Metarank is an open-source, low-code ranking service designed to enhance user engagement by personalizing search results, recommendations, and listings. It targets developers and data scientists seeking to integrate machine learning-based ranking into existing systems without extensive custom coding, offering faster iteration and improved relevance.

How It Works

Metarank employs a real-time, stateless architecture that leverages Redis for state management, enabling horizontal scaling and high throughput. It supports various machine learning models, including LambdaMART and collaborative filtering (ALS), and can ingest signals from diverse streaming sources. The service automatically computes dozens of common ranking features (e.g., CTR, user agent) and integrates with LLMs for semantic understanding, optimizing for low reranking latency (10-20ms).

Quick Start & Requirements

• Install/Run: docker run -i -t -p 8080:8080 -v $(pwd):/opt/metarank metarank/metarank:latest standalone --config /opt/metarank/config.yml --data /opt/metarank/events.jsonl.gz
• Prerequisites: Docker, curl.
• Resources: Requires downloading events.jsonl.gz and config.yml (approx. 100MB compressed).
• Links: Docs, Website, Demo

Highlighted Details

• Real-time personalization and reranking for search and recommendations.
• Supports LLM-based semantic search and traditional collaborative filtering.
• Automated feature generation and model retraining (AutoML).
• Optimized for low reranking latency and horizontal scalability.

Maintenance & Community

• Active development with a public roadmap and community Slack channel.
• Community Slack

Licensing & Compatibility

• Apache 2.0 License.
• Permissive license suitable for commercial use and integration into closed-source applications.

Limitations & Caveats

The README indicates a "[TODO]" for semantic neural search features, suggesting this functionality may be incomplete or under development.