@memberjunction/ai-recommendations-rex

MemberJunction recommendation provider for the rasa.io Rex recommendation engine. Retrieves content recommendations from Rex and saves them into MJ Recommendation/Recommendation Item entities.

Overview

This package implements RecommendationProviderBase for the rasa.io Rex engine. It authenticates with the Rex API, looks up vector IDs from MJ Entity Record Documents, requests content-based recommendations via Rex's /suggest endpoint, and persists the results as RecommendationItem entities in the MemberJunction database.

graph TD
    PROVIDER["RexRecommendationsProvider"] --> TOKEN["Get Rex Access Token\n(/tokens)"]
    PROVIDER --> DOCS["Load Entity Record Documents\n(Vector IDs)"]
    PROVIDER --> SUGGEST["Get Recommendations\n(/suggest)"]
    PROVIDER --> SAVE["Save Recommendation Items\n(MJ Entities)"]

    TOKEN --> REXAPI["Rex API"]
    SUGGEST --> REXAPI
    DOCS --> RUNVIEW["RunView\n(Entity Record Documents)"]
    SAVE --> MJDB["MJ Database"]

    style PROVIDER fill:#7c5295,stroke:#563a6b,color:#fff
    style TOKEN fill:#2d6a9f,stroke:#1a4971,color:#fff
    style DOCS fill:#2d6a9f,stroke:#1a4971,color:#fff
    style SUGGEST fill:#2d6a9f,stroke:#1a4971,color:#fff
    style SAVE fill:#2d6a9f,stroke:#1a4971,color:#fff
    style REXAPI fill:#b8762f,stroke:#8a5722,color:#fff
    style RUNVIEW fill:#2d8659,stroke:#1a5c3a,color:#fff
    style MJDB fill:#2d8659,stroke:#1a5c3a,color:#fff

Features

• Rex API integration: Authenticates and retrieves recommendations from the rasa.io Rex engine
• Batch processing: Processes recommendations in configurable batch sizes (default 200)
• Vector ID resolution: Looks up Pinecone vector IDs from MJ Entity Record Documents
• Type mapping: Configurable mapping from Rex content types to MJ entity names
• Score clamping: Normalizes recommendation scores to a 0-1 probability range
• Error tracking: Optionally logs errors to a MJ List for review
• Auto-registration: Registers with MJ class factory as "rasa.io Rex"

Installation

npm install @memberjunction/ai-recommendations-rex

Configuration

Set the following environment variables:

| Variable | Required | Description | |---|---|---| | REX_API_HOST | Yes | Rex API base URL for authentication | | REX_RECOMMEND_HOST | Yes | Rex recommendation service URL | | REX_USERNAME | Yes | Rex API username | | REX_PASSWORD | Yes | Rex API password | | REX_API_KEY | Yes | Rex API key | | REX_BATCH_SIZE | No | Batch size for processing (default: 200) |

Usage

import { RexRecommendationsProvider } from '@memberjunction/ai-recommendations-rex';
import { RecommendationRequest } from '@memberjunction/ai-recommendations';

// The provider is auto-registered as "rasa.io Rex" via @RegisterClass.
// It can also be instantiated directly:
const provider = new RexRecommendationsProvider();

const request: RecommendationRequest = {
  Recommendations: recommendationEntities,  // Array of RecommendationEntity
  RunID: 'run-id',
  CurrentUser: contextUser,
  Options: {
    EntityDocumentID: 'entity-doc-id',  // Required
    type: 'course',                     // Rex content type
    filters: [{ type: 'course', max_results: 10 }],
    TypeMap: {                          // Optional: custom type-to-entity mapping
      'course': 'Contents',
      'person': 'Contributors'
    }
  }
};

const result = await provider.Recommend(request);

if (result.Success) {
  console.log('Recommendations generated successfully');
} else {
  console.error('Errors:', result.Errors);
}

Key Dependencies

| Dependency | Purpose | |---|---| | @memberjunction/ai-recommendations | RecommendationProviderBase abstract class | | @memberjunction/core | Metadata, RunView, LogError, LogStatus | | @memberjunction/core-entities | Entity types (RecommendationEntity, RecommendationItemEntity, etc.) | | @memberjunction/global | @RegisterClass decorator | | axios | HTTP client for Rex API calls |

Related Packages

• @memberjunction/ai-recommendations - Base recommendation provider abstraction
• @memberjunction/ai-vectors-pinecone - Pinecone vector database provider (source of vector IDs)

License

ISC