-
Notifications
You must be signed in to change notification settings - Fork 5
Tutorials
Welcome to Kaas, this guide will walk you through the installation process and show you how to start chatting with a remote LLM in no time.
- Download the latest release for your operating system from the Releases page. Kaas currently supports Windows(64-bit), macOS and Linux.
- Install the application:
- Windows: Run the Microsoft Installer (.msi file) or NSIS Installer (-setup.exe file)
- macOS: Run the .dmginstaller
- Linux: Use the appropriate package manager for your distribution (e.g., .deb for Debian/Ubuntu, .rpm for Fedora)
- Launch Kaas from your applications menu or desktop shortcut.
-
Hover over the side navigation bar on the left and click Models. This will bring you to the Models page.
-
Click Create new model button, it will open a dialog that asks you to pick a provider you want to create model with
-
Each provider requires a difference set of fields to be filled, so that you can successfully access its APIs. For example, to use OpenAI, you need to fill it your API Key (kept in a database on your device) and name of the model you want to use. For OpenAI and Ollama, you can load available models by clicking the **Load from API" button, or you can choose to fill in the model name manually.
Note: if you have enabled Proxy on the Settings page, the request to load models will go through proxy too. -
Once you filled all the required fields, click Save.
-
You should now see the model you just created under the Your Models section.
-
You can create multiple models from the same provider and use the in different conversations.
-
Make sure you have created at least one model
-
Hover over the side navigation bar on the left and click Conversations. This will bring you to the Conversations page.
-
Click "Start new conversation" button
-
It will ask you to pick a model for the new conversation, pick one and click Use this model
-
A conversation with default title "New conversation" is created, you can now type your question in the input box, upload an image (if the model supports vision) or use a prompt template for more complex prompt (how to define a prompt template).