What is the Google Cloud Inference API?

Stay organized with collections Save and categorize content based on your preferences.

The Cloud Inference API allows you to:

  • Index and load a dataset consisting of multiple data sources stored on Google Cloud Storage.
  • Execute Inference queries over loaded datasets, computing relations across matched groups (see below for data organization).
  • Unload or cancel the loading of a dataset.
  • Get simple status updates for a dataset sent for processing.


Get Started!


Brief overview of the Google Cloud Inference API

Notes on data organization in the Cloud Inference API:

  • A single data entry is called an 'event'. For example: pressure is 110 PSI.
  • Each event must have a start time and optionally an end time as well.
  • Each event has a type that is called the data name. For example: 'pressure' for pressure readings and 'temperature' for temperature readings.
  • A logical grouping of events is called a 'group'. Groups can include data of any type. For example, all the sensor readings for pressure and temperature during a day can be represented as a group with an 'ID' of absolute day number and with the data items individual readings of pressure and temperature sensors from 12:00:00am to 11:59:59pm of that day.
  • Each group is also populated with a set of events for doing time-based aggregation. This data is automatically generated from the timestamps of the provided events.
    • ty: Calendar Year
    • tew: Epoch Week (weeks since Unix epoch)
    • ted: Epoch Day (days since Unix epoch)
    • teh: Epoch Hour (hours since Unix epoch)
  • Scoring in the system works across groups. When you query the system for patterns and trends, groups are used as the unit of aggregation.

The Cloud Inference API enforces the following per-project quotas:

  • Clients can list, create, and delete datasets at a rate of up to 1 QPS
  • Clients can query at a rate of up to 10 QPS
  • Clients can ingest data items at a rate of up to 100 QPS