DEVOURED
Neocarta (GitHub Repo)
Neocarta is an open-source library that maps database schemas and business logic into a Neo4j graph to provide AI agents with semantic context.
What: The library, developed by the Neo4j field team, uses connectors for BigQuery and Dataplex to ingest metadata into a graph, which is then exposed via the Model Context Protocol (MCP) to help AI agents reliably generate SQL queries.
Why it matters: By centralizing metadata, business glossaries, and query history in a graph database, developers can move away from fragile prompt-based schema descriptions toward a structured, verifiable context layer for agents.
Takeaway: If you are struggling with unreliable Text2SQL performance, try installing Neocarta using 'pip install neocarta' to serve your data context via MCP.
Deep dive
- Neocarta constructs a semantic layer that links tables, columns, business terms, and query history in a graph.
- Connectors extract metadata from BigQuery and Dataplex without duplicating the raw source data.
- Uses the Model Context Protocol (MCP) to provide agents with tools for schema discovery and relationship traversal.
- Supports optional embedding generation to enable semantic search over table and column metadata.
- Provides a performance-optimized Rust extension for the Neo4j Python driver for faster bulk ingestion.
- Includes query log analysis to map actual data usage patterns against declared schemas.
- Can be used alongside frameworks like LangGraph to guide agentic SQL generation.
Decoder
- MCP (Model Context Protocol): An open standard created by Anthropic that allows AI applications to connect to external data sources and tools through a unified interface.
- Semantic Layer: A business representation of data that defines how different datasets relate to one another and what terms mean, abstracting away the underlying database structure.
- Information Schema: A set of read-only views in most relational databases that provide information about all tables, views, columns, and procedures in the database.
Original article
Full article content is not available for inline reading.
Explore all 59 articles in today's digest
Fresh dev news, AI-digested daily. Never miss a trend at devoured.fyi