- Home
- Feature Store Software
- Metarank
Metarank
United States
What it is
Metarank is an open-source ranking service for personalized semantic search and recommendations. It uses customer activity such as clicks and purchases to adjust rankings in real time. Listed capabilities include recommendations, LambdaMART reranking, automatic feature generation and retraining, and serving multiple models for A/B testing. For semantic search and reranking, it can use large language models in bi-encoder and cross-encoder modes. Events can come from Apache Kafka, Apache Pulsar, AWS Kinesis Streams, a REST API, or files; files support imports but not inference. Persistence choices include local memory, a local directory, Redis, and S3. Metarank runs as a Docker container or JAR on Linux, Windows, and macOS, and exposes endpoints for feedback ingestion, model training, ranking, recommendations, LLM inference, and Prometheus metrics. It supports online and offline stores and point-in-time joins. The software is free and self-hosted. Production guidance recommends Redis persistence, while multiple API instances require an external load balancer because clustering is not provided out of the box.
For and against
For
- Open-source ranking and recommendation service
- Supports Kafka, Pulsar, Kinesis, REST, and file events
- Offers LambdaMART reranking and multiple-model A/B testing
- Runs on Linux, Windows, and macOS
Against
- Production guidance recommends Redis persistence
- Multiple API instances need an external load balancer
- Files support import but not inference
The verdict
Metarank offers ranking, recommendations, and semantic search capabilities in a self-hosted service. Its production deployment has specific persistence and scaling considerations, so account for those when evaluating it.
Who it suits
Metarank suits teams building personalized search, recommendations, or learning-to-rank systems who can deploy and operate self-hosted software. It may fit teams that want to use customer interaction signals and connect event sources or search systems.
On the record
- API
- Its API includes feedback, training, ranking, recommendations, LLM inference, and Prometheus metrics endpoints.docs.metarank.ai · read 8 Oct 2026
- Data integrations
- Supported event sources include Apache Kafka, Apache Pulsar, AWS Kinesis Streams, REST API, and files; files are supported for import but not inference.docs.metarank.ai · read 8 Oct 2026
- Deployment
- Metarank can run as a JAR or Docker container, and the docs also describe Kubernetes deployment.docs.metarank.ai · read 8 Oct 2026
- Deployment limit
- The production guide says standalone mode cannot be scaled across instances that share a database and recommends serve mode for production.docs.metarank.ai · read 8 Oct 2026
- Error reporting
- Error reporting uses Sentry by default, with breadcrumbs and personally identifiable information tracking disabled; tracking can be disabled in configuration or with METARANK_TRACKING=false.docs.metarank.ai · read 7 Oct 2026
- Event sources
- Supported event sources include Apache Kafka, Apache Pulsar, AWS Kinesis Streams, a REST API, and files.docs.metarank.ai · read 7 Oct 2026
- Maker support
- Metarank Labs describes itself as providing professional services and integration support for Metarank.linkedin.com · read 8 Oct 2026
- Models
- The product overview lists LambdaMART reranking, collaborative filtering recommendations, automatic feature generation, model retraining, and A/B testing among its features.docs.metarank.ai · read 8 Oct 2026
- Performance
- The performance guide lists typical reranking latency of 10–30 ms, Redis memory use of 1–10 GiB, and data import throughput of 1,000–3,000 events per second.docs.metarank.ai · read 7 Oct 2026
- Personalization
- It uses customer signals such as clicks and purchases to personalize ranking in real time.docs.metarank.ai · read 8 Oct 2026
- Privacy
- The docs say default anonymous usage analytics does not track IP addresses, collected data is not shared, and data is stored for one year before removal.docs.metarank.ai · read 7 Oct 2026
- Product
- Metarank is an open-source ranking service for personalized semantic search and recommendations.docs.metarank.ai · read 8 Oct 2026
- Production requirement
- For production, the docs recommend Redis persistence and say multiple API instances require an external load balancer because clustering is not provided out of the box.docs.metarank.ai · read 7 Oct 2026
- Production requirements
- The Kubernetes deployment guide lists Helm, Redis, and an event bus as prerequisites for distributed deployment.docs.metarank.ai · read 8 Oct 2026
- Ranking
- It uses customer signals such as clicks and purchases to improve ranking and support real-time personalization.docs.metarank.ai · read 7 Oct 2026
- Ranking features
- It computes ranking signals including click-through rate, referrer, user agent, and time.docs.metarank.ai · read 8 Oct 2026
- Search
- The docs describe using LLMs in bi-encoder and cross-encoder modes for semantic search and reranking.docs.metarank.ai · read 7 Oct 2026
- Search integrations
- The docs describe Metarank as a secondary reranking service that can work with search systems including Elasticsearch, OpenSearch, and SOLR.docs.metarank.ai · read 8 Oct 2026
