It knows your documents –
and your tone
An off-the-shelf model knows nothing about your company, your clients or the way you write. In Qwirbel you store your own knowledge and train profiles from examples – both stay on your device.
起价 60 € · 一次付清，没有订阅 · 适用于 Windows、Linux 和 macOS
要点
Store your own knowledge
Your documents go into a knowledge system Qwirbel draws on – without anything being uploaded.
Profiles from examples
A handful of good examples is enough for answers to come out in your tone instead of officialese. Profiles can be kept separate and switched.
Understanding images
A vision model describes what is on a photo or screenshot – useful for receipts, error messages and templates.
Dictate instead of type
Speech recognition records and writes along. With the lean engine it needs a fraction of the memory it used to.
Profiles set the tone
How knowledge and style actually take effect
Above is what it does. Here is how – and above all what does NOT happen.
⚠ The model is not changed in the process
Neither your knowledge nor your style alters the model weights – no retraining takes place. Both are supplied at runtime: the agent searches your knowledge and is given your examples. The difference matters, because it also means you can take everything back out at any time, and none of it is irreversibly baked into a file.
What you can store
Manuals, notes, contracts, minutes. When needed the agent searches them and answers from them instead of guessing. For companies that is the difference between a generic answer and one that knows your rules – without those rules travelling to a provider.
How the style is picked up
You show a few examples of how you write, and Qwirbel keeps to them. Several profiles can exist side by side, one for customers and one for internal use. Switching costs no restart.
The long-term memory can be inspected
What Qwirbel remembers about you – preferences, recurring goals, connections – you can view, change and delete. It sits on your drive, not in an account. A program that remembers things without letting you look inside would be exactly what Qwirbel otherwise stands against.
Separate per account
In the Server edition every account has its own knowledge and its own profiles. There is no shared pool everyone reads from – otherwise one department's knowledge would show up in another's answers.
关于这一点大家会问
Are my documents uploaded anywhere?
No. Knowledge, profiles and chats stay on your device. Only if you explicitly add an external provider for the answer does the text of that request go there – files and knowledge still stay local.
Is this really fine-tuning the model?
No, and that matters: no model weights are changed. It works with examples and your knowledge as context. That is plenty for style and terminology – but it does not turn a small model into a big one.
How much knowledge can I store?
The limit is your disk space, not a plan. Large collections do need to be organised, otherwise search finds the wrong thing.
Does it survive an update?
Yes. Your data lives separately from the program. An update replaces the application, not your content.
这个 Qwirbel 也能做
所有本领都住在同一个程序里——买一次，一把钥匙。