Streaming

By default, Atlas streams responses as they’re generated — you see words appear in real time rather than waiting for the entire response to finish. This makes conversations feel faster and more natural.

How Streaming Works

When you send a message, the Atlas cloud backend connects to the AI model and forwards each chunk of text to your app as soon as it’s produced. The response builds up word by word in the chat panel.

Under the hood, this uses a standard web technology called Server-Sent Events (SSE) — the cloud backend keeps a connection open and pushes each piece of text through as it arrives. You don’t need to know the details; it just means the streaming is reliable and works through firewalls and proxies the same as any normal web request.

Tool Calls During Streaming

If Atlas decides to use a tool partway through generating a response, you’ll see the tool call appear in the chat while streaming is happening. For example:

Atlas starts typing a response
It realizes it needs to look something up in your vault
A tool-call indicator appears
The tool runs
Atlas continues (or restarts) its response with the new information

This all happens in sequence and you can watch it unfold in real time.

Turning Off Streaming

If you prefer to see the complete response appear all at once — or if you’re on a connection where streaming causes issues — you can turn it off.

Go to Settings > AI > Enable Streaming and toggle it off.

With streaming disabled:

The chat input is locked while Atlas thinks
The full response appears at once when it’s ready
Tool calls still happen, but you won’t see them mid-response — the final result appears when everything is done