Skip to main content

The Media Hub

Generative AI is no longer just about text. You might be testing an agent that analyzes receipts, a system that OCRs PDF documents, or a bot that answers questions about audio files.

How do you supply a dynamic "image" to your LLM-as-Judge metric inside a CSV dataset?

Enter the Media Hub and Asset Paths.

Asset Paths​

CertOps allows you to toggle the schema type of any column in your dataset from standard Text to an Asset Path.

When you define a column as an Asset Path, you are telling CertOps: "The text string in this column is actually a pointer to a file."

An Asset Path can easily be one of two things:

  1. A Public URL: e.g., https://example.com/images/receipt_01.jpg
  2. A Media Hub Relative URL: e.g., /eval_images/receipt_01.jpg

When evaluating a multi-modal metric, CertOps natively understands these paths, streams the underlying file representation, and seamlessly passes the actual image/document into your chosen LLM-as-Judge as multimodal content.

The Media Hub​

Relying on public URLs for proprietary evaluation data is often a security nightmare. We created the Media Hub so you don't have to host your evaluation assets yourself.

The Media Hub is you and your team's private, unified file storage system right inside the CertOps platform.

How to use it:

  1. Navigate to the Media Hub in the CertOps dashboard.
  2. Create folders to organize your assets (e.g., /eval_images).
  3. Upload your multi-modal testing files (.jpg, .pdf, .mp3) directly into the hub.
  4. In your CSV dataset, simply reference the file by its relative path (e.g., /eval_images/receipt_01.jpg) and toggle the column type to Asset Path.

CertOps securely links your Media Hub files directly into your evaluation runs, keeping your data safe while unlocking the full testing power of modern multi-modal AI!