Fictional default OpenAI model IDs cause change summaries and triage to crash with 404 error

[Adityakk9031]

Describe the bug

The default model options configured for the openai and vercel_gateway AI providers are set to fictional/mock model IDs (gpt-5.4-nano, gpt-4.1-nano, and gpt-4.1-mini). Because these models do not actually exist on the OpenAI API, any check that triggers an AI change summary or triage using the default configuration fails immediately with a model-not-found HTTP 404 error.

Steps to reproduce

1. Configure an OPENAI_API_KEY on the server or in per-account Settings.
2. Enable AI Change Summary on any monitor and run a check that detects a change.
3. The LLM completion request fails, throwing a model-not-found error, and no change summary is generated.

Expected behavior

The default models and recommendations should point to real, production-ready, and cost-effective models hosted by OpenAI (such as gpt-4o-mini as the default/recommended, and gpt-4o).

Code Location

In ai-models.ts:L5-L16:

export const AI_MODELS_BY_PROVIDER: Record<AiProvider, AiModelOption[]> = {
  openai: [
    { id: "gpt-5.4-nano", label: "GPT-5.4 Nano (recommended)" },
    { id: "gpt-4.1-nano", label: "GPT-4.1 Nano" },
    { id: "gpt-4.1-mini", label: "GPT-4.1 Mini" },
  ],
  vercel_gateway: [
    { id: "openai/gpt-5.4-nano", label: "GPT-5.4 Nano (recommended)" },
    { id: "openai/gpt-4.1-nano", label: "GPT-4.1 Nano" },
    { id: "openai/gpt-4.1-mini", label: "GPT-4.1 Mini" },
  ],
};