Support image messages in the playground

[jeff-an]

Describe the feature or potential improvement

Many other LLM observability platforms have the ability to add messages that contain images in the playground.

The UX I've seen usually allows the user to either paste in an image URL, or drop a file from their own machine onto the message. Presumably the platform then converts the file to base 64 and puts the value into the message.

Almost all of our agents require images, so this would be a great way to rapidly iterate on prompts. Today, we would have to run a script to extract the base64 from an image we have on our machine and then paste that giant block into the frontend. The frontend today also doesn't seem to support messages with type image at all.

Additional information

No response

[marcklingen]

thanks for the feedback!