mgenrique/prophet-api repository overview

Container with python:3.11-slim for arm that implements an api to generate predictions with Prophet. Here’s a possible description for your Docker Hub repository:

⁠Repository Description

Prophet API Add-on for Home Assistant
This Docker container provides a lightweight REST API based on Flask for time series forecasting using Meta's (formerly Facebook) Prophet model. It is specifically designed to integrate as an add-on in Home Assistant, enabling easy forecasting based on historical data.

⁠Features

• REST API: Receives JSON-formatted data and returns predictions in JSON.
• Prophet Model: Utilizes Prophet, a robust and accurate time series model, ideal for trend and seasonality-based data.
• ISO Date Format: Returns dates in ISO format to ensure compatibility.
• ARM Compatibility: Optimized for ARM architectures, ideal for devices like the Raspberry Pi.

⁠Usage

1. Installation: Pull the container from Docker Hub or clone the repository and build the image.
2. Requests: Send data in JSON format to receive forecasts.
3. Integration with Home Assistant: Ideal for generating real-time predictions within Home Assistant.

⁠Example

Send a POST request to the API with date and value data in the following format:

{
  "data": [
    {"ds": "2024-01-01", "y": 10},
    {"ds": "2024-01-02", "y": 15},
    {"ds": "2024-01-03", "y": 13}
  ]
}

The API will return a forecast of future values.

⁠Notes

This container is ideal for home automation and applications that require local time series forecasting, especially on ARM architectures.

Example request:

import requests
import json

# API URL
url = "http://<container IP>:5000/forecast"

# sample data (dates and values)
data = {
    "data": [
        {"ds": "2024-01-01", "y": 10},
        {"ds": "2024-01-02", "y": 15},
        {"ds": "2024-01-03", "y": 13},
        {"ds": "2024-01-04", "y": 18}
    ]
}

# Hacer la solicitud POST
response = requests.post(url, json=data)

# Verificar la respuesta
if response.status_code == 200:
    forecast = response.json()
    print("Forecast:")
    print(json.dumps(forecast, indent=2))
else:
    print("Error:", response.status_code)
    print(response.text)

You get a response with forecast for 30 periods