (up to 10 MB)
Each cell is independent — you can have a text cell next to an image cell next to a PDF cell in the same row. You can also have different modalities across rows within the same column.

​
Text

Text is the default modality. Simply click on a cell and type or paste your content. There is no additional configuration needed. For example, if your prompt has a {{user_question}} variable, you would add a user_question column and type the question directly into each cell:
user_question
What is the capital of France?
Summarize this article in three sentences.
Translate the following to Spanish: “Hello, how are you?”

​
Images

To add an image to a cell, paste one of the following:
  • A publicly accessible image URL — e.g., https://example.com/photos/front.png
  • A base64-encoded image string — e.g., data:image/png;base64,iVBORw0KGgo...
Adaline automatically detects the image and renders a preview in the cell. Supported image formats: PNG, JPEG/JPG, GIF, and WebP. Image cells rendered in a dataset

​
Example: image via URL

Paste a direct link to a publicly hosted image into the cell: 
https://upload.wikimedia.org/wikipedia/commons/thumb/0/02/Knobby_ADALINE.jpg/960px-Knobby_ADALINE.jpg
Adaline fetches the image and displays a thumbnail.

​
Example: image via base64

Paste a base64-encoded image string with the data URI prefix: 
data:image/jpeg;base64,/9j/4AAQSkZJRgABAQAAAQABAAD...
Base64 images in datasets currently have a size limit of 1 MB. We’re increasing this limit to 10 MB soon!

​
PDFs

To add a PDF to a cell, paste one of the following:
  • A publicly accessible PDF URL — e.g., https://example.com/docs/report.pdf
  • A base64-encoded PDF string — e.g., data:application/pdf;base64,JVBERjQK...

​
Example: PDF via URL

Paste a direct link to a publicly hosted PDF into the cell: 
https://example.com/docs/quarterly-report.pdf

​
Example: PDF via base64

Paste a base64-encoded PDF string with the data URI prefix: Maximum PDF size: 10 MB. 
data:application/pdf;base64,JVBERi0xLjQKMSAwIG9iago8PA...
Base64 PDFs in datasets are hosted on Adaline storage, users can access them via Dataset API as well.

​
Importing multimodal data via CSV

You can bulk-import multimodal datasets using CSV files. In the CSV, text values are entered as-is, and image or PDF values are entered as their publicly accessible URLs or base64 strings. Adaline automatically detects the modality of each cell based on the value.

​
Text-only CSV

A standard CSV where every cell contains plain text: 
user_question,context
What is the capital of France?,"France is a country in Western Europe."
Summarize this article.,"The article discusses the impact of AI on healthcare."

​
Mixed modalities CSV (text + images)

A CSV that combines text and image URLs in different columns: 
user_question,product_image
Describe this product.,https://example.com/images/product-a.png
What color is this item?,https://example.com/images/product-b.jpeg
Is this item suitable for outdoor use?,https://example.com/images/product-c.webp

​
Mixed modalities CSV (text + images + PDFs)

A CSV that combines text, image URLs, and PDF URLs across columns: 
user_question,reference_image,supporting_document
Describe what you see in the image.,https://example.com/images/storefront.png,https://example.com/docs/brand-guidelines.pdf
Compare this image to the brand guidelines.,https://example.com/images/new-design.jpeg,https://example.com/docs/brand-guidelines.pdf
Summarize the key findings from this report.,https://example.com/images/chart.png,https://example.com/docs/q4-report.pdf

​
Mixed modalities CSV with base64

You can also include base64-encoded content directly in the CSV. This is useful when your images or PDFs are not publicly hosted: 
user_question,product_image
Describe this product.,data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAA...
What color is this item?,https://example.com/images/product-b.jpeg
When mixing URLs and base64 strings in the same column, Adaline handles each cell independently. You can freely mix both formats within a single column.

​
Importing the CSV

Once your CSV is ready, click Upload CSV in the dataset editor to import it. Adaline parses each cell value and automatically detects the modality:
  • Plain text values are treated as text.
  • Values starting with http:// or https:// pointing to image or PDF files are treated as the corresponding modality.
  • Values starting with data:image/ or data:application/pdf;base64, are treated as base64-encoded images or PDFs.
See Import CSV into Dataset for the full import guide.

​
Size limits and storage

Content typeMaximum size
TextUp to 1 MB
Image URLNo limit (fetched at runtime)
Image base64Up to 1 MB (larger sizes coming soon)
PDF URLNo limit (fetched at runtime)
PDF base64Up to 10 MB

​
Next steps

Setup Dataset

Create datasets and configure columns for evaluation.

Import CSV into Dataset

Bulk-import test cases including multimodal data.

Evaluate Prompts

Run evaluations on your multimodal dataset.

Use Images in Prompt

Learn how to use images in your prompts.