# X Recommendation Algorithm

## Docs

- [CR Mixer API](https://mintlify.wiki/twitter/the-algorithm/api/cr-mixer-api.md): API reference for CR Mixer (Candidate Recommendation Mixer), Twitter's candidate generation service
- [Data Record Formats](https://mintlify.wiki/twitter/the-algorithm/api/data-records.md): DataRecord format used for machine learning features, labels, and predictions in the X Recommendation Algorithm
- [Follow Recommendations Service API](https://mintlify.wiki/twitter/the-algorithm/api/frs-api.md): API reference for Twitter's Follow Recommendations Service (FRS), providing personalized account suggestions
- [Home Mixer API](https://mintlify.wiki/twitter/the-algorithm/api/home-mixer-api.md): API reference for Twitter's Home Mixer service, which constructs and serves Home Timeline content
- [Thrift API Definitions](https://mintlify.wiki/twitter/the-algorithm/api/thrift-definitions.md): Core Thrift type definitions used across the X Recommendation Algorithm services
- [System architecture](https://mintlify.wiki/twitter/the-algorithm/architecture.md): Overview of the components, services, and data flows that power X's recommendation system
- [Timelines Aggregation Framework](https://mintlify.wiki/twitter/the-algorithm/data/aggregation-framework.md): Flexible framework for computing real-time and batch aggregate features from user behavior signals
- [Retrieval Signals](https://mintlify.wiki/twitter/the-algorithm/data/retrieval-signals.md): User behavior signals powering candidate sourcing in Twitter's recommendation algorithm
- [Unified User Actions](https://mintlify.wiki/twitter/the-algorithm/data/unified-user-actions.md): Centralized real-time stream of user actions consumed across Twitter's product, ML, and marketing teams
- [User Signal Service](https://mintlify.wiki/twitter/the-algorithm/data/user-signals.md): Centralized platform for comprehensive user action and behavior data across Twitter
- [Building and Testing](https://mintlify.wiki/twitter/the-algorithm/development/building.md): Learn how to build and test the X Recommendation Algorithm using Bazel
- [Contributing Guidelines](https://mintlify.wiki/twitter/the-algorithm/development/contributing.md): Learn how to contribute to the X Recommendation Algorithm project
- [Development Setup](https://mintlify.wiki/twitter/the-algorithm/development/setup.md): Set up your development environment for the X Recommendation Algorithm
- [How the recommendation algorithm works](https://mintlify.wiki/twitter/the-algorithm/how-it-works.md): End-to-end explanation of how X generates personalized recommendations from candidate generation to final ranking
- [X's recommendation algorithm](https://mintlify.wiki/twitter/the-algorithm/introduction.md): Open-source recommendation system that powers For You Timeline, Search, Explore, and Notifications across X
- [Candidate Generation](https://mintlify.wiki/twitter/the-algorithm/ml/candidate-generation.md): Multi-source candidate retrieval system that reduces billions of tweets to thousands using graph-based algorithms and user engagement signals
- [Navi ML Serving](https://mintlify.wiki/twitter/the-algorithm/ml/navi.md): High-performance machine learning model serving infrastructure built in Rust with support for TensorFlow, ONNX, and PyTorch runtimes
- [Product Mixer Framework](https://mintlify.wiki/twitter/the-algorithm/ml/product-mixer.md): Composable service framework for building scalable content recommendation pipelines using reusable components and declarative configuration
- [Ranking Systems](https://mintlify.wiki/twitter/the-algorithm/ml/ranking.md): Two-stage ranking architecture using light ranker for candidate pre-filtering and heavy ranker for final scoring in X's recommendation pipeline
- [TWML Framework](https://mintlify.wiki/twitter/the-algorithm/ml/twml.md): Legacy TensorFlow-based machine learning framework used for training light ranker models in X's recommendation pipeline
- [Graph Feature Service](https://mintlify.wiki/twitter/the-algorithm/models/graph-features.md): Distributed system for computing graph-based features between user pairs
- [Real Graph](https://mintlify.wiki/twitter/the-algorithm/models/real-graph.md): Machine learning model to predict user interaction likelihood
- [SimClusters](https://mintlify.wiki/twitter/the-algorithm/models/simclusters.md): Community-based embeddings for users, tweets, and content recommendation
- [Trust and Safety Models](https://mintlify.wiki/twitter/the-algorithm/models/trust-and-safety.md): ML models for detecting NSFW, toxic, and abusive content
- [TwHIN](https://mintlify.wiki/twitter/the-algorithm/models/twhin.md): Dense knowledge graph embeddings for users and tweets
- [CR Mixer](https://mintlify.wiki/twitter/the-algorithm/services/cr-mixer.md): Candidate Retrieval and Mixing service for fast iteration on candidate generation
- [Follow Recommendations Service](https://mintlify.wiki/twitter/the-algorithm/services/follow-recommendations.md): Personalized account recommendations and Who-To-Follow suggestions
- [Home Mixer](https://mintlify.wiki/twitter/the-algorithm/services/home-mixer.md): The main service for constructing and serving Twitter's Home Timelines
- [Pushservice](https://mintlify.wiki/twitter/the-algorithm/services/pushservice.md): Push notification recommendation service for personalized user notifications
- [Timeline Ranker](https://mintlify.wiki/twitter/the-algorithm/services/timelineranker.md): Legacy service providing relevance-scored tweets from search and graph sources
- [Tweetypie](https://mintlify.wiki/twitter/the-algorithm/services/tweetypie.md): Core Tweet service for reading and writing Tweet data