Neocarta: Build a Semantic Layer Graph for Text-to-SQL Agents to Fix Table Understanding Failures Upstream
JeremyCMorgan · x · 2026-07-30
Jeremy Morgan introduces Neocarta, an open-source library built on Neo4j that provides AI agents with a semantic understanding of data. The core problem: Text-to-SQL agents fail before SQL generation—they don't know what table names mean or which joins matter. Neocarta constructs a knowledge graph from BigQuery metadata, business terms, and observed query logs, then serves it to agents over MCP (Model Context Protocol), attacking the failure upstream. The project is open-sourced on GitHub (neo4j-labs/neocarta) with agent integration, evaluation, and examples.
More from coding & agent
- GrokTerm 0.1.20 Released: Integrates Grok Voice 2.0 — Daniel_Farinax · 2026-07-30
- Pi School: A Free, Self-Paced Course for the Minimal Pi Agent Harness — craigsdennis · 2026-07-30
- Gradio 6.21.0 Released: Improved Proxy Support and Event Chain Reliability — Gradio · 2026-07-30
- TanML: Open-Source Tabular Model Validation Toolkit Released — AccomplishedLeg1508 · 2026-07-30
- Y Combinator Highlights Hiloop: An Autonomous AI Research Platform — ycombinator · 2026-07-30
- AgentSnap: Open-Source Snapshot Testing to Catch LLM Agent Regressions — iamfaham5 · 2026-07-30