Skip to content

Chat

The Chat page (/chat) is the main interface for interacting with AI models.

Chat page

When you start a new conversation, a notice explains that the built-in chat is intended for quick, lightweight interactions rather than a full-featured chat client, and links to the Desktop chat clients section of the Connect guide for using a dedicated client with the API. The notice disappears once you send a message.

Page Layout

  • Left sidebar — Lists your conversations. Click any conversation to reload its message history.
  • Model selector — At the top of the chat area. Choose which model to send your message to. Models that require acknowledgment show a warning (⚠) if you haven't acknowledged them yet. Models are omitted when your account or project has no coin pool, because acknowledging them would still leave them unusable.
  • Chat area — Displays message bubbles. Your messages appear on the right; assistant replies appear on the left.
  • Message detail — Each assistant message has an info icon (ⓘ). Click it to see the token counts, response duration, and output speed for that reply.

Conversation Management

  • New conversation — Click the + New button in the sidebar.
  • Switch conversations — Click any conversation in the sidebar to load its history.
  • Remove conversation — Hover over a conversation and click the ✕ button.
  • Conversation title — Automatically generated from your first message.
  • Conversation storage — Conversations are saved automatically. If you have disabled conversation storage on your profile page, chats still work but are not saved, and the sidebar shows a notice instead of a history.

Sending Messages

  1. Select a model from the dropdown at the top.
  2. Type your message in the input area at the bottom.
  3. Press Enter to send. Hold Shift+Enter to insert a newline without sending.

Streaming and Thinking

Responses stream in character by character as the model generates them. For models that support reasoning, a collapsible Thinking… section appears above the final answer, showing the model's chain-of-thought.

File Attachments

Attach a file to your message in two ways:

  • File picker — Click the paperclip (📎) button.
  • Drag and drop — Drag a file directly onto the message input bar.

Supported file types:

Type Extensions
Documents .txt, .md, .csv, .json, .py, .js, .ts, .html, .css, .xml, .yaml, .yml
PDFs .pdf
Images .png, .jpg, .jpeg, .gif, .webp
  • Images are sent as part of a "see and answer" message, for models that can interpret images.
  • Documents are read and prepended to your message text.
  • PDFs are parsed and appended as text.

An attachment chip appears above the input while composing. Click ✕ on the chip to remove it before sending.

Markdown and Math

Assistant responses are rendered with full formatting support:

  • Markdown — Headers, code blocks, lists, links, tables, and more.
  • Math — Inline $…$ and display $$…$$ LaTeX expressions rendered with KaTeX.
  • Code blocks — Dollar signs inside fenced code blocks are preserved and not treated as math.

On desktop, click the ←/→ arrow button in the header to collapse or expand the conversation sidebar. On mobile, the sidebar opens as an overlay via the hamburger menu.