Features/Images in chat

Feature

Images in chat

Chat Images is Diorama's planned attachment flow for sending one image with an AI chat message. It is designed for quick screenshot-style context when text alone is not enough.

Sofia was trying to make sense of a confusing chart in another app, and describing it in words took three tries and still left gaps. With image attachments she sent one screenshot and got a straight answer in a single reply. A back-and-forth that used to stall became a thirty-second clarification.

Why it matters

Many support and coaching questions are faster to answer when people can show what they are seeing. A safe image attachment flow can reduce back-and-forth and make AI responses more relevant on the first reply.

How Diorama uses this

This flow is still being completed and is not fully available in the active composer UI yet. Once ready, you'll select one photo or screenshot from your library or camera before sending a message. You'll see a preview chip with options to remove or replace the image before hitting send.

What you can do

Chat Images is currently in an in-between state. Core preprocessing and transport plumbing exist, but the full user-facing attach flow has not fully shipped in the chat composer yet.

The intended experience is simple: attach one image, preview it before sending, remove or replace it if needed, and send with your prompt for multimodal analysis. The emphasis is on transient usage for better context, especially for screenshots.

Until that flow is completed, users should continue to treat chat as text-first.

Notes and limitations

End-to-end attachment handling is still incomplete, including composer affordances, persistence of attachment metadata in all write paths, and explicit attachment state transitions.

The roadmap also requires user-facing privacy disclosure and reliable error handling before this should be considered generally available.

The science

Showing is lower-bandwidth than describing. When you can point at the exact chart or screen you mean, the assistant spends its effort solving the question instead of guessing what you are looking at — and guessing is where these exchanges usually go sideways.

Limitations

Treat image-enabled chat as unavailable until the composer, persistence metadata, and send-state handling are completed end-to-end. Image attachments are intended to be transient for request-time analysis rather than long-term chat media history. Current implementation work is still missing full attachment UI, analytics instrumentation, and reliable failure-state UX.

Good tools meet questions where they actually occur, which is usually mid-task with something on screen. Letting you show before you explain matches how people already think when they are stuck.

Diorama

See your whole picture.

Diorama brings training, recovery, and health together so you keep competing — not just keep going.