Provider settings per agent

[two70]

Summary

This pull request introduces enhancements to how agent profile-specific settings are loaded and applied in the python/helpers/settings.py module. The primary change is the addition of a mechanism to load and merge settings from agent profile directories when the active profile changes.

Profile settings loading and merging:

• Added the get_profile_settings function to load and merge settings from usr/agents/{profile}/settings.json and agents/{profile}/settings.json, returning a merged dictionary or None if no settings file is found.
• Updated the set_settings function to check if the agent_profile has changed, and if so, load and merge profile-specific settings into the active settings before normalization and saving.

This way, if you want to use an expensive model, or even completely different provider for one agent, and a cheaper model for another agent you can quickly switch by just switching agents.

Examples

Simply create a settings.json file in the agent directory and set the provider and model settings.

/a0/agents/developer/settings.json:

{
  "chat_model_provider": "anthropic",
  "chat_model_name": "claude-sonnet-4",
  "util_model_provider": "anthropic",
  "util_model_name": "claude-3-5-haiku-latest"
}

/a0/agents/researcher/settings.json:

{
  "chat_model_provider": "openai",
  "chat_model_name": "gpt-4o",
  "util_model_provider": "openai",
  "util_model_name": "gpt-4o-mini"
}