Forecast any time series
in one API call

Three steps to your first forecast

No ML background needed. No model training. No infrastructure to provision.

Three models. One endpoint.

Pass "model": "auto" to let the API select, or specify one explicitly. The API automatically falls back to ARIMA if the GPU backend is unavailable.

Automatic fallback: If the GPU backend (Modal.com) is unreachable, the API silently falls back to ARIMA and sets fallback_used: true in the response. Your request always completes.

Real numbers. Published results.

Rolling-window backtesting — 5 windows per dataset. Source code and raw JSON results are available in the GitHub repository. No cherry-picked metrics.

ETT-h1: ETDataset — Exchange Rate: Time-Series-Library — M5-sample: synthetic retail demand with M5-style weekly seasonality. Methodology: 5-window rolling backtesting, evaluated on held-out test windows.

Why TSFA over alternatives

Honest comparison. TSFA is not the right tool for every use case — but for univariate forecasting with a REST API, it's the simplest and cheapest path.

Honest caveat: TSFA does not yet support multivariate forecasting with covariates (Phase 2). If you need exogenous variables today, Nixtla or AWS Forecast are better fits.

Built for real forecasting problems

Three domains where TSFA is production-ready today.

Endpoints

Base URL: https://api.eymdey-network.com/v1 — all requests require X-RapidAPI-Key header.

Integrate in minutes

Copy-paste examples for the three most common use cases. Replace YOUR_KEY with your RapidAPI key.

import requests

# Install: pip install requests
url = "https://api.eymdey-network.com/v1/forecast/univariate"
headers = {
    "X-RapidAPI-Key": "YOUR_KEY",
    "Content-Type": "application/json",
}
payload = {
    "series": [120, 132, 128, 145, 139, 152, 148, 160, 155, 168, 163, 175],
    "horizon": 7,
    "frequency": "D",
    "model": "auto",
    "confidence_levels": [0.80, 0.95],
}

resp = requests.post(url, json=payload, headers=headers)
resp.raise_for_status()
data = resp.json()

print(data["model_used"])             # "arima" or "chronos"
print(data["forecast"]["mean"])      # [176.3, 179.1, ...]
print(data["forecast"]["lower_95"]) # 95% lower bound
print(data["diagnostics"]["trend"])  # "upward"
print(data["meta"]["credits_used"]) # 1

// Node.js or browser fetch
const resp = await fetch("https://api.eymdey-network.com/v1/forecast/univariate", {
  method: "POST",
  headers: {
    "X-RapidAPI-Key": "YOUR_KEY",
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    series: [120, 132, 128, 145, 139, 152, 148, 160, 155, 168, 163, 175],
    horizon: 7,
    frequency: "D",
    model: "auto",
    confidence_levels: [0.80, 0.95],
  }),
});

if (!resp.ok) throw new Error(`HTTP ${resp.status}`);
const data = await resp.json();

console.log(data.model_used);            // "arima"
console.log(data.forecast.mean);         // [176.3, ...]
console.log(data.forecast.lower_95);     // 95% lower bound
console.log(data.diagnostics.trend);     // "upward"
console.log(data.meta.credits_used);     // 1

# Basic forecast
curl -X POST "https://api.eymdey-network.com/v1/forecast/univariate" \
  -H "X-RapidAPI-Key: YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "series": [120, 132, 128, 145, 139, 152, 148, 160, 155, 168, 163, 175],
    "horizon": 7,
    "frequency": "D",
    "model": "auto",
    "confidence_levels": [0.80, 0.95]
  }'

# Check your credit usage
curl "https://api.eymdey-network.com/v1/usage" \
  -H "X-RapidAPI-Key: YOUR_KEY"

# List available models
curl "https://api.eymdey-network.com/v1/models" \
  -H "X-RapidAPI-Key: YOUR_KEY"

Response structure

{
  "status": "success",
  "model_used": "arima",                       // model that actually ran
  "forecast": {
    "timestamps": ["2024-01-13", "2024-01-14", "..."],
    "mean":      [178.2, 181.5, 184.8, "..."],   // point forecast
    "lower_80":  [171.0, 173.5, 176.0, "..."],   // 80% lower bound
    "upper_80":  [185.4, 189.5, 193.6, "..."],   // 80% upper bound
    "lower_95":  [164.5, 166.8, 169.2, "..."],   // 95% lower bound
    "upper_95":  [191.9, 196.2, 200.4, "..."]    // 95% upper bound
  },
  "diagnostics": {
    "trend": "upward",                          // "upward" | "downward" | "stable"
    "seasonality_detected": false,
    "seasonality_period": null,
    "series_length": 12,
    "missing_values": 0,
    "stationarity": "non_stationary"           // ADF test result
  },
  "meta": {
    "inference_time_ms": 234.0,
    "request_id": "req_abc123",               // use in bug reports
    "credits_used": 1
  }
}

Simple, volume-based plans

All plans include confidence intervals, diagnostics, and backtesting. Upgrade or cancel anytime via RapidAPI.

Error codes

All errors follow the same JSON structure: {"status":"error","code":"...","message":"...","details":null}

Frequently asked questions

Answers to the most common questions.