Getting started with function-backed time series（函数支持的时间序列入门指南）

This guide will help you get started with creating a function-backed time series that can be used in Quiver analyses to enable real-time forecasting workflows.

Prerequisites

To set up function-backed time series, you need the following resources:

• A Foundry function that returns a numeric time series using the time series SDK.
• A tagged version of that function to reference in Quiver.

Set up a function-backed time series in Quiver

1. Install the time series SDK: In Code Repositories, add the timeseries-sdk library to your code repository using the Libraries tab.
   
   
   
   

2. Write a Python function: Create a function that returns a serialized pandas DataFrame with the following columns:

3. timestamp: A pandas timestamp.

4. value: A numeric value, for example int32 or float.

The time series SDK provides a library for serializing a pandas DataFrame as shown in the example below.

from functions.api import function
from timeseries_sdk.types import TimeSeries
from ontology_sdk.ontology.objects import Machine
import pandas as pd

@function
def performance_prophet_forecast(
    machine: Machine,
    periods: int = 96,      # number of future steps
    freq: str = "15min",    # sampling frequency
) -> list[bytes]:
    """
    Forecast a machine's performance_score using Prophet.
    """
    from prophet import Prophet

    # Retrieve performance score data (columns: timestamp, value)
    df = machine.performance_score.to_pandas(all_time=True)

    if df.empty:
        return TimeSeries.serialize(pd.DataFrame(columns=["timestamp", "value"]))

    # Prepare the data for Prophet
    df = df.rename(columns={"timestamp": "ds", "value": "y"}).sort_values("ds")
    model = Prophet().fit(df)

    # Create future data frame for forecast
    future = model.make_future_dataframe(periods=periods, freq=freq, include_history=True)
    forecast = model.predict(future)

    # Format the forecast output
    out = forecast[["ds", "yhat"]].rename(columns={"ds": "timestamp", "yhat": "value"})
    return TimeSeries.serialize(out)

1. Publish the function: Publish your function and tag a version.

2. Use the function in Quiver:

3. Add a Code function time series card.
   
   
   
   
4. Select your function, choose the correct version, and configure its inputs.
   
   
   
   
5. Optionally, add another instance of the card to compare multiple scenarios side-by-side.

---

中文翻译

---

函数支持的时间序列入门指南

本指南将帮助您创建函数支持的时间序列（function-backed time series），该序列可用于 Quiver 分析，实现实时预测工作流。

前提条件

要设置函数支持的时间序列，您需要以下资源：

• 一个使用时间序列 SDK（time series SDK）返回数值型时间序列的 Foundry 函数（function）。
• 该函数的已标记版本（tagged version），用于在 Quiver 中引用。

在 Quiver 中设置函数支持的时间序列

1. 安装时间序列 SDK： 在代码仓库（Code Repositories）中，使用 Libraries 选项卡将 timeseries-sdk 库添加到您的代码仓库。
   
   
   
   

2. 编写 Python 函数： 创建一个函数，返回包含以下列的序列化 pandas DataFrame：

3. timestamp： 一个 pandas 时间戳。

4. value： 一个数值，例如 int32 或 float。

时间序列 SDK 提供了一个用于序列化 pandas DataFrame 的库，如下例所示。

from functions.api import function
from timeseries_sdk.types import TimeSeries
from ontology_sdk.ontology.objects import Machine
import pandas as pd

@function
def performance_prophet_forecast(
    machine: Machine,
    periods: int = 96,      # 未来步数
    freq: str = "15min",    # 采样频率
) -> list[bytes]:
    """
    使用 Prophet 预测机器的 performance_score。
    """
    from prophet import Prophet

    # 获取性能评分数据（列：timestamp, value）
    df = machine.performance_score.to_pandas(all_time=True)

    if df.empty:
        return TimeSeries.serialize(pd.DataFrame(columns=["timestamp", "value"]))

    # 为 Prophet 准备数据
    df = df.rename(columns={"timestamp": "ds", "value": "y"}).sort_values("ds")
    model = Prophet().fit(df)

    # 创建用于预测的未来数据框
    future = model.make_future_dataframe(periods=periods, freq=freq, include_history=True)
    forecast = model.predict(future)

    # 格式化预测输出
    out = forecast[["ds", "yhat"]].rename(columns={"ds": "timestamp", "yhat": "value"})
    return TimeSeries.serialize(out)

1. 发布函数： 发布您的函数并标记一个版本。

2. 在 Quiver 中使用函数：

3. 添加一个 代码函数时间序列卡片（Code function time series card）。
   
   
   
   
4. 选择您的函数，选择正确的版本，并配置其输入。
   
   
   
   
5. （可选）添加另一个卡片实例，以并排比较多个场景。