[BOT ISSUE] fix: add missing OpenAI model gpt-realtime-2

[github-actions]

Missing model

The model gpt-realtime-2 is available on the OpenAI API but absent from the model catalog (packages/proxy/schema/model_list.json). The catalog currently has gpt-realtime ($4/$16, 32K context, 4K max output) and gpt-realtime-1.5, but not gpt-realtime-2 which is a significantly upgraded model with a 128K context window, 32K max output, and updated output pricing ($24/MTok).

OpenAI describes it as "our most capable realtime voice model" with support for speech-to-speech interactions, configurable reasoning effort, stronger instruction following, and reliable tool use.

Model details

Field	Value
Model ID	gpt-realtime-2
Format	openai
Flavor	chat
Multimodal	true (audio + image input)
Input cost (per 1M tokens)	$4.00
Output cost (per 1M tokens)	$24.00
Cached input cost (per 1M tokens)	$0.40
Max input tokens	128000
Max output tokens	32000
Display name	GPT Realtime 2
Available providers	["openai"]

Verification checklist

• Cross-source confirmation: Model confirmed on (1) OpenAI model detail page with pricing, context window, and max output tokens, and (2) OpenAI models listing page as an available model.
• Recent commits check: Not added in any recent commits touching model_list.json. No match for gpt-realtime-2 as a model key in the catalog.
• ID format validation: Follows existing gpt-realtime-* convention used by sibling models gpt-realtime (line 11245) and gpt-realtime-1.5 (line 11202).

Verification notes

Field	Source
Model ID, context window, max output tokens	OpenAI model detail page
Text token pricing ($4.00/$24.00/$0.40 cached)	OpenAI model detail page
Model availability	OpenAI models listing page

Fields NOT verified from official sources:

• Azure availability — the existing gpt-realtime and gpt-realtime-1.5 entries list ["openai", "azure"], but Azure availability for gpt-realtime-2 specifically is not confirmed. The downstream fix job should verify.
• Audio token pricing ($32/$64 per MTok) is published but omitted from the spec since the catalog schema uses text token pricing.

{
  "kind": "missing_model",
  "provider": "openai",
  "models": ["gpt-realtime-2"],
  "status": "active",
  "model_specs": {
    "gpt-realtime-2": {
      "format": "openai",
      "flavor": "chat",
      "multimodal": true,
      "input_cost_per_mil_tokens": 4.0,
      "output_cost_per_mil_tokens": 24.0,
      "input_cache_read_cost_per_mil_tokens": 0.4,
      "displayName": "GPT Realtime 2",
      "max_input_tokens": 128000,
      "max_output_tokens": 32000,
      "available_providers": ["openai"]
    }
  },
  "source_urls": [
    "https://developers.openai.com/api/docs/models/gpt-realtime-2",
    "https://developers.openai.com/api/docs/models/all"
  ]
}