Chatting
Sending messages
Type in the composer and press Enter to send (Shift+Enter for a new line). While a reply is streaming you'll see it render live, with a thinking/elapsed-time indicator — press the stop button (or Esc) to cut a response short. You only pay for what was generated.
Attaching files
Paste or upload files into the composer, up to 20 MB each:
| Type | Formats & handling |
|---|---|
| Images | PNG, JPEG, WebP, or GIF. Models with the Vision capability can read them. |
| Documents | PDF, Word (.docx), Excel (.xlsx), Markdown, plain text, CSV, or HTML. The text is extracted in your browser and sent to the model; the raw file never leaves your device. |
Choosing a model
The model picker in the composer lets you browse every model on the network, filter by text/image capability, search by name, and star favorites. Auto mode picks the best model for each message for you. See Models & operators.
Message actions
Hover a message (or tap it on mobile) for actions:
| Action | What it does |
|---|---|
| Copy | Copy the message text. |
| Retry / Regenerate | Ask the model to answer again. |
| Edit | Rewrite one of your own messages and resend it; any attachments (images and documents) are preserved. |
Who answered
Each answer is labeled with the model that produced it — historical attribution, so it stays correct after you switch models. The ⓘ beside that label opens the details of that turn: the model's description and context window, plus the operator that served it — its on-chain operator and node ids, and its NFD name and avatar where it has one (see How operators are named). If that operator has since left the network the ids remain, but there's no name left to resolve.
Sources and citations
When a model cites sources, citations appear as superscript markers in the text with the full source list at the end of the message.
Reasoning
Models that support reasoning show their thinking in a collapsible block above the answer — expand it to see how the model got there.
Tool activity
When a model uses tools mid-reply, each call shows up as a compact labeled
line in the response saying what it's doing — Searching for "algorand consensus", Reading algorand.co — with the sites a search turned up
listed beneath it as clickable chips, as they land. A run of consecutive
tool calls collapses into a one-line summary — "2 web searches · 1 page
read" — that expands to the full list with each call's own detail, so
tool-heavy replies stay readable. Operators on an older node version don't
send the detail; there the lines read generically ("Searched the web"),
which is normal. These tools run on the operator's node, not in your
browser — see Built-in tools.
Generating and editing images
Models with the Generation or Edit capability can create and modify images mid-conversation. Click any image to open it full-screen, where you can download it, copy it, or hand it back to the model for edits. The Images setting controls how many images a single request may produce — see Settings and Pricing.
Keyboard shortcuts
The shortcuts you'll use most in the composer:
| Shortcut | Action |
|---|---|
Enter | Send the message |
Shift+Enter | New line |
Esc | Stop a streaming reply |
See Keyboard shortcuts for the full list, including search, settings, and conversation actions.