First Backend

The default kit transport sends a JSON request with serialized field values under inputs.

{
  "inputs": {
    "prompt": "Example text",
    "threshold": 0.75
  }
}

Return a reports object keyed by report id. meta is optional and can carry request ids, model version, timing, or audit information.

{
  "reports": {
    "prediction": {
      "label": "Approved",
      "confidence": 0.91
    }
  },
  "meta": {
    "model": "credit-risk-v2"
  }
}

Express

import express from "express";

const app = express();
app.use(express.json());

app.post("/api/predict", (req, res) => {
  const { prompt, threshold } = req.body.inputs ?? {};

  if (!prompt) {
    res.status(400).json({ message: "Prompt is required." });
    return;
  }

  res.json({
    reports: {
      prediction: {
        label: "Approved",
        confidence: Number(threshold ?? 0.75),
      },
    },
    meta: { model: "demo" },
  });
});

FastAPI

from fastapi import FastAPI, HTTPException
from pydantic import BaseModel, Field

app = FastAPI()

class PredictRequest(BaseModel):
    values: dict = Field(alias="inputs")

@app.post("/api/predict")
def predict(request: PredictRequest):
    prompt = request.values.get("prompt")
    if not prompt:
        raise HTTPException(status_code=400, detail="Prompt is required.")

    return {
        "reports": {
            "prediction": {
                "label": "Approved",
                "confidence": request.values.get("threshold", 0.75),
            }
        },
        "meta": {"model": "demo"},
    }

For non-2xx responses, createJsonTransport reads the response body and uses a message property when present.