HomeChatSign in
Delaverse
Delaverse
ChatChangelogContact us
Chat
Start here
Getting startedYour first chat in a minute
HomeChat
Quick start

Getting started with the chat

The AIWidget chat is a web chat that runs right inside the dashboard. Talk to OpenAI, Anthropic and Google models, keep several conversations going at once, and have your history stored on the server — all paid from one shared wallet.

Nothing to installHistory kept on the serverParallel chatsPay for what you use

At a glance

Live chat in the browser

Live conversation with OpenAI, Anthropic and Google models, streamed token by token (SSE) — exactly like the VS Code extension.

History on the server

Every chat is saved server-side, so your past conversations follow you to any browser or device. Up to 30 are kept.

A model per chat

Each chat remembers its own model and reasoning level. Keep a quick chat on Haiku and a deep one on Opus at the same time.

Paid from your wallet

Token costs come out of the same AIWidget wallet as WijCode. A request you stop costs nothing.

Getting started

1

Sign in at delaverse.ai. If you don't have an account yet, sign up now — new accounts get a free daily message allowance to try things out.

2

From the side menu go to My widgets → Chat and open the Chat tab.

3

Pick a model from the menu at the top (for example anthropic · claude-haiku-4-5 for a cheap, fast start), type your message and press Enter.

That's it — you're set

The answer types itself out live, and the real token cost appears underneath the message when it finishes.

What it can do

Several chats at once

Use + New chat in the sidebar to start a fresh conversation. Nothing is lost — every chat stays in the sidebar, ordered by most recent activity. Click one to open it and load its history.

Automatic titles

Your first message becomes the chat's title (up to 48 characters). There's no need to name anything by hand.

Deleting a chat

Hover a chat in the sidebar and a delete button appears. You're asked to confirm first.

Stopping a response

While an answer is streaming you can press Esc or click Stop. Stopped requests cost nothing — the backend only charges your wallet after a successful completion event.

Reasoning and web search

Models that support these show their own chips in the chat header:

  • Reasoning effort — the GPT-5.4 family and Claude Opus/Sonnet 4.5+
  • Web search — currently the GPT-5.4 family only

Keyboard shortcuts

Send the message
Sends what's in the composer.
Enter
New line
Adds a line break in the composer.
Shift+Enter
Stop streaming
Stops an answer that's still coming in.
Esc

Limits

Attaching files

You can attach up to 3 files per message, each under 25 MB — images (JPEG, PNG, WebP), documents (PDF, Word, Excel, PowerPoint) and plain text. Charts inside Word and PowerPoint are not read; convert those to PDF first. Attaching a file needs credit in your wallet.

  • Each message can be at most 200,000 characters.

Security and privacy

  • Your messages only reach the AIWidget server when you press Send.
  • Message content passes through the provider of the model you chose to generate the answer. The billing ledger stores usage metadata only — model, token counts, cost — never the text itself. Chat history is kept in AIWidget's database purely so you can read it back.
  • Signing out clears every chat from local storage, but the server-side history stays until you delete it yourself.
ChangelogContact usTermsPrivacy