Multi Model Config

[amphetamarina]

Related Issue: #47

• Refactor ifs/elifs into separate prepare{model} functions

  • [ ] Discuss if this refactor makes sense

• Adds 'custom' option on devon configure, giving the user freedom to choose whatever LiteLLM-compatible model they want.

• Example .devon.config working with mixtral using OpenRouter:

{
  "modelName": "openrouter/mistralai/mixtral-8x22b-instruct",
  "apiKey": "sk-or-v1-<openrouter-api-key>",
  "apiBase": "https://openrouter.ai/api/v1",
  "promptType": "llama3"
}

[amphetamarina]

So, might have going out of the way with the refactorings. Also pulled from main and added changes so not to break ollama support.

I'm doing more comprehensive regression testing here, just wanted to know if the refactor is a GO or not, I can just change back to ifs/elifs if it doesn't make sense to extract them now.

[akiradev0x]

This is great work! Will review tomorrow. Thanks!