Skip to main content

Why Pixyle.ai and Geins?

Skip tedious manual work and automate the process of tagging products with meta data from product images in Geins with AI technology by Pixyle.

Try the demo from Pixyle here: https://www.pixyle.ai/demo to see how it works on your product catalog.

Launchpad for building a Pixyle.ai integration

This Pixyle.ai integration launchpad is built on Azure Functions. The launchpad is a starting point for building Pixyle.ai integrations. The launchpad is an fork of the Geins Integration Launchpad

Repository​

The source code for this launchpad is available on GitHub pixyle-ai-geins-integration-serverless

Pre-requisites​

Features​

Auto tag products with meta data from product images in Geins with AI technology by Pixyle.

"Replace manual, repetitive and costly labeling with our state-of-the art visual AI than can auto-tag and process your entire catalogue in minutes instead of hours." - Pixyle

Suggested usecase​

Use this to automatically tag your products with meta data from product images. For example, you can use this to tag your products with color, pattern, material, style, etc. From this you can create filters and search for your products. This will most likely increase your conversion rate and customer satisfaction.

Getting started​

Run the following command to install the dependencies:

npm install

Add the local.settings.json file to the root of the project. The local.settings.json file is used to store the credentials for the Azure Storage Account and Geins Management API. The local.settings.json file is not checked in to the repository.

{
  "IsEncrypted": false,
  "Values": {
    "ENVIRONMENT" : "development",
    "FUNCTIONS_WORKER_RUNTIME": "node",
    "AzureWebJobsStorage": "my-storage-connection-string",

    "AZURE_ACCOUNT_NAME": "my-account-name",
    "AZURE_ACCOUNT_KEY": "my-key",
    "AZURE_TABLE_NAME": "log",
    "AZURE_QUEUE_NAME": "queue-items",

    "GEINS_MGMT_API_KEY": "my-key",
    "GEINS_MGMT_API_USERMAME": "my-username",
    "GEINS_MGMT_API_PASSWORD": "my-pwd",

    "PIXYLE_USERMAME": "my-username",
    "PIXYLE_PASSWORD": "my-pwd",
  },
  "Host": {
    "CORS": "*"
  }
}

Run the following command to start the function app:

func start

Authentication​

@azure/core-auth is used for authentication. You can find more information here: https://www.npmjs.com/package/@azure/core-auth

  • For the table storage and queue storage, you can use the connection string or credentials.
  • For the Geins API, you use api-user credentialas and api-key. Read more at docs.geins.io.

Update the local.settings.json file with your credentials:

{
  "IsEncrypted": false,
  "Values": {
    "ENVIRONMENT" : "development",
    "FUNCTIONS_WORKER_RUNTIME": "node",
    "AzureWebJobsStorage": "my-storage-connection-string",

    "AZURE_ACCOUNT_NAME": "my-account-name",
    "AZURE_ACCOUNT_KEY": "my-key",
    "AZURE_TABLE_NAME": "log",
    "AZURE_QUEUE_NAME": "queue-items",

    "GEINS_MGMT_API_KEY": "my-key",
    "GEINS_MGMT_API_USERMAME": "my-username",
    "GEINS_MGMT_API_PASSWORD": "my-pwd",

    "PIXYLE_USERMAME": "my-username",
    "PIXYLE_PASSWORD": "my-pwd",
  }
}

How it works​

The serverless function app is triggered by a message in the queue. The message contains the product id and the image url. The function app then downloads the image from the url and sends it to Pixyle for processing. Pixyle returns the meta data for the image. The function app then updates the product in Geins with the meta data.

Parsers​

PixyleParser​

This is a wrapper around the Pixyle API. It handles authentication and parsing of the response. The parser returns a list of PixyleAttributes.

createDataset()​

Creates an dataset out of the product images and stores the dataset id in the process table.

processDataset()​

Gets the dataset id from the processTable and starts the processing of the dataset if datset is uploaded and ready for processing.

resultDataset()​

Gets the dataset id from the processTable and gets the result of the processing. The result is stored in the processTable.

GeinsParser​

This is a wrapper around the Geins Management API. It handles authentication and parsing of the response. The parser returns a list of GeinsAttributes.

productSave()​

Saves the result of the processing to the product in Geins.

parameterBroker()​

Gets the parmeterId from the brokerTable based on the pixyle attribute.

Broker table​

The broker table is used to translate pixyle attributes into your own attributes. For example, if Pixyle returns the attribute color and you want to map it to a parmeter of choice in Geins. Application automaticly creates an azure stroage table called brokerTable.

The table has the following structure:

  • PartitionKey: pixyle attribute
  • RowKey: pixyle attribute
  • paramterId: geins parameter id

Process table​

The table has the following structure:

  • PartitionKey: pixyle datasetid
  • RowKey: geins product id
  • status: status of the process can be creating, processing, saving, saved
  • data: json data from pixyle

Flow​

  1. Create dataset in pixyle - from webhook to httpTrigger to queueTrigger
  2. Process dataset in pixyle - form timerTrigger to queueTrigger
  3. Get result of dataset from pixyle - from timerTrigger to queueTrigger
  4. Update product in geins with result - from timerTrigger to queueTrigger

Setting up Pixyle tags with Geins Product Meta Data​

Create a new parameter/properties group in geins or use an existing one. Add the parameters/properties you want to use in the group. Add the id of the parameter/properties to the brokerTable with the pixyle attribute as PartitionKey and RowKey.

All returned pixyle attributes will be added to the brokerTable. You can then map the pixyle attributes to your geins parmeterId in the broker table.

Refrences​