Relayby decks.ai

Docs

Build with Relay.

Developer-first documentation for the unified LLM gateway.

Make your first request in under a minute.

Bearer tokens, key scopes, BYOK.

Chat completions

/v1/chat/completions reference.

Routing modes, rules, and policies.

Reliability chains and retries.

Provisioning, scoping, rotation.

Per-key and per-team spend caps.

Request, response, and trace IDs.

Spend, error, and routing events.

JS, Python, Go — OpenAI-compatible.

HTTP and provider error reference.

Quickstart

Make your first request

1Create an API key
2Set base URL to https://api.decks.ai/v1
3Choose a model (or use auto/best)
4Send a chat completion
5View logs in your dashboard

Base URL

https://api.decks.ai/v1

Authentication

Authorization: Bearer RELAY_API_KEY

Endpoint

POST /chat/completions

curl

curl https://api.decks.ai/v1/chat/completions \
  -H "Authorization: Bearer $RELAY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "auto/best",
    "messages": [{ "role": "user", "content": "Hello, world" }]
  }'

Response

json

{
  "id": "cmpl_8wYz...",
  "model": "openai/gpt-4.1",
  "choices": [{ "message": { "role": "assistant", "content": "Hello!" } }],
  "usage": { "prompt_tokens": 12, "completion_tokens": 6, "total_tokens": 18 },
  "relay": { "routed_to": "openai/gpt-4.1", "fallback_used": false, "latency_ms": 412 }
}