DarkSeek API

A JSON-over-HTTPS API for the DarkSeek inference service, with bounded streaming and a chat-completions-compatible interface.

Authentication

Every model and generation request requires a customer API key. Keep it in a server-side environment variable. Do not put keys in query strings, browser bundles, shared screenshots, or analytics.

Base URL

Use the origin of a deployment with its production Go API binding connected, followed by /v1. A separate API hostname can point to the same Go ingress once DNS and TLS are configured. No public API origin is declared ready yet.

API version & compatibility

The currently verified contract is small text chat, streaming, model enumeration, authentication, admission and cancellation through the isolated Go gateway. Full DeepSeek protocol parity is a separate acceptance gate, including tool lifecycle, reasoning controls, Responses, Messages, FIM, files, search/fetch, caching metadata and balance semantics.

Errors and request identity

Errors use a JSON error object with type, code and message. Request IDs are returned in X-Request-ID. Once an SSE stream begins, failures are explicit error events or incomplete streams, never a fabricated successful completion.

Customer accounts & operations

The platform manages account-owned keys and a shared durable admission allowance through Better Auth and D1. Verified email and an explicitly enabled allowance are required for inference. Public registration, verification/recovery delivery, the public inference connection, credits, billing and production support remain launch gates.

Service identity

DarkSeek is independently operated. DeepSeek API keys, account balances, pricing, legal registrations and support addresses do not transfer to DarkSeek.