feat: add Requesty as an OpenAI-compatible binding

[Thibaultjaigu]

Add Requesty as an OpenAI-compatible binding

This PR adds a new requesty binding to zoos/bindings_zoo/, mirroring the existing open_router binding as closely as possible. Requesty is an OpenAI-compatible LLM gateway/router, so the request/response shapes and the binding logic are identical to OpenRouter.

What it does

• Base URL: https://router.requesty.ai/v1 (chat at /v1/chat/completions, model list at /v1/models)
• Auth: Authorization: Bearer <key> via a requesty_key config field (or the REQUESTY_API_KEY environment variable)
• Model naming: provider/model (e.g. openai/gpt-4o-mini)
• The /v1/models endpoint is OpenAI-shaped (a data array), consumed by the models scraper

New files (zoos/bindings_zoo/requesty/)

• __init__.py — class Requesty(LLMBinding), binding_name = "Requesty", binding_folder_name = "requesty", DEFAULT_CONFIG with requesty_key, override_api_url = "https://router.requesty.ai/v1", API_URL = "https://router.requesty.ai/v1/models". All LLMBinding method logic is identical to the OpenRouter binding (OpenAI-compatible).
• models_scraper.py — class RequestyModelFetcher, API_URL = "https://router.requesty.ai/v1/models". Same OpenAI-shaped data-array parsing as the OpenRouter scraper. Requesty's model objects expose context_window and capability flags (supports_vision, etc.) instead of OpenRouter's architecture/pricing objects, so the field reads are guarded defensively with .get() and sensible fallbacks (mirroring the OpenRouter scraper's defensive style). The output YAML structure is preserved field-for-field.
• binding_card.yaml — Requesty metadata.
• models.yaml — generated by running the new scraper against the live /v1/models endpoint (550 models).
• requirements.txt, .gitignore — copied unchanged from the OpenRouter binding.

Discovery / registry

Bindings in zoos/bindings_zoo/ are auto-discovered by directory scan — there is no central binding registry/list to edit. A grep for open_router/openrouter across *.py/*.yaml/*.json (excluding the binding dir and tests) found only scraped model artifacts, no code-level registry. So the new directory is sufficient.

Logo

The OpenRouter binding ships logo.png/logo2.png. I intentionally omitted these rather than ship the OpenRouter-branded image under a Requesty binding; the binding falls back to the default icon. Happy to add a proper Requesty logo if you'd like one included.

Verification

• ast.parse on both __init__.py and models_scraper.py — OK.
• yaml.safe_load on binding_card.yaml and models.yaml — OK.
• Confirmed the requesty/ directory mirrors open_router/ file-for-file (minus the omitted logos), with all OpenRouter identifiers/URLs/keys renamed.
• Live POST https://router.requesty.ai/v1/chat/completions (model openai/gpt-4o-mini) → HTTP 200, real completion returned.
• Live GET https://router.requesty.ai/v1/models → HTTP 200, data array with 572 models (proves the scraper path works end-to-end).

Docs: https://requesty.ai , https://docs.requesty.ai

I work at Requesty. This mirrors the existing OpenRouter provider as closely as possible. Happy to adjust or close it if it's not a fit.