Developer docs

Guides

Multi-page docs — not a single blurb. Each guide labels maturity as Ships Partial Not yet.

Why DragonLair vs a pure aggregator

Aggregators maximize catalog reach. DragonLair is an OpenAI-compatible gateway and a self-optimizing inference control plane: hard governance before ranking, evidence-driven Observe → Experiment → Evaluate → Promote/rollback, and explainable decisions.

• Route explanation headers on decisions (for example X-Darouter-Route-Explanation).
• model=auto resolves to a concrete model — see X-Darouter-Selected-Model.
• Governance explanations via X-Darouter-Governance-Explanation; cache-aware rank reasons via X-Darouter-Cache-Reason.

Deeper category table: /compare. Maturity: /features. No GA savings, no 500-model absolutes.

Canonical API host

Use https://api.cybertronic.dev only. Do not send X-Darouter-Edge-Secret from a public client. /admin/* and POST /v1/registration are not on the public hostname.

Use the status page to confirm current service availability before production rollout.

Authentication Ships

Tenant inference and catalog routes take Authorization: Bearer <Darouter API key> (or a configured JWT). Full guide: /docs/auth.

• Opt-in public signup — POST /v1/signup when Operator Platform settings has public_signup=true (off by default).
• Portal — sign in at /portal, then POST /v1/keys.
• Operator-issued — admin-token POST /v1/registration (not on the public host).

OpenAPI cross-links (no recipe farm)

Every guide below cites the operations that exist in /v1/openapi.json. Use OpenAI Python, OpenAI TypeScript, or curl with base_url — we do not ship a first-party SDK and we do not publish cookbook/recipe trees.

When a portal panel mentions a route not yet in OpenAPI, treat OpenAPI as the stable contract and the panel as Partial.

Quickstart

Common clients only: curl, OpenAI Python, OpenAI TypeScript. DragonLair does not ship a first-party SDK — OpenAPI is the contract. No cookbook trees.

1. List models (safe — no billed dispatch)

curl -sS https://api.cybertronic.dev/v1/models \
  -H "Authorization: Bearer $DAROUTER_API_KEY"

2. Chat completion

curl -sS https://api.cybertronic.dev/v1/chat/completions \
  -H "Authorization: Bearer $DAROUTER_API_KEY" \
  -H 'Content-Type: application/json' \
  -d '{"model":"auto","messages":[{"role":"user","content":"Hello"}]}'

3. OpenAI Python SDK

from openai import OpenAI
client = OpenAI(
    api_key="dar_…",
    base_url="https://api.cybertronic.dev/v1",
)
print(client.chat.completions.create(
    model="auto",
    messages=[{"role": "user", "content": "Hello"}],
))

4. TypeScript / Node (openai)

import OpenAI from "openai";
const client = new OpenAI({
  apiKey: process.env.DAROUTER_API_KEY,
  baseURL: "https://api.cybertronic.dev/v1",
});
const res = await client.chat.completions.create({
  model: "auto",
  messages: [{ role: "user", content: "Hello" }],
});
console.log(res.choices[0]?.message);

Full common-client guide: /docs/sdk (OpenAI Python / TypeScript / curl only).

OpenAPI & agent brief

curl -sS https://api.cybertronic.dev/v1/openapi.json
# same document on the marketing host:
#   https://dragonlayer.ai/v1/openapi.json

Also available: OpenAPI and the agent-readable documentation index.