Literal AI offers multimodal logging. In this example, you see how an image is logged by Literal AI in an application using OpenAI’s GPT4-Turbo model. A chat Thread is initiated, in which the assistent start with a welcome message. Then, a user query comes in, with a photo as attachment. In this example, this is hardcoded, but of course you can implement similar behavior in your chatbot. The attachment is added to the user message Step. This picture is then sent to the multimodal AI model, and the response is logged in Literal AI as well.

Here’s an example of how different kinds of attachments are displayed in Literal :

Attachments on a step

Examples of attachments