Hi,

I would like to ask whether Starveri API is in scope for the OpenModels registry before preparing any data PR.

Starveri API is an independent OpenAI-compatible GPT/Codex API gateway. It is not affiliated with OpenAI. The public API exposes model and pricing metadata, and the buyer path uses prepaid credits.

Public references:

• Buyer/pricing page: https://api.starveri.net/credits
• Docs with pricing table: https://api.starveri.net/docs
• Public model metadata: https://api.starveri.net/models
• OpenAPI: https://api.starveri.net/openapi.json
• LLM discovery: https://api.starveri.net/llms.txt

Potential provider shape:

id: starveri
name: Starveri API
description: Independent OpenAI-compatible GPT/Codex API gateway with public model metadata and prepaid credits.
api_base_url: https://api.starveri.net/v1
auth_type: bearer
compatibility: openai

The current public model IDs are:

• gpt-5.5
• gpt-5.4
• gpt-5.4-mini
• gpt-5.3-codex
• gpt-5.3-codex-spark
• gpt-5.1-codex

Pricing is published per 1M tokens and includes input, output, and cached-input prices. The /models endpoint also notes that context windows are subject to Codex-available limits.

Questions before a PR:

1. Are independent OpenAI-compatible gateway providers in scope for OpenModels?
2. Should Starveri gpt-5.4, gpt-5.4-mini, and gpt-5.5 be mappings to the existing canonical model IDs, or should gateway-served models be represented separately?
3. Should Starveri-specific gpt-5.1-codex, gpt-5.3-codex, and gpt-5.3-codex-spark get new canonical model files first, or should they be omitted until there is a canonical upstream model definition?
4. How should cached-input pricing map to the schema: cache_read_per_million, or another field?
5. For rate limits, should the mapping use the highest published paid tier, the default paid tier, or omit tier-specific limits if a single mapping cannot represent tiering?
6. Should context_window_override stay null until exact Starveri-specific token limits are published?

I am not providing an API key here and am not claiming OpenModels has tested Starveri end-to-end. If this is in scope, I can prepare a minimal PR using only the public docs and model metadata above.