Skip to main content

Enrichment Service

Trustana's Enrichment Service is the core AI-powered engine that transforms basic product information into rich, channel-ready content. This service leverages advanced machine learning to understand product context and generate high-quality, relevant content when manually triggered through the Portal interface.

Portal-Only Operations

Enrichment tasks are initiated and managed exclusively through the Trustana Portal. API access does not include enrichment triggering, status monitoring, or task management capabilities.

How Enrichment Works​

Data Enrichment​

Data Enrichment comprises two primary input sources that feed into the generative engine:

First Party Sources (User-Provided)​

Any user-edited or user-added attribute value will be used as source material for enrichment. The system leverages all available user-provided context along with attribute configuration and prompts to regenerate, combine, and format content.

Key Components:

  • User Attribute Data: Any user-edited or user-added attribute values serve as source material for enrichment
  • User Images: Product images may contain important details in nutritional tables or specifications, which are extracted using AI vision
  • Attribute Configuration: User-added prompts on how the output should look, following good prompt engineering practices
  • Content Processing: Transform and combine user data into optimized attribute values

Generative Engine​

The AI engine processes first-party data and external sources to create enriched product content:

  • Content Regeneration: Transform existing data into optimized formats
  • Content Combination: Merge multiple data sources into cohesive product information
  • SEO Optimization: Generate search-friendly titles and descriptions
  • Templated Generation: Generate descriptions following HTML section requirements defined in prompts
  • Prompt-Driven Output: Follow user-defined prompts and formatting instructions

External Data Sourcing (Optional)​

External Data Sourcing searches the public web for the product to be enriched, sourcing contextual information about the product to be used as source material for the generative engine.

Web Sources

  • Web Search: Country-specific search with domain inclusion and exclusion controls
  • Search Results: Any publicly accessible web content found through search algorithms
  • Geographic Targeting: Country selection to find region-specific product information
  • Domain Filtering: Include or exclude specific domains from search results

Preferred Sources

  • User-Provided URLs: Specific URLs where users have identified correct, high-quality product information
  • Quality Assurance: User pre-approved sources that receive preferred treatment as context
  • Brand Examples: Product detail pages from official brand sites that users trust
  • Priority Processing: Preferred sources get higher priority in the enrichment context

Sourced Content Types​

  • Technical Specifications: Detailed product measurements, materials, and features
  • Product Descriptions: Marketing copy and feature explanations
  • Compatibility Information: Cross-reference data and compatibility matrices
  • Regulatory Information: Certifications, compliance data, and safety information

Image Enrichment​

Image Enrichment focuses on sourcing images for products, often helping customers reach the desired objective of 10 images per product, which greatly impacts sales conversion rates.

Common Use Cases​

  • Nutritional Tables: Food and supplement facts panels and ingredient lists
  • Different Angles: Multiple product views including front, back, side, and detail shots
  • Updated Imagery: Current product photos from official and public sites
  • Lifestyle Images: Products in use or context-setting environments
  • Packaging Variations: Different size options and packaging configurations

Source Channels​

  • Official Brand Sites: High-quality product photography and marketing images
  • Public Marketplaces: User-generated and merchant-provided imagery
  • Public Resources: Available product catalogs and press materials
  • Retailer Websites: Professional product photography from authorized dealers

Offline Product Benefits​

Image enrichment is particularly valuable for offline products where images are mandatory requirements:

  • Own Sales Channels: Meet internal image requirements for website listings
  • Marketplace Compliance: Fulfill mandatory image requirements for Amazon, eBay, etc.
  • Conversion Optimization: Provide comprehensive visual product information
  • Professional Appearance: Ensure consistent, high-quality product presentation

Task-Based System Architecture​

The enrichment system operates on a task-based architecture where products can only be present in one task at a time. This ensures processing integrity and prevents conflicts during enrichment operations.

