Is your feature request related to a problem? Please describe.
Currently, KeyType requires downloading models locally. This works well for many users, but it makes it difficult for people who already manage their models through an Ollama server.
For example, I already have multiple LLMs running in Ollama and shared across different applications. Having to download the same models again for KeyType wastes disk space and requires maintaining two separate model environments.
Describe the solution you'd like
I'd like KeyType to support connecting to an Ollama-compatible API endpoint (e.g. http://localhost:11434 or a remote Ollama server) instead of requiring locally downloaded models.
The workflow could be:
- Configure an Ollama endpoint URL.
- Automatically list available models from the server.
- Allow selecting one of those models for typing generation.
- Continue using the existing local model workflow as an option.
This would make KeyType more flexible while remaining fully compatible with the current implementation.
Describe alternatives you've considered
The current workaround is to download the same model specifically for KeyType, even if it already exists in Ollama. This duplicates storage usage and makes model management less convenient.
Additional context
Many AI desktop applications now support Ollama as a backend because it provides a standard local API and allows multiple applications to share the same models.
Supporting Ollama would also make it possible to use remote inference servers (on a local network or another machine) without copying large models to every device.
Is your feature request related to a problem? Please describe.
Currently, KeyType requires downloading models locally. This works well for many users, but it makes it difficult for people who already manage their models through an Ollama server.
For example, I already have multiple LLMs running in Ollama and shared across different applications. Having to download the same models again for KeyType wastes disk space and requires maintaining two separate model environments.
Describe the solution you'd like
I'd like KeyType to support connecting to an Ollama-compatible API endpoint (e.g.
http://localhost:11434or a remote Ollama server) instead of requiring locally downloaded models.The workflow could be:
This would make KeyType more flexible while remaining fully compatible with the current implementation.
Describe alternatives you've considered
The current workaround is to download the same model specifically for KeyType, even if it already exists in Ollama. This duplicates storage usage and makes model management less convenient.
Additional context
Many AI desktop applications now support Ollama as a backend because it provides a standard local API and allows multiple applications to share the same models.
Supporting Ollama would also make it possible to use remote inference servers (on a local network or another machine) without copying large models to every device.