API docs

POST /v1/answer

Send only a prompt. ElseLane classifies, routes, and fails over.

curl https://api.elselane.com/v1/answer \
  -H "Authorization: Bearer el_live_..." \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: optional-unique-id" \
  -d '{"prompt":"Explain CAP theorem simply"}'

Optional body fields: system (string) and response_format ({ "type": "json_object" }).

{
  "id": "...",
  "answer": "...",
  "task_type": "chat",
  "model_used": "gemini-3.6-flash",
  "provider": "google",
  "fallback_used": false,
  "attempts": [
    { "provider": "google", "model": "gemini-3.6-flash" }
  ],
  "usage": {
    "input_tokens": 12,
    "output_tokens": 40,
    "credits_charged": 10
  }
}

attempts lists each provider tried. Failed steps include error (e.g. quota, circuit open, provider disabled, invalid_json_response).

JSON structured output

Set response_format: { "type": "json_object" } on either endpoint. ElseLane only routes to providers with native JSON mode (OpenAI, Google, Groq, Together — Anthropic steps are skipped), passes the provider constraint, and validates that the answer parses as a JSON object. Invalid JSON fails that attempt and fails over. Streaming is not supported with JSON mode in v1.

curl https://api.elselane.com/v1/chat/completions \
  -H "Authorization: Bearer el_live_..." \
  -H "Content-Type: application/json" \
  -d '{
    "model": "auto",
    "response_format": { "type": "json_object" },
    "messages": [
      { "role": "system", "content": "Return JSON with keys title and summary." },
      { "role": "user", "content": "Write a short blog brief about failover APIs." }
    ]
  }'

Which model answered

Top-level model on chat completions stays "auto". Read the actual model from the body or headers:

• /answer → model_used, provider, attempts
• /chat/completions → elselane.model_used, elselane.provider
• Headers on success: X-ElseLane-Model, X-ElseLane-Provider, X-ElseLane-Fallback, X-ElseLane-Task-Type

Deno / Edge (fetch)

Works in Supabase Edge Functions and other Deno runtimes — no Node SDK required.

const res = await fetch("https://api.elselane.com/v1/chat/completions", {
  method: "POST",
  headers: {
    Authorization: `Bearer ${Deno.env.get("ELSELANE_API_KEY")}`,
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    model: "auto",
    response_format: { type: "json_object" },
    messages: [
      { role: "system", content: "Return JSON with keys title and summary." },
      { role: "user", content: "Brief on multi-provider failover." },
    ],
  }),
});

if (!res.ok) throw new Error(await res.text());

const model = res.headers.get("X-ElseLane-Model");
const data = await res.json();
const content = data.choices[0].message.content; // JSON string
const parsed = JSON.parse(content);

POST /v1/chat/completions

OpenAI-compatible shim. Use model: "auto" only. ElseLane metadata is under elselane in the response. Set stream: true for SSE (text/event-stream) chunks ending with data: [DONE]; final chunk includes usage and elselane. Do not combine stream: true with JSON mode.

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

curl --no-buffer https://api.elselane.com/v1/chat/completions \
  -H "Authorization: Bearer el_live_..." \
  -H "Content-Type: application/json" \
  -d '{
    "model": "auto",
    "stream": true,
    "messages": [{"role":"user","content":"Hello"}]
  }'