Installation & Integration

Virtual Try On is an AI-based project that allows the customer to try on the relevant product on a photo they have uploaded. The flow is completed through 3 core applications:

  • Virtual Try On Extension

  • Frontend (Next or Python)

  • Mobile App


1. Installation

Extension

Variable
Value / Description

DB_SSL

true

DB_SCHEMA

.public

APP_TEST_MODE

false

APP_SERVER_PORT

:8008

SEC_API_AUTH_PASSWORD

20-character combination of uppercase, lowercase letters and numbers (must be entered as a secret)

API_AUTH_USERNAME

15-character combination of uppercase, lowercase letters and numbers

PROCESS_MONITORING

false

GOOGLE_OAUTH_ADDRESS

https://oauth2.googleapis.com/token

APP_LIMITED_CATEGORIES

Comma-separated FE category IDs for categories where try-on should be disabled

APP_SERVER_ACCESS_LOGS

true

VERTEX_AI_IMAGE_FORMAT

base64

INF_TRY_ON_RATE_LIMITER

250-M

API_ALLOWED_IMAGE_DOMAINS

Root domain names of the CDNs where product images are hosted (e.g. akinoncloud.com)

DB_CONNECTION_POOL_MAX_SIZE

20

DB_CONNECTION_POOL_MIN_SIZE

10

TOGGLE_IMAGE_GENERATION_ENABLED

true

S3_PRESIGN_EXPIRE_MINUTES

30

S3_PRESIGN_LISTING_EXPIRE_MINUTES

300

FE-Next

(To be filled)

FE-Python

(To be filled)

Mobile App

(To be filled)


2. Observation Mode Controls

Before the application is fully released to end customers, controls must be performed in observation mode. In a deployment made in this mode, end customers cannot directly see the new features. However, the features become accessible when enable_virtual_try_on=true is added to the browser's local storage.

Comprehensive controls must be performed including:

  • Query speed for both direct queries using the product's existing image and queries made with cropping

  • Similar product checks

  • Error checks

The product must NEVER be released directly to the end customer without these controls being completed.


3. Release to End Customer

Once the controls are completed, the FE applications must make the feature available to end customers either by deploying a new release or by making a parameter change.

Last updated

Was this helpful?