> ## Documentation Index > Fetch the complete documentation index at: https://langwatch.ai/docs/llms.txt > Use this file to discover all available pages before exploring further. # View images in datasets > View image datasets in LangWatch to support multimodal evaluations and agent testing scenarios. With your images column type set to `image (URL)`, the dataset shows the images themselves, so you can analyze them at a glance. The screenshot below shows how to set the column type to `image (URL)`. You can also set this type when creating a new dataset.
LangWatch
Once you select the image type, the dataset shows the image column. You can edit the column value by clicking on the image cell. Keep in mind the column value is the URL of the image. LangWatch ## How attachment URLs reach the model When an image URL feeds a vision workflow or evaluation, LangWatch fetches the URL server-side and delivers the actual image to the model as image content, so the model sees the picture instead of the link text. This works across every model provider, whether or not the provider fetches URLs itself. The attachment type is detected from the response rather than the file extension, so extension-less links such as signed S3 or CDN URLs are classified correctly. Audio and PDF URLs are delivered the same way when referenced from a prompt: an audio URL reaches the model as audio for it to hear, and a PDF URL reaches it as a document for it to read. If a field is declared as an image and its URL cannot be loaded as an image (the URL is unreachable, returns an error status, is too large, or serves something other than an image), the run fails with a clear message naming the URL and the reason, instead of passing the raw link to the model as text. A plain link mentioned in prose is left as text and never fails the run.