Gateway

Unified Inference API

One OpenAI-compatible endpoint for every supported model and provider

Send chat, completion, and embedding requests through a single OpenAI-compatible endpoint. Switch between providers without rewriting client code.

Drop-in OpenAI compatibilitySingle integration surfaceVendor decoupling

기능 스냅샷

1 API

OpenAI-compatible endpoint

multi

model providers

unified

request schema

실제 프로덕션 워크플로에 매핑된 주요 가치 포인트.

이점

Drop-in OpenAI compatibility

Reuse the existing OpenAI SDK, libraries, and tooling. Only the base URL and API key change.

이점

Single integration surface

One contract for chat, streaming, function calling, and embeddings across every supported provider.

이점

Vendor decoupling

Move workloads between providers from configuration alone. No code change required.

게이트웨이 호출, 라우팅 정책, 과금 제어에 이 기능을 통합하는 3단계 경로.

단계 1

Create an API key

Issue a key under your account or organization with scoped permissions.

단계 2

Point your SDK to the gateway

Set the base URL to the DeepToken endpoint and continue using OpenAI-compatible request shapes.

단계 3

Select a model

Pass any supported model id and let the gateway resolve provider, region, and quota.

Is the API fully OpenAI compatible?

Yes. Standard chat, completion, embedding, and streaming endpoints follow the OpenAI request and response schema.

Can I call multiple models from one key?

Yes. One API key can route to every model your plan and organization permits.

다음 단계

문서, 모델 카탈로그 또는 대시보드에서 시작하여 실제 트래픽으로 이 기능을 검증하세요.