Ollama and optional API providers
Choose a local model or add a provider in each app.
Updated 21 October 2026 · Steps describe the prepared Windows apps. Options vary by document mode.
AI is optional
Ordinary editing works without AI. Models and cloud API usage are not included with the apps. AI output can be wrong; review it before inserting or relying on it.
Connect an installed Ollama model
- Install and run Ollama separately, and download a model supported by your computer.
- Open AI assistant → Settings in the KeepSuite app you want to use.
- Choose an Ollama provider. For Ollama running on this computer, the usual local base URL is http://127.0.0.1:11434.
- Use the provider/model controls to load the available models and choose an installed model. Save the settings.
- Send a short, non-sensitive test request. Review the response before using the insert/apply action.
A local endpoint is not the same as a guaranteed local model. If Ollama is configured to use a remote service or cloud model, that service’s network and privacy practices apply.
Add your own API provider
Use AI assistant → Settings to add a supported provider, its base URL, model identifier and your API key. Add provider settings separately in each app where needed. A compatible API endpoint must support the protocol you select.
Keys are stored using Windows-protected encryption. They belong in the desktop app’s settings, not on this help website or in a support message. Requests send the prompt and included document context to the chosen provider; its costs and terms apply.
Ask, review and apply
Select the text, cells or objects you want to work on when the task supports it, then open AI assistant and choose or write a request. Review the response and the target before applying it. KeepSuite may stop an apply operation if the document changed after the request, so it does not overwrite a newer edit.
Use Undo when available if an applied result is unsuitable. Keep a native backup before asking AI to make extensive changes.
If the connection fails
- Check that Ollama is running and the chosen model is installed.
- Check the provider type, base URL and exact model name.
- For a cloud provider, confirm internet access, a valid key and provider account availability.
- If a long request is slow, try a shorter selection or smaller local model.
Do not disable protections or expose a local model server to the internet just to make a connection work. If you need support, report the provider type and error message with keys removed.
Still need a hand?
Email support@hcus.llc with the app name and a brief description.
← All help guides