support for more than one ollama backend

[nilsherzig]

Is your feature request related to a problem? Please describe.
It takes some time to switch between the embeddings model and the agent model. Especially with large (70b+) models.

Describe the solution you'd like
Since the embeddings model is so small, it's possible to just have a second ollama instance running for it.

Describe alternatives you've considered
Check if its possible to keep both in memory?

[nilsherzig]

ollama/ollama#3418