Chatting with the AI

The chat workspace is where you talk with Grengin’s AI. Type a question, attach photos or files, pick a model, turn on extra abilities like web search, and read replies as they stream in. This guide walks through everything you can do, task by task.

See also: Guide home · Settings. Administrators can also read the Control Hub guide.

Start a new chat

Select New Chat in the sidebar (it’s icon-only when the sidebar is collapsed).
You’ll land on a fresh welcome screen with a floating chat-bubble icon and a short greeting.
The message box is focused and ready. Just start typing.

Your past conversations live in the sidebar under Chats. Select any one to reopen it; while it loads you’ll see a brief loading placeholder.

Write and send a message

The message box sits at the bottom of the screen and grows as you type.

Type your message. The box expands to fit longer text.
Press Enter to send.
To add a new line without sending, press Shift+Enter.

You can also select the Send button (the arrow) to send. It’s disabled while the box is empty, and shows a spinner while your message is sending or files are still uploading.

A short AI disclaimer sits beneath the box as a reminder that replies can be imperfect.

Heads up: Once a reply starts, it can’t be stopped. There’s no stop button. Let it finish, then send a follow-up if you need to redirect.

Attach photos and files

You can send images and documents along with your message.

Select the + (Add content) button next to the message box.
Choose Add photos or Add files.
Pick the items you want. They upload right away and appear as pills above the message box.

Each attachment pill shows its name, size, and an upload status:

A spinner while it’s uploading
A checkmark when it’s done
An x if the upload failed

Preview before sending:

Select an image thumbnail to open a full-image preview.
Select a file to open a read-only text preview (for text files).
Close any preview with its x, by selecting outside it, or by pressing Esc.

Remove an attachment: select the x on its pill.

The + button is unavailable while a message is sending.

Choose the AI model

Different models have different strengths. The current model’s name appears as the message box placeholder.

Select the model selector (the provider icon, model name, and caret near the message box).
Pick a model from the menu. Models are grouped by provider.
A small eye icon marks models that can understand images, so look for it if you’re attaching photos.

If models are still loading or fail to load, the menu tells you.

Turn on Connectors (tools)

Connectors let the AI reach out to your connected services and tools to get things done.

Select the Tools (Connectors) button near the message box.
Toggle on the connectors you want the AI to use for your next message.
Send your message as usual.

If you have no connectors set up, the menu says so. You manage which services are connected from your Settings.

When a tool needs you to sign in to an outside account, the AI shows a connect prompt right in the conversation, with the tool and service named (and the permissions it’s requesting). Select Connect account to open a sign-in popup, or Not now to skip it. You’ll see a green confirmation when it connects, or a red message with a retry option if it doesn’t.

Turn on Web search

Let the AI look things up online for fresher or broader answers.

Select the Search toggle (the globe) near the message box.
It highlights when it’s on.
Send your message. The AI can now search the web while answering.

Dictate with your voice

Prefer to speak? Use voice dictation to turn speech into text.

Select the microphone button near the message box.
It pulses red while it’s listening, and the placeholder shows the recording state.
Speak. Your words are transcribed into the message box, where you can edit before sending.

The microphone is unavailable while a message is sending. (Voice input depends on your browser, so it may not appear everywhere.)

Read the AI’s reply

Replies appear in the conversation above the message box and stream in as they’re written.

A three-dot typing indicator shows while the AI is preparing its answer.
A blinking cursor follows the text as it streams.
The view scrolls automatically to keep the newest text in sight. If you scroll up to read something, auto-scroll pauses so you don’t get yanked back down; scroll to the bottom to resume.

Remember: a reply can’t be interrupted once it begins.

Copy a message or a code block

Copy a whole message: hover over it (desktop), or focus or tap it (mobile), to reveal its action toolbar, then select Copy. You’ll see a brief checkmark to confirm. This works on both your messages and the AI’s. The toolbar stays hidden while a reply is still streaming.

Copy a code block: code in a reply has its own Copy button in its header (which also shows the language). Select it and it changes to Copied! On smaller screens the label is hidden, but the button still works.

Listen to a reply aloud

You can have the AI’s replies read out loud (where your browser supports text-to-speech).

On an AI reply, open its action toolbar (hover, focus, or tap).
Select Listen.
While it speaks, you can Pause, Resume, or Stop speaking from the same toolbar.

View images and files in a reply

When a reply includes attachments:

Images appear as thumbnails in a grid (with a brief placeholder while they load). Select one to open a full-screen preview. Close it with the x, by selecting the backdrop, or by pressing Esc.
Files appear as a labeled box with an icon for their type. Select it to download or open the file in a new tab.

See web-search sources

When the AI searched the web, a Searched the web card appears with the results it used.

The card shows a green check when the search is done (or a spinner while it’s running).
Select the card header to expand it.
Inside you’ll see the search query, a count of results, and each source as a link with its title and site. Select any source to open it in a new tab.

See tool-call results

When the AI uses a connector or tool, its work shows up as a step in the conversation.

Each step shows a status (check, spinner, or red x for an error), the tool’s name, which service it ran on, and how long it took.
Select the step to expand its details. You can view the inputs and the results, switch between a Formatted and a Raw (JSON) view, and Copy the result.
Results are shown in whatever shape fits best: tables, key-value cards, lists, plain text, inline images, or file references. If a tool fails, a red error row explains what went wrong.

When something goes wrong

Error banners appear above the message box. Budget warnings show in amber; other errors show in red. Dismiss a banner with its x. (Banners stay hidden while a reply is streaming.)
Per-message errors show as a small red row attached to the affected message.
Brief pop-up notices (toasts) confirm actions or report failures, and some are clickable to take you straight to the relevant place.

If a reply seems to stall, remember you can’t stop it from the screen. Wait for it to finish, then try again or rephrase in a new message.