Open navigation

External Analytics Deployment

Developers can deploy their own external analytics in platform using the Command Line Interface. The integration of custom models enables to use the platform ressources and features for visual and non-visual data management, adding external analytics to end-to-end workflows.

This tutorial shows how to configure an external analytic description and deploy the analytic.

1. Analytic Components

The analytic is mainly defined by :

  1. A name
  2. An algorithm
  3. Inputs
  4. Parameters
  5. Outputs
  6. Computation requirements

2. Analytic Description File

The analytic description file is a single YAML file. Here is an example :

 spec-version: v0.1
 name: eiffel-corp/rust-detector
 display-name: Rust detector
 description: Detects rust on historical buildings
 instance-type: small
volume-size: 50
 groups: [Maintenance]
  - name: building_mesh
    display-name: Building 3D Model
    description: 3D Model of the model (mesh dataset)
    kind: dataset
      type: mesh
    required: true
  - name: detection-mode
    display-name: Rust detection mode
    description: Level of detection (Fast, Standard or Deep)
      type: string
        - Fast
        - Standard
        - Deep
    required: true
  - name: rust-report
    display-name: Rust detection report
    description: Pdf report containing insights about rust detected in the 3D model
    kind: dataset
      type: file
    required: true

3. Model Deployment

Deployment is achieved in three steps :

Step 1 - Package your model in a Docker image

Step 2 - Push it to your Docker registry

Step 3 - Create the analytic with the Command Line Interface

4. Analytic Availability in Platform

Step 1 - Go to the Analytics section from the left bar to open the catalog :

Step 2 - Open the newly created section (in our example below called BETA) and select your newly integrated analytic (Geo Scatter Plot in example) :

Step 3 - Run your analytic :

Learn here about SDK. 

Did you find it helpful? Yes No

Send feedback
Sorry we couldn't be helpful. Help us improve this article with your feedback.