AI chat
A streaming chat workspace for drafting, explaining and reviewing, with the conversations kept in your account so you can return to them.
Draft content — not yet reviewed by an editor.
How to use it
The chat workspace keeps the model picker, the message thread and the input box on one screen.
- Sign in, then pick a chat model from those the catalogue lists as available.
- Send a message. Attach an image if the model supports vision.
- Read the streamed reply, then edit your message and ask again if the direction is wrong.
- Reopen any earlier conversation from the history list.
What changes between models
Chat models differ in context length, whether they accept images, whether they expose a reasoning mode, and what they cost per token. The workspace shows the options each model declares instead of assuming them.
- Thinking or reasoning controls: only on models that declare them.
- Image input: only on models whose capabilities include vision.
- File attachments: handled through the service upload endpoint when the model accepts them.
Cost
Chat is charged by tokens rather than per message, so a long history costs more than a short question. The server quotes the request before it is sent and the ledger records the final amount.
Limits worth knowing
Replies are generated text and can be wrong; check anything you plan to rely on. Very long conversations may need to be trimmed before a model with a smaller context window will accept them.
Questions
Is history shared between models?
No. A conversation belongs to the model it was started with, so switching models starts a separate thread.
Can I edit an earlier message?
Yes. Editing a message discards the replies that followed it so the thread stays consistent, and the exchange is charged again when you resend.
Can I send a screenshot?
Only to a model whose capabilities include image input. The attach control is hidden when the selected model has no vision support.