[Feature] Support additional LLM providers (OpenAI, Anthropic, local models)

[ASGCompute]

The MiroFish LLM agent currently uses NVIDIA NIM with DeepSeek V3.2. Supporting other providers would make the system more accessible.

Providers to consider:

• OpenAI GPT-4
• • Anthropic Claude
• • • Local models via Ollama
• • • • Google Gemini

[IlyasFardaouix]

I think the issue might be related to the fact that the Config class in __init__.py doesn't seem to have any attributes set for the additional LLM providers (e.g. OpenAI, Anthropic, local models). The ALLOWED_EXTENSIONS attribute is used to check file extensions, but it doesn't seem to be used for LLM providers. Can you clarify how the LLM providers are currently being configured, or if there's a separate configuration file for them?