hermes - 💡(How to fix) Fix Copilot /models context display uses prompt limit instead of context window [1 pull requests]

Hi Hermes team — I found a small Copilot metadata mismatch while checking the live "/models" catalog.

What happens

For GitHub Copilot models, Hermes currently reads capabilities.limits.max_prompt_tokens as the model context size. The newer Copilot catalog can expose a separate max_context_window_tokens, which is the value users expect to see as the context window.

Example live shape from the Copilot /models endpoint:

{
  "id": "gpt-5.5",
  "capabilities": {
    "limits": {
      "max_context_window_tokens": 400000,
      "max_prompt_tokens": 272000,
      "max_output_tokens": 128000
    }
  }
}

With the old logic, /model and context resolution display 272k instead of the actual 400k context window.

Expected behavior

Hermes should prefer max_context_window_tokens when present, while keeping max_prompt_tokens as a fallback for older catalog payloads.

Notes

I have a minimal fix ready that:

• uses max_context_window_tokens first
• falls back to max_prompt_tokens when needed
• adds coverage for the distinct 400k context / 272k prompt case

Thanks for maintaining Hermes — this should be a very small, backwards-compatible cleanup.