Your first chat¶

In this tutorial you ask Nawa a question it has to look up, watch it work, and read the answer with its sources. It takes about five minutes and assumes a connected provider (Connect a provider).

On a new installation, the guided tour opens Settings at LLM to help you connect a model, then highlights the workspace controls where you use them. You can skip it at any point, or replay it from Settings → General → Getting started → Take a tour.

1. Turn on the web¶

Nawa researches on the web only when you allow it. In Settings → Web, Automatic uses a configured search API, otherwise Web search (DDGS) without a key or an open browser. You can also choose a provider and enter its key. Close Settings.

Click New chat at the top of the sidebar. In its composer, the web toggle beside the input is now available; turn it on. It stays on for this chat.

The composer’s reasoning menu shows this chat’s Main and Coder models and reasoning choices. Existing chats retain saved selections when Settings defaults change. See LLM settings.

2. Ask¶

Type into the composer:

What changed in the latest release of Node.js, and which version is it?

Press Send (or Enter; Shift+Enter makes a new line).

At the end of the chat, your message moves to the top and the answer appears below it; once the answer fills the view, the transcript follows it. If you have scrolled up to read, sending leaves the view where it is, and Jump to latest above the composer takes you to the new turn. Scrolling while Nawa works always keeps you where you scroll.

3. Watch the turn¶

Nawa narrates its searches and page reads above the answer. Each step has a receipt you can open. Sending another message while it works puts that message in the prompt queue to run after the current turn completes. To change the current task, use that queued prompt's Steer button instead. Queue and steer prompts explains both paths. With an empty composer, the send button becomes Stop generating. Stop cancels running Windows commands through their process owner and checks descendant cleanup; other chats stay responsive. During model-server connection retries, a notice shows the attempt count and clears on connection or Stop.

If a step needs your permission, a card appears in the transcript with Allow once, Always allow and Deny, and a clock showing how long it has waited. Permissions explains when that happens; for a web question in the default mode it usually does not.

4. Read the answer¶

The answer arrives as it is written. A numbered web citation opens the source page. Up to four source cards appear after a sourced answer; the circular chevron opens the drawer when there are more. The drawer contains sources attached to that answer, not a complete log of every page visited. When you ask for a search-result link, Nawa can give you the address without opening the page. Numbered citations refer to sources it actually read. Older chats may contain saved search listings, labeled as pages not opened, with links to the listed websites.

5. Keep going¶

Ask a follow-up in the same chat; Nawa remembers the conversation, and compacts it quietly when it grows long.
Attach a file or an image with the paper-clip button beside the composer; Nawa reads documents and looks at images.
The chat's menu (the three dots in its header) offers Rename, Files — the chat's workspace — Automate and Delete.

What you have seen¶

A turn: your message, the steps Nawa took with its tools, and the answer, all recorded so the receipts stay readable later. The next tutorial, Your first app, uses the same turn to build something.

Verbosity

Settings → General → Transcript verbosity chooses how much of the work is narrated: Normal, Verbose, or Summary.