Laya AI. Try your next workflow online.

1. Create a key and connect

Sign in, open API Keys in your account settings and create a key. Keep enough credits in your account, then send a POST request to the URL below. The API path has no language prefix, even when this page is translated.

Use your own application server to make API calls. Store the key in a server environment variable named LAYA_API_KEY; never put it in browser code or a public repository. The playground uses your signed-in session instead of asking for a key.

2. Choose a model and define the request

Set model to english, multilingual or typed-decisions. All three options accept choice, score and noul questions.

3. Send your first request

Set LAYA_API_KEY to the key you created. This example evaluates one message with all three question types. Run JavaScript on Node.js 18+ (as an .mjs file), or use Python 3 with its built-in urllib module; no Python package installation is needed.

export LAYA_API_KEY="YOUR_API_KEY"

curl --fail-with-body --request POST 'https://thejevai.com/laya/v1/systemone' \
  --header "Authorization: Bearer $LAYA_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "model": "english",
  "state": "I was charged twice for my subscription. Please refund the extra payment today.",
  "questions": {
    "department": {
      "type": "choice",
      "instructions": "Which team should handle this request?",
      "criteria": {
        "billing": "Payments and refunds",
        "technical": "Bugs and outages"
      }
    },
    "priority": {
      "type": "score",
      "instructions": "How urgent is this request?",
      "criteria": [
        "Low",
        "Medium",
        "High"
      ]
    },
    "refund": {
      "type": "noul",
      "instructions": "Is the customer requesting a refund?",
      "criteria": {
        "true": "A refund is requested",
        "false": "No refund is requested"
      }
    }
  }
}'

4. Read answers, usage and credits

Check both the HTTP status and code === 0 before using the result. Success returns answers inside data.result; it does not return a chat completion or a streaming text response.

Illustrative response: decisions, token counts and latency vary with the request.

{
  "code": 0,
  "message": "ok",
  "data": {
    "result": {
      "model": "english",
      "answers": {
        "department": {
          "type": "choice",
          "choice": "billing",
          "probabilities": {
            "billing": 0.97,
            "technical": 0.03
          }
        },
        "priority": {
          "type": "score",
          "score": 1.8,
          "legend": {
            "0": "Low",
            "1": "Medium",
            "2": "High"
          }
        },
        "refund": {
          "type": "noul",
          "noul": 0.98
        }
      },
      "usage": {
        "input_tokens": 1200,
        "output_tokens": 0
      },
      "elapsedMs": 142
    },
    "creditsUsed": 2
  }
}

data.result.answers

Look up each result by the question ID you sent. Read choice, score or noul according to its type; additional probability fields may be present.

data.result.usage

input_tokens and output_tokens report measured token usage. See data.creditsUsed for the credits consumed.

data.result.elapsedMs

Time in milliseconds spent handling the request. This is an observation, not a latency guarantee.

data.creditsUsed

Credits deducted for this request. Use this field to show the actual charge in your application.

Use the request and response format documented here when integrating this endpoint. If you switch to another Laya service, check its authentication and response structure and update your client as needed.

5. Plan for usage limits and failures

400

Invalid JSON, model, state, question type or criteria. Correct the request before retrying.

401

Missing or invalid authentication. Check your Bearer API key or sign in to the playground.

402

Insufficient credits. Add credits and retry with enough balance for the request.

413

Request body exceeds 32 KiB. Shorten the context or split the request.

429

Requests are too close together. Respect the Retry-After header before retrying.

502

Inference failed, timed out or returned an invalid result. Check the error message and retry only when appropriate.

503

The service is unavailable. Try later or contact support.

Read the HTTP status and message on errors. This endpoint has no idempotency key: blindly retrying after a client timeout may create another charge if the first request completed. Confirm the outcome before retrying paid requests.