Skip to main content

How to Add Images to a Datasource

Learn how to add image files (PNG, JPEG, GIF, WebP) to a datasource so your agent can view and reference them as pictures.

Your agent can answer questions from documents and spreadsheets, but many teams also keep critical information in pictures: product photos, screenshots, diagrams, scanned receipts, or whiteboard captures. Previously, datasources rejected image files. Now you can add images (PNG, JPEG, GIF, WebP) to a datasource so your agent can see them as pictures when answering questions.

You might need images in a datasource when:

  • Your product catalog lives partly in photos your agent should describe or compare

  • Your team stores architecture diagrams or screenshots the agent should reference

  • You want receipts or scanned forms visible to the agent alongside your documents

This guide shows you how to add image files to a datasource, which lets your agent view them as pictures in its answers.

Before you begin

  • Your company must have images-for-datasources enabled (ask your admin if you do not see image formats listed)

  • Supported formats are PNG, JPEG, GIF, and WebP; images over 5 MB are compressed in your browser before upload (animated GIFs are never re-encoded)

Steps

  1. Open your datasource and click Add files to see the supported file formats, which now list image types when the feature is on

  2. Drag images into the upload drop zone or pick them through the file picker, which accepts image files alongside documents

  3. Wait for any "Compressing..." indicator to finish, since the datasource cannot be saved while an image is still compressing

  4. Save the datasource and confirm each image row shows an image icon with "Processed" status

Common questions

Are images searched like documents?

No. Images are stored and shown to the AI as pictures; they are not searched. Text documents use search, images use vision.

What happens to files picked from Google Drive or other connectors?

Those are downloaded server-side and are not browser-compressed, so very large images may be rejected instead of shrunk.

What's next?

After adding images, ask your agent questions that reference them. To learn more about datasource types and configuration, see Datasources in elvex.

Did this answer your question?