Bring files into the conversation
Attachments for richer requests
Some requests are easier to show than to type. Attachments let users add images, PDFs, and text files right in the chat—a screenshot of an error, a PDF invoice, a photo of their room—so the assistant can see and reason over exactly what they mean. There's nothing to integrate: the widget handles the attach button, drag-and-drop, and uploads.
What you get
- Users add files with a paperclip button or drag-and-drop—no host-page integration
- Supports images (PNG, JPEG, WebP, GIF), PDFs, and text/Markdown, up to 25 MB each
- Images and PDFs preview in a side panel; every file is passed to the assistant with the message
- Per-application toggle, enforced server-side—uploads are rejected even if the UI is bypassed
- Uploads stream to a private CDN via short-lived signed upload URLs, and files auto-delete after 7 days
- The foundation for image generation—upload a photo, get your product composited into the scene
How it works
Enable it for your app
Attachments are on by default; toggle them per application in the dashboard's Behavior settings, where the switch is a server-side gate the host page can't override.
Users attach files in chat
From the paperclip button or by dragging files onto the panel, each upload streams to a private CDN through a short-lived signed URL with a progress thumbnail.
The assistant reasons over them
On send, ready files attach to the message as file parts and reach the model with the text, so it can answer about the screenshot, invoice, or photo the user shared.