[AICORE-367] Register models - Nuxeo Issue Tracker

XML

Word

Printable

Details

Type: User story
Status: Resolved
Priority: Blocker
Resolution: Fixed
Affects Version/s: None
Fix Version/s: 2.4.3
Component/s: AI Cloud Services

Epic Link:
Plug and Play: Configuration
Tags:
- nxAI
Team:
AI
Sprint:
AI 3a - Plug and Play

Description

As an Insight Admin
I want to be able to select which models are registered on each Nuxeo Application.
So that I can dynamically add or remove models and optimize suggestion requests

When a Nuxeo Application synchronizes with Nuxeo Insight, Nuxeo Insight should provide the list models that are active for the Nuxeo Application.
The list should have the models that are published with the same label (datasource label = prediction label) as the Nuxeo Application.
This way, on auto enrich and form suggestions we only query Insight with content that we have models to.

Previously, we had to register a model and what prediction label (modellabel) we wanted to use. This had two problems:

if we make another model available, we need to change the addon app xml configuration
if we unpublish a registered model, when the addon side asks for suggestions/predictions, that model will still be queried and an error will yield

This configuration should allow the suggestion enrichment to be saved on the metadata.

Example:

<extension point="models" target="org.nuxeo.ai.model.serving.ModelServing">
  <model id="TotalProdTalent">
    <filter primaryType="ProductImage" />
    <info name="modelName">totalprodtalent</info>
    <config name="useLabels">false</config>
    <info name="modelLabel">${nuxeo.ai.model.label}</info>
    <inputProperties>
      <property name="file:content" type="img" />
    </inputProperties>
    <outputProperties>
      <property name="pvh_related:Talent" type="cat" />
    </outputProperties>
    <config name="minConfidence">0.60</config>
  </model>
</extension>

Example of configuration to auto-enrich and save enrichment.

<extension target="org.nuxeo.runtime.stream.service" point="streamProcessor">
 <streamProcessor name="bulkFashionBotProcessor" defaultConcurrency="2" defaultPartitions="4" logConfig="bulk" class="org.nuxeo.ai.enrichment.EnrichingStreamProcessor">
 <option name="source">bulk.enrich</option>
 <option name="sink">enrichment.fashion.bulk</option>
 <option name="enrichmentProviderName">Project</option>
 </streamProcessor>
</extension>

<extension target="org.nuxeo.runtime.stream.service" point="streamProcessor">
 <streamProcessor name="sbulkSaveMetadataProcessor" defaultConcurrency="2" defaultPartitions="4" logConfig="bulk" class="org.nuxeo.ai.functions.SaveEnrichmentFunction">
 <option name="source">enrichment.fashion.bulk</option>
 </streamProcessor>
 </extension>