feat(ai-gateway): add Tiyuvta BYOK provider

[avifenesh]

Ticket

N/A — new provider integration.

Component/Service

• Web (Frontend)
• Jawn (Backend)
• Worker (Proxy)
• Bifrost (Marketing)
• AI Gateway
• Packages
• Infrastructure/Docker
• Documentation

Type of Change

• Bug fix
• New feature
• Breaking change
• Documentation update
• Performance improvement
• Refactoring

Deployment Notes

• No special deployment steps required
• Database migrations need to run
• Environment variable changes required
• Coordination with other teams needed

Context

Add Tiyuvta as a first-class BYOK provider in the AI Gateway. Tiyuvta exposes an OpenAI-compatible chat-completions API, so Helicone can route requests using a customer-supplied Tiyuvta key while retaining its normal gateway observability and request controls.

This initial contribution registers the two live Tiyuvta chat models:

• qwen3.8-27b → qwen/qwen3.8-27b
• ornith-1.5-35b-a3b → ornith-ai/ornith-1.5-35b-a3b

Both entries include the published 262,144-token context/output limits, input/output/cached-input pricing, multimodal metadata, tool and structured-output support, and reasoning controls.

What changed

• Added the TiyuvtaProvider with bearer authentication and https://api.tiyuvta.ai/v1/chat/completions routing.
• Added Qwen3.8 27B and Ornith 1.5 35B-A3B registry/model metadata and exact cached-input cost multipliers.
• Added provider detection, database mappings, priority, OpenAI usage parsing, and Responses-to-chat compatibility.
• Added the provider settings entry, logo, and an attributed quickstart link.
• Added cost-registry and end-to-end Worker gateway coverage for routing, authentication, model translation, request controls, auto-selection, and 429 handling.

Scope

• BYOK only (ptbEnabled: false); this does not add pass-through billing or Helicone-managed Tiyuvta credentials.
• Embedding and reranking models are intentionally excluded from this first PR because they use different endpoint shapes from chat completions.

Validation

• Packages Jest: 42 suites passed, 1 skipped; 582 tests passed, 3 skipped; 8 snapshots passed.
• Worker Vitest: 33 files and 1,433 tests passed.
• Worker TypeScript/ESLint: passed with zero errors.
• Web production build: passed.
• Tiyuvta provider pages, attributed quickstart, and live /v1/models metadata were rechecked before submission.

Sources

• Model catalog: https://api.tiyuvta.ai/v1/models
• Pricing: https://inference.tiyuvta.ai/pricing
• Qwen model card: https://huggingface.co/Qwen/Qwen3.8-27B
• Ornith model card: https://huggingface.co/ornith-ai/Ornith-1.5-35B-A3B

Screenshots / Demos

N/A — the visible change is a provider entry using the included Tiyuvta logo.

Extra Notes

The provider documentation link includes ?c=helicone-gateway so Tiyuvta can measure activation originating from this integration.

[greptile-apps]

Your trial has ended. Reactivate Greptile to resume code reviews.

[vercel]

@avifenesh is attempting to deploy a commit to the Helicone Team on Vercel.

A member of the Team first needs to authorize it.