Table of Contents

Ollama Setup

Ollama is an open-source tool for running large language models locally. WindowSill supports Ollama as an AI provider, giving organizations full control over their AI infrastructure without relying on cloud services.

Configuration Paths

WindowSill offers three ways to configure Ollama, each with different network requirements and scenarios:

Scenario A: Administrators restrict AI providers and models to a defined list

Method Request Flow Network Requirement Best For
Dashboard App → WindowSill server → Ollama Ollama must be publicly accessible Centralized management, cloud-hosted Ollama
WindowSill App App → Ollama directly Ollama reachable from user's device Centralized management while hosting Ollama on a Private network or local installation

Scenario B: Administrators let user configure AI providers and models

Method Request Flow Network Requirement Best For
Registry App → Ollama directly Ollama reachable from user's device Per-organization member configuration, while using Ollama on a Private network or local installation

Dashboard Configuration

When you configure Ollama through the WindowSill Dashboard, all LLM requests are proxied through the WindowSill web server:

WindowSill App  →  WindowSill Server (getwindowsill.app)  →  Your public Ollama Server

Requirements:

  • Your Ollama server must be publicly accessible from the internet (with or without restrictions).
  • The WindowSill server IP (51.77.212.201) must be able to reach your Ollama endpoint.
  • HTTPS is strongly recommended.

Advantages:

  • Centralized configuration for all organization members.
  • No client-side setup required.
  • Ollama's endpoint is never stored on the client's machine.

Disadvantages:

  • You can not use an Ollama server that is deployed on a private enterprise network or on user's local machine. The server has to be public on the internet.

How to configure (step-by-step)

  1. Navigate to WindowSill Dashboard.
  2. Select your organization and navigate to WindowSill tab, then AI Providers & Models.
  3. Change Configuration Mode to Restrict AI providers and models to a defined list.
  4. Click Add an AI Provider.
  5. Select Ollama (public server)
  6. Enter the URL or IP address of your public Ollama server, then click Add.
  7. Then, click Add Model and Ollama (public server) to select the model(s) you want to allow your clients to use.

WindowSill App Configuration

When you configure Ollama through the WindowSill App, the WindowSill app connects directly to your Ollama server:

WindowSill App  → Your public, private or local Ollama Server

Requirements:

  • Your Ollama Server must be reachable directly from the client's machine. It can be done by either hosting Ollama on the user's machine, deploying it on a private enterprise server, or a public server.

Advantages:

  • Centralized configuration for all organization members.
  • You can use an Ollama server deployed on a private enterprise network or on user's local machine.
  • No internet exposure required.

Disadvantages:

  • If the list of models change on the Ollama's server, or from a local machine to another, the WindowSill dashboard won't be aware of it until the configuration has been updated by an Administrator.
  • Ollama's endpoint is always stored on the client's machine.

How to configure (step-by-step)

  1. Navigate to WindowSill Dashboard.
  2. Select your organization and navigate to WindowSill tab, then AI Providers & Models.
  3. Change Configuration Mode to Restrict AI providers and models to a defined list.
  4. On your Windows desktop, install and run the WindowSill app.
  5. Sign-in in the app using an organization Administrator account.
  6. Right-click on the WindowSill bar, Settings.
  7. Navigate to Account and ensure you select the Organization for which you want to edit the Ollama's settings.
  8. Click Sync now to refresh your settings (just in case).
  9. Go to AI Writing & Analysis, AI Providers, WindowSill AI Pro.
  10. If you are administrator of the selected organization, you will see an Administrator Zone appearing at the bottom.
  11. Click Configure & Test Ollama.
  12. Ensure the displayed organization name corresponds to the one you wish to edit.
  13. Enter your public, private network or local Ollama server's endpoint, then click Connect
  14. If the connection was successfully established, enter an LLM prompt for test, select a model, then click Test. An LLM request will be send to the server. This is to ensure the WindowSill app can successfully interact with your potentially private or local Ollama server.
  15. If the test was successful, click Upload Configuration to finish.
  16. Navigate back to WindowSill Dashboard. Refresh the page and navigate to WindowSill tab, then AI Providers & Models.
  17. Confirm that you see Ollama (private network or local) listed in Providers, and that all the Ollama's models are listed.
  18. Delete the model(s) you do not want to allow your clients to use.

Registry Configuration

When you configure Ollama via Windows Registry, the WindowSill app connects directly to your Ollama server:

WindowSill App  → Your public, private or local Ollama Server

Requirements:

  • Your Ollama Server must be reachable directly from the client's machine. It can be done by either hosting Ollama on the user's machine, deploying it on a private enterprise server, or a public server.
  • Registry keys deployed via Group Policy, Intune, PowerShell or similar.

Advantages:

  • You can use an Ollama server deployed on a private enterprise network or on user's local machine.
  • No internet exposure required.
  • Customization of the configuration is possible for each member of an organization.

Disadvantages:

  • No centralized configuration.
  • Ollama's endpoint is always stored on the client's machine.

How to configure

See Registry Keys for configuration details.