The welcome
The first time Your Own AI opens, a welcome takes three steps: get your first model, meet your AIs, go - with an optional Flowsta sign-in on the last one. It is built so that nothing blocks you - you can be chatting before the download finishes.
Step 1 - your first model
The welcome offers one model: the one graded best for this computer, sized to your memory and graphics card (the biggest one, on a computer with several). Nothing on the page leads away from it. Models run on this computer, nothing you say leaves it, and there is nothing to pay. Download and continue starts the download and moves on.
The words are honest about small machines. A pick that will run on your processor says "good for everyday chat, slower on long answers". A machine with less memory than the smallest model needs is told so, in amber, and the optional online plan is named as the fast route - but the download still goes ahead and everything else in the app works as normal.
Used the app before, with a Flowsta Vault backup? A note on this step says where to restore it: Settings, under Your Flowsta Identity - see Backups and recovery.
Step 2 - meet your AIs
You start with three AIs, and all three share the model that is downloading. Pick the set that fits you:
| Set | The three |
|---|---|
| Personal | Veebo (playful and curious), Teresa (warm and supportive), Reeves (straight answers, no personality in the way) |
| Work | Assistant (writing, planning, email drafts), Coder (code and debugging), Analyst (data, summaries, documents and numbers) |
A set is an edit of the same three AIs, never a fourth. Change any name, picture or personality right here - the names are yours to type over. The Change tile offers six personalities to start with; all eighteen, and your own, live on the Your AIs page, where you can also put your own photo or a generated one on any AI.
Step 3 - go
The last step offers, optionally, to connect your Flowsta Vault: "Backs up your AIs and conversations, syncs them between devices, and signs you in to online models. Free, no password." Depending on what is on this computer the button reads Sign in with Flowsta, Open Flowsta Vault or Get Flowsta Vault - free; a locked Vault is told to unlock first, and later is fine - it is in Settings under Your Flowsta Identity. Signed in, the step says so by name.
If the model has landed, the page says You're all set and Start chatting opens the chat.
If it is still downloading, the page says so with the percentage, and Take me to my AIs opens the chat anyway. You do not have to wait: ask any of your AIs a question and the answer arrives the moment the model lands. The message field shows the progress while you wait, and a note appears when the model is ready.
The download never starts over
- Close the app mid-download and the welcome picks it up where it stopped next time, never from zero.
- A download still running when you come back is joined, not restarted.
- If the download fails, the welcome says so with the reason, offers Try the download again, and your AIs are ready either way - they only need a model, which you can also pick on the Offline Models page later.
After the welcome
The home page offers the small memory model once, and from the second launch may offer the optional helper when none of your installed models is graded well enough for memory work. Neither is part of the first download; see The helper model. Then carry on with Your first chat.
Replies keep going when you leave
Ask a question, go to another page, and the answer is there when you come back. A reply or an agent turn keeps running on any page; opening the chat again shows the running reply at once, with your question at the top, and a reply that finished while you were away in the last minute opens the same way. One you watched finish is left where it is.
The activity tray
Everything running that you are not looking at shows in one tray at the bottom right, on every page: replies and agent turns on other pages, model, engine and Projects helper downloads, the documents being read. It never repeats the page you are on.
A reply row says who is doing what: "Teresa is answering..", "Reeves is working..", "Reeves is waiting for the engine" while another AI answers first, "Teresa answered", or "Reeves needs you" when an agent is waiting on a permission - and then the row carries Allow and Don't allow, so you can answer from wherever you are. Every reply row has Open, and Stop while it runs. A finished reply stays listed for a minute unless you open it; a download row says its progress, then "Downloaded", and the NVIDIA engine and the Projects helper say when they are installed. Finished rows clear themselves after a moment; one that could not finish stays until you dismiss it, with the reason.