Introduction

This is the reference documentation for the Everything API with an overview of all the available methods.

Everything is the Platform from Mira which provides generic computer vision capabilities, enabling 3rd party applications to easily and cost effectively detect and recognize concepts from images without the need of large datasets, manual annotations or extensive trainings.

Foundational Pillars

Before using the platform, it’s important to understand the following foundational pillars:

Taxonomy: A structured definition of concepts, including their relationships and hierarchical organization.
Asset: A container for resources within the platform, such as images or image crops.
Annotation: Metadata associated with assets, used to enrich them with information, such as detected instances in an image and their corresponding concepts.

Everything Platform annotates assets automatically, associating them directly to concepts defined in the taxonomy.

Further details about these pillars can be found in the subsequent sections.

Get started with the API

Mira’s RESTful web API enables you to access resources and functionalities in Everything.

A valid API token is needed to perform API requests. Tokens can be created from the Device and API management section of each Workspace in your tenant.
The API generation wizard enables to define specific grants to each token.

Once the access token is created within the context of a workspace, it can be used to access or perform actions only within that specific workspace.

HTTP methods

Everything API is HTTP-based; It accepts JSON and responds with JSON.

Method	Description
GET	Read data from a resource
POST	Create a resource, or perform an action
DELETE	Remove a resource
PUT	Replace a resource with a new one

HTTP Status Codes

Everything API uses conventional HTTP response codes to indicate the success or failure of an API request. In general, codes in the 2xx range indicate success, codes in the 4xx range indicate an error caused by the information provided (i.e., a required parameter was omitted, a method was not found, etc.), and codes in the 5xx range indicate an error with servers.

Pagination

When dealing with large datasets in the platform, the API supports pagination to control the amount of data returned in a single response.

Pagination enables efficient querying by dividing results into pages and providing mechanisms to retrieve subsequent or previous subset of data.

The following query parameters are used to control pagination:

limit (integer): Specifies the maximum number of items to return
offset (integer): Specifies the starting point (index) from which to retrieve the items.

The API returns an empty dataset if the offset overflows the size of the dataset.

GET api/taxonomy/v1/taxonomy?limit=100&offset=0

Common annotation workflow

Everything Platform is designed to detect and recognize instances of generic items in visual inputs by assigning concepts to them.

The platform provides a range of functionalities for building and executing inference workflows, enabling various business cases that require extracting structured concepts from visual data.

Below is the recommended standard workflow for performing an end-to-end inference on a given image:

Define Taxonomy: A predefined taxonomy is created to specify the concepts that will be detected and recognized in the visual input.
Image Encoding: The input image, such as a JPG file, is serialized into a Base64-encoded byte string.
Data Upload: The encoded image data is uploaded to the Everything Platform as a newly created element.
Asset Creation: An asset encapsulating the uploaded element is created.
Automatic Annotation: The Everything Platform automatically generates annotations for the asset, based on the taxonomy.