AI Chat — Overview

AI Chat is VyroAI.vip's general assistant. How conversations are saved, renamed, starred, exported or linked to a chat project, and how to pick the model.

What it is

AI Chat is VyroAI's general assistant — the place to ask questions, think something through, or get help that is not tied to a project. It runs on the leading frontier models, and you choose which one.

Conversations

Every chat is saved as a conversation in the sidebar, so you can come back to it later. Opening one picks up exactly where you left off. The open conversation is part of the address — /chat?c=… — so refreshing the page reopens it, and you can bookmark it to come straight back. The link only opens the conversation on your own account; anyone else gets an empty chat.

What you can do with one

The menu on a conversation in the sidebar offers more than most people find:

Rename it, and star it so it stays near the top
Export it — as Markdown, plain text or JSON, which is the one people go looking for when they want to keep an answer outside VyroAI
Duplicate it, to branch off a long conversation without losing the original
Link to Project — one of your chat projects, the name-and-colour groups described below. Once one is linked the entry reads Change Project, with Unlink beside it. (Pointing a conversation at your code is a different feature, done from inside the chat itself — see Connecting a Chat to a Project.)
Delete it

The sidebar itself moves

You can drag the conversation sidebar wider or narrower, and move it to the other side of the screen if you would rather have your chats on the right.

Chat projects — the part that does steer the AI

Conversations can be grouped into a project: give it a name, a colour and, most importantly, instructions. Every message in a conversation filed under that project is sent with those instructions, so "answer in Dutch", "we are on Paper 1.21" or "keep answers short" holds for the whole group without being retyped.

The sidebar filters by All Chats, Starred and Projects, and a conversation is moved into one from its own menu. This is a different thing from Connecting a Chat to a Project, which points a conversation at a Code Studio project so it can read your files.

The per-conversation settings do nothing yet

Do not confuse the above with the conversation settings dialog, which offers its own instructions, a temperature slider and a max tokens ceiling. None of the three changes any answer: the instructions are saved and then ignored, and the temperature and max-tokens values are not even saved — close the dialog and they are gone. Every answer uses the same standard limits regardless of what you set.

To steer one conversation, file it under a chat project and put the rule in the project's instructions. Inside the Code Studio, AI Memory does the same job — see Project Memory.

You do not have to wait

You can keep typing while the assistant is still answering. Extra messages line up in a queue — up to 10 — and are sent one after the other, so a train of thought does not have to wait for the reply.

What else it can do

Chat is not only text:

Voice — talk to it instead of typing
Images — generate pictures without leaving the conversation
Charts — ask for numbers to be drawn and the answer comes back as a chart
Project context — point a chat at one of your projects so it is told its name and its top-level files and folders

Most of these have their own article.

Ask it where something is

A question like "where do I add a dependency?" gets a different kind of answer: a small card instead of a paragraph. It carries a path strip — Plugin editor › Tools › Dependency Manager — then numbered steps, ending with how you know it worked, and every page it names is a button you can press to go straight there.

It is the fastest way to find a corner of VyroAI you have not used before, and it beats hunting through menus. If the assistant genuinely does not know, the card says so and points at the nearest pages and at support, rather than guessing.

The Web search switch does not search the web

The Web search switch in the + menu beside the message box, under Add to your message, looks like it connects the assistant to the internet. It does not. Turning it on adds one sentence to your message saying you would like search results, and the server has no search of any kind to act on it — the model answers from what it already knows, exactly as it would with the switch off.

Treat it as decoration. When you need something current, paste the page or the figures into the message yourself.

Chat has its own memory, separate from a project's

AI Chat keeps an account-wide memory: rules you have saved that are applied to every conversation, whichever model you are on. It is the same memory system the Code Studio uses, sharing one list — but general chat is deliberately shown only the rules that belong to no project, so instructions written for one plugin do not leak into an unrelated conversation.

The assistant can also propose a new rule mid-conversation; it arrives as a card you approve or dismiss, and nothing is remembered until you approve it. Project Memory describes the same system from the Code Studio side, including the private half the AI never sees.

The controls on the page itself

Seven things on the chat screen have no label to look up, so here they are:

Retry — the circular arrow in the footer of the newest answer, beside Copy and the 👍 / 👎. It asks for that answer again. Only the newest reply can be retried, and the button is missing when the message it would replay carried attachments. The right-hand end of the same footer names the model that answered and what the answer cost.
Stop generating — while an answer is streaming, the Send button becomes a black square. Press it and the answer stops where it is.
Thoughts — on a model that shows its reasoning, a folded row appears above the answer: Thinking… with a pulsing brain while it works, Thoughts once the answer starts. It stays folded unless you open it.
Documents — once the AI has written a titled code block, a Documents chip with a count appears in the message box, beside the model chip. (The bar that used to sit above the messages is gone.) It opens Documents in this chat: press a name to unfold the file, or the copy icon beside it to copy the whole thing. It is a list of what the AI made in this conversation, read straight out of the messages.
Show N earlier messages — a long conversation folds everything but the newest 40 messages behind one line at the top. Press it and they are all back; nothing is hidden for good.
Jump to latest — a pill that appears when you scroll up. It reads N new messages if replies arrived while you were reading, and one press takes you to the bottom.
Your draft is kept — text in the box that you have not sent survives leaving the page and coming back in the same tab. It is kept in the tab, so closing the tab is what clears it.

The model chip in the message box lists a cost multiplier next to each model (1×, 2× and so on) rather than a credit price — it tells you how much dearer one model is than another, not what the next message will cost.

Also in AI Chat

Next in AI Chat: Voice Conversations
Ways to talk to it: Connecting a Project — Let a conversation know which project you mean — and what it can and cannot see.
Ways to talk to it: Asking It to Make Something — Ask AI Chat to make, build or create something and you get guided cards instead of a chat answer. The six tools, what costs credits, and when the cards stay away.
What comes back: Images inside Chat — Generate pictures inside AI Chat without switching pages, including wearable Minecraft skins with a 3D view, and when the Image Creator is the better pick.
What comes back: Charts in Chat — Ask for numbers and the answer can come back drawn instead of written.

1 of 10 in AI Chat · Last checked 11 September 2026 · Previous: Crashes & Projects That Will Not Open