pyTSI

pyTSI is a read-only Python SDK for Microsoft Azure time series insights. It provides methods to conveniently retrieve your data and is designed for analysts, data scientists and developers working with time series data in Azure TSI.

The main objective of this library are to:

• Not transform the data in any way. The information will be provided to you as delivered by the TSI, gaps and all. The only exception to this is that when requesting time series you can ask for rows comprised entirely on NaN's to be removed.
• Make it easy and pythonic to obtain the information in the TSI as a pandas DataFrame.

Please note that the code is under heavy development, the API will change frequently, and the documentation is not (yet) updated. Also, please note that even though this code started as a fork of TSClient, pyTSI is not compatible at all with TSIClient.

Documentation

• Azure time series REST APIs: https://docs.microsoft.com/en-us/rest/api/time-series-insights/

Quickstart

Instantiate the TSIClient to query your TSI environment. Use the credentials from your service principal in Azure that has access to the TSI environment (you can also use environment variables to instantiate the pyTSI or provide a specific TSI API version, check the documentation).

from pyTSI import TSIClient as TSI

client = TSI.TSIClient(environment_name='<your-tsi-env-name>',
                       client_id='<your-client-id>',
                       client_secret='<your-client-secret>',
                       tenant_id='<your-tenant-id>',
                       application_name='<your-app-name>')

# List the instances in the TSI, also list their types
# and variables.
for instance in client.time_series():
    print(f'\t{instance}')
    print('\tInstance type description:')
    print(f'\t\t{instance.series_type}')
    print('\t\tType vars:')
    for v in instance.series_type.vars:
        print(f'\t\t\t{v}')

You can now query each instance

You can query your timeseries data by timeseries id, timeseries name or timeseries description. The Microsoft TSI apis support aggregation, so you can specify a sampling freqency and an aggregation method. Refer to the documentation for detailed information.

import datetime

# Define the start & end times for the series that we want to retrieve
t0 = datetime.datetime(year=2020, month=10, day=22, hour=10, minute=53, second=00,
                       tzinfo=datetime.timezone.utc)
t1 = datetime.datetime(year=2020, month=10, day=22, hour=11, minute=53, second=30,
                       tzinfo=datetime.timezone.utc)

# Select a time series by selecting from the list of by filtering by ID
ts = client.time_series.get_by_id(['Time series ID'])
t = ts.series_type
# Get raw event data
raw_events = ts.get_events(start=t0, end=t1, 
                           variables=['temperature', 'humidity'],
                           types=['Double', 'Double'],
                           drop_nans=True)
# Get series data for raw events & composed variables
series_data = ts.get_series(start=t0, end=t1, 
                            variables=[t.temperature, t.humidity, t.series_sum],
                            drop_nans=True)
# Aggregate series
aggregated_data = ts.aggregate_series(start=t0, end=t1, interval='PT1M', 
                                      variables=[t.temperature, t.EventCount],
                                      drop_nans=True)

Each of these functions return a DataFrame, with variable names as columns.

License

pyTSI is licensed under the MIT license. See LICENSE file for details.