Documentation
Quickstart

Quickstart

Step 1: Install Jan

You can run Jan either on your desktop using the Jan desktop app or on a server by installing the Jan server. To get started, check out the Desktop installation pages.

Once you have installed Jan, you should see the Jan application as shown below without any local model installed:


Default State


Step 2: Turn on the GPU Acceleration (Optional)

If you have a graphics card, boost model performance by enabling GPU acceleration:

  1. Open Jan application.
  2. Go to Settings -> Advanced Settings -> GPU Acceleration.
  3. Click the Slider and choose your preferred GPU.
  4. A success notification saying Successfully turned on GPU acceleration will appear when GPU acceleration is activated.

Ensure you have installed your GPU driver. Please see Desktop for more information on activating the GPU acceleration.


Turn on GPU acceleration

Step 3: Download a Model

Jan offers various local AI models tailored to different needs, all ready for download directly to your device:

  1. Go to the Hub.
  2. Select the models that you would like to install. To see model details, click the model name.
  3. You can also paste the Hugging Face model's ID or URL in the search bar.

Ensure you select the appropriate model size by balancing performance, cost, and resource considerations in line with your task's specific requirements and hardware specifications.

  1. Click the Download button.

Download a Model


  1. Go to the Thread tab.
  2. Click the Model tab button.
  3. Choose either On-device or Cloud section.
  4. Adjust the configurations as needed.

    Please see Model Parameters for detailed model configuration.


Parameters

Step 4: Customize the Assistant Instruction

Customize Jan's assistant behavior by specifying queries, commands, or requests in the Assistant Instructions field to get the most responses from your assistant. To customize, follow the steps below:

  1. On the Thread section, navigate to the right panel.
  2. Select the Assistant tab menu.
  3. Provide a specific guideline under the Instructions field.

Assistant Instruction


Step 5: Start Thread

Once you have downloaded a model and customized your assistant instruction, you can start chatting with the model.


Chat with a Model


Step 6: Connect to a Remote API

Jan also offers access to remote models hosted on external servers. You can link up with any Remote AI APIs compatible with OpenAI. Jan comes with numerous extensions that facilitate connections to various remote AI APIs. To explore and connect to Remote APIs, follow these steps:

  1. On the Thread section, navigate to the right panel.
  2. Select the Model tab menu.
  3. Next to the OpenAI models -> click the Gear Icon (⚙️).
  4. Enter your OpenAI API Keys.

Connect Remote API


What's Next?

Now that Jan is up and running, explore further:

  1. Learn how to download and manage your models.
  2. Customize Jan's application settings according to your preferences.