ShapedQL – The SQL engine for search, feeds, and AI agents

Summary: ShapedQL is a domain-specific SQL engine that compiles queries into real-time ranking pipelines for search, personalized feeds, and AI agent memory. It integrates retrieval, filtering, scoring, and reordering into a single interface, replacing complex multi-service stacks with concise SQL code and native multi-modal embeddings.

What it does

ShapedQL compiles simple SQL into multi-stage pipelines that retrieve candidates from various sources, apply filters, rank results using live ML models, and reorder outputs to ensure diversity. It supports hybrid search, collaborative filtering, and real-time model scoring within one query.

Who it's for

It is designed for engineering teams building relevance-based systems like "For You" feeds, search engines, or AI agents requiring personalized memory and ranking.

Why it matters

ShapedQL eliminates the need for stitching together vector databases, feature stores, and reranking services, reducing thousands of lines of infrastructure code to about 30 lines of SQL for faster development and maintenance.