Add Requesty as a first-class OpenAI-compatible gateway #2240

Summary

Add Requesty as a first-class OpenAI-compatible gateway in OpenClaude, following the existing gateway descriptor pattern (same shape as the OpenRouter gateway).

Requesty is an OpenAI-compatible LLM router: one key for several hundred models from OpenAI, Anthropic, Google, DeepSeek, xAI and others, with per request cost reporting, fallback policies and an EU endpoint.

Docs: https://docs.requesty.ai
Model list: https://router.requesty.ai/v1/models (public, no key needed)

Problem

Requesty already works through the generic OpenAI-compatible route, but users have to type the base URL, key and model ids by hand, and there is no model discovery or dedicated credential for it.

Proposed Direction

Smallest useful version, mirroring src/integrations/gateways/openrouter.ts:

• Gateway id requesty, label Requesty, transport openai-compatible
• Base URL https://router.requesty.ai/v1
• Credential env REQUESTY_API_KEY, Bearer auth
• Hybrid catalog: a short curated list verified against the live catalog, plus discovery from GET /v1/models (same response shape as OpenAI, with context_window and max_output_tokens per model)
• Tests and docs lines matching what the OpenRouter gateway has
• Not added to any default or automatic selection

The EU endpoint (https://router.eu.requesty.ai/v1) could be documented as a base URL override rather than a second gateway.

Alternatives Considered

The generic OpenAI-compatible route with a manual base URL, which is what users do today.

Additional Context

I'm putting together a branch that follows the gateway descriptor pattern and will test it against the live API. I'll link it here as a PR, and I'm happy to adjust scope first if you'd prefer something different.

Disclosure: I work at Requesty.