Websites and images

Websites#

Add websites to the Dataroom by clicking on the Add Website button.

To keep AITEZA fast for everyone, it crawls only a limited number of websites at the same time. When many crawls are running, the website shows the status Queued and starts by itself as soon as a place is free. People with fewer running crawls go first, so many websites of one person do not hold up everyone else’s. Your websites keep their place in the queue, even if AITEZA restarts in between.

Crawled URLs#

For websites with subpage crawling enabled, you can view all crawled subpage URLs along with metadata such as max depth, max pages, and the last scanned date.

Website Version History#

Web sources have version history showing snapshots of crawled content over time. This allows you to track how website content has changed between crawls.

Automatic Website Updates#

AITEZA can crawl a website again on its own, so that changes to its pages, such as a revised standard, reach the Dataroom without anyone rescanning it by hand. Each automatic update is saved as a new version in the website’s version history.

To stop the automatic updates, open the same dialog and choose Off. If the website is still being crawled when an update is due, AITEZA tries again an hour later. When many crawls are already running, the update waits until one finishes. You need editor rights in the Dataroom to change the schedule.

Retry Embeddings for Websites#

If a website was crawled successfully but embedding creation failed, you can retry the embedding step without re-crawling.

Images#

The Images tab lets you upload image files that are made available to multimodal models for true visual analysis in agent mode. This is fundamentally different from uploading images via the Files tab, where only text is extracted via OCR.

When you upload an image here, AITEZA automatically generates a short and a long description of the image. These descriptions are vectorized and used for text-based semantic search, so the agent can find and retrieve the right image based on your query. Once located, the agent can inspect the actual image file using its Get Image tool, not just the text representation.

Images are turned upright automatically before they are described: photos by their orientation information, drawings and scans saved sideways or upside down by the direction in which their text reads best. A preview of each image appears in the list right away, while it is still uploading.

Images from Documents#

AITEZA also extracts images from documents that are uploaded via the Files tab: diagrams, charts, screenshots, and other images contained in a document are recognized automatically and become available in the Images tab, just like individually uploaded image files.

You can also ask AITEZA to include images in its answers. If you do, the relevant image, for example a diagram from one of your documents, is added to the text in the response.

Supported Image Formats#

All common image formats are supported for upload in the Images tab, including JPEG, PNG, GIF, BMP, TIFF, WebP, and others. You can also upload PDF files directly in the Images tab: each page of the PDF is automatically split and processed as an individual image, so PDF content becomes usable for visual analysis wherever images are supported.

Downloading Images#

You can download images directly from the Images tab.

Image Version History#

Images support full version history, similar to files. You can view previous versions and upload new versions while preserving history.

Retry Embeddings for Images#

If an image was processed successfully but embedding creation failed, you can retry the embedding step.