oximedia-recommend

Status: [Stable] | Version: 0.1.4 | Tests: 969 | Updated: 2026-04-20

Content recommendation and discovery engine for OxiMedia. Provides comprehensive recommendation capabilities including content-based filtering, collaborative filtering, hybrid approaches, and advanced personalization.

Part of the oximedia workspace — a comprehensive pure-Rust media processing framework.

Features

• Content-based Filtering - Recommend similar content based on features
• Collaborative Filtering - User behavior-based recommendations via matrix factorization (SVD)
• Hybrid Approach - Combine multiple recommendation methods
• User Profiles - Build and manage user preference profiles
• View History - Track and analyze viewing patterns
• Rating System - Handle explicit ratings and implicit feedback
• Trending Detection - Identify trending content in real-time
• Personalization - Context-aware personalized recommendations
• Diversity Enforcement - Ensure recommendation diversity and avoid filter bubbles
• Freshness Balancing - Balance popular and new content
• A/B Testing - Experimentation framework for recommendation strategies
• Bandit Algorithms - Multi-armed bandit for exploration-exploitation
• Cold Start - Strategies for new users and new content
• Sequence Models - Sequential recommendation based on viewing history
• Exploration Policy - Configurable exploration strategies
• Decay Model - Time-based relevance decay
• Score Cache - Cached recommendation scores
• Feature Store - User and item feature management
• Feedback Signal - Implicit and explicit feedback processing
• Impression Tracker - Impression and click tracking
• Item Similarity - Item-to-item similarity computation
• Popularity Bias - Popularity de-biasing
• Ranking - Multi-objective ranking
• Session - Session-based recommendations

Usage

Add to your Cargo.toml:

[dependencies]
oximedia-recommend = "0.1.4"

use oximedia_recommend::{RecommendationEngine, RecommendationRequest, RecommendationStrategy};
use uuid::Uuid;

let mut engine = RecommendationEngine::new();

// Get recommendations for a user
let request = RecommendationRequest {
    user_id: Uuid::new_v4(),
    limit: 10,
    strategy: RecommendationStrategy::Hybrid,
    ..Default::default()
};

let results = engine.recommend(&request)?;

// Record user interaction
engine.record_view(user_id, content_id, watch_time_ms, completed)?;
engine.record_rating(user_id, content_id, 4.5)?;

API Overview

Core types:

• RecommendationEngine — Main engine coordinating all recommendation strategies
• RecommendationRequest — Request with user ID, limit, strategy, and context
• RecommendationStrategy — ContentBased, Collaborative, Hybrid, Personalized, Trending
• Recommendation / RecommendationReason — Result item with score and explanation
• DiversitySettings — Category diversity and serendipity configuration
• RecommendationContext — Device, location, time-of-day context

Modules:

• ab_test — A/B testing framework
• bandits — Multi-armed bandit algorithms
• calibration — Recommendation calibration
• cold_start — Cold start strategies
• collab_filter, collaborative — Collaborative filtering (filter, predict, SVD)
• content, content_based — Content-based filtering
• context_signal — Context signal processing
• decay_model — Time-based decay
• diversity — Diversity enforcement
• error — Error types
• explain — Recommendation explanation generation
• exploration_policy — Exploration strategies
• feature_store — Feature management
• feedback_signal — Feedback processing
• freshness — Content freshness scoring
• history — View/listen history tracking
• hybrid — Hybrid recommendation strategies
• impression_tracker — Impression and click tracking
• item_similarity — Item similarity computation
• personalize — Personalization engine
• popularity_bias — Popularity de-biasing
• profile — User preference profiles
• rank, ranking — Multi-objective ranking
• rating — Explicit rating handling
• recommendation_score — Score computation
• score_cache — Score caching
• sequence_model — Sequential recommendation
• session — Session-based recommendations
• trending — Trending content detection
• user_profile — User profile management

License

Apache-2.0 — Copyright 2024-2026 COOLJAPAN OU (Team Kitasan)