Task Management​

  • Single Task Assignment: Each product can only be assigned to one active enrichment task
  • Task Queuing: Multiple enrichment requests for the same product are queued sequentially
  • Status Tracking: Each task maintains detailed status and completion metrics
  • Resource Management: System resources are allocated efficiently across active tasks

Product Metadata After Enrichment​

Once enrichment is completed, product metadata includes detailed task information and completion statistics:

{
  "dataEnrichment": {
    "taskId": "ER00002689",
    "status": "FINISHED",
    "completedAt": 1722934566131,
    "webSourceCount": 0,
    "preferredWebSourceCount": 0,
    "searchWebSourceCount": 0
  },
  "attributeSummary": {
    "enrichableTotal": 55,
    "enrichableCompleted": 27,
    "enrichableCompletionRate": 49.09090909090909
  }
}

Metadata Fields Explained​

  • taskId: Unique identifier for the enrichment task
  • status: Current task status (FINISHED, IN_PROGRESS, FAILED, etc.)
  • completedAt: Unix timestamp of task completion
  • webSourceCount: Number of external web sources used
  • preferredWebSourceCount: Sources from preferred/configured websites
  • searchWebSourceCount: Sources found through web search
  • enrichableTotal: Total number of attributes eligible for enrichment
  • enrichableCompleted: Number of attributes successfully enriched
  • enrichableCompletionRate: Percentage of successful attribute enrichment

Enrichment Blame Tracking​

AI-enriched values include a "blame" field that tracks the source and approval status of enriched content. This enables transparency, quality control, and workflow management for enriched attributes.

Blame Status Types​

{
  "attributes": [
    {
      "key": "smartShortDesp",
      "value": "Experience the ultimate collaboration with the swatch BIOCERAMIC MOONSWATCH. This masterpiece features a sleek white dial, black hands, tachymeter scale, and distinctive subdials. Combining modern innovation with classic design, it's perfect for watch enthusiasts.",
      "blame": "ENRICHMENT"
    },
    {
      "key": "productFeatures",
      "value": "Premium bioceramic construction, Swiss movement, water resistant",
      "blame": "ENRICHMENT_APPROVED"
    },
    {
      "key": "marketingColor",
      "value": "Green",
      "blame": "ENRICHMENT_REJECTED"
    }
  ]
}

Blame Field Values​

  • ENRICHMENT: AI-generated content awaiting user review
  • ENRICHMENT_APPROVED: User-approved AI content via Portal interface
  • ENRICHMENT_REJECTED: User-rejected AI content, eligible for refresh in next enrichment task

Monitoring Progress​

Use the product search API to track enrichment progress:

curl -X POST https://api.trustana.com/v1/products/search \
  -H "X-API-Key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "filter": {
      "dataEnrichment.status": {
        "$in": ["FINISHED"]
      }
    },
    "pagination": {
      "limit": 50,
      "offset": 0
    },
    "sort": {
      "field": "dataEnrichment.completedAt",
      "order": "DESC"
    }
  }'

Workflow Benefits​

  • Transparency: Clear indication of content source and approval status
  • Quality Control: Users can track which content has been reviewed and approved
  • Refresh Logic: Rejected content can be regenerated in subsequent enrichment tasks
  • Audit Trail: Complete history of user decisions and content evolution
  • User Tracking: System maintains user and timestamp information for rejection decisions

Integration Considerations​

Portal-Centric Workflow​

All enrichment operations must be initiated through the Trustana Portal:

  • Task Creation: Set up enrichment tasks with specific product selections
  • Configuration Management: Define enrichment parameters and source preferences
  • Progress Monitoring: Track task status and completion metrics. Products can be QUEUED for a task, IN_PROGRESS, or FINISHED (with completedAt only filled when successful). Use the product search API for remote progress tracking: https://api.trustana.com/v1/products/search
  • Quality Review: Approve or reject enriched content before publishing

API Limitations​

The OpenAPI does not provide:

  • Enrichment task triggering capabilities
  • Real-time enrichment status monitoring
  • Task management or cancellation functions
  • Direct access to enrichment configuration settings

Data Access​

Enriched product data becomes available through standard product retrieval APIs once enrichment tasks are completed.