Making Databases LLM-Ready Building Semantic Layers with Semantido
summary
Claims a production-grade semantic layer called Semantido that ties SQLAlchemy models to operational metadata, keeps semantic definitions in sync with migrations, and enforces descriptions before deploy. Shows concrete tools and patterns: Semantido generates semantic metadata from SQLAlchemy models, FastAPI + LangChain (and Pydantic AI) for text-to-SQL and validation, Neon/Postgres for data and Docker deployment patterns. Demo includes a banking schema, bridge-table failure modes, observability and evals; live demo has a transient DB connection failure but code and repos are available.
AI-written summary · about
chapters (16)
- 0:00Introduction
- 1:07Why semantic layers matter
- 5:46LLM reliability with databases
- 8:58History of semantic computing
- 13:14Semantics beyond words
- 21:03Text-to-SQL challenges
- 27:25LLM SQL failure modes
- 32:24Introducing Semantido
- 41:24Semantic layer architectures
- 48:14Why Semantido from scratch
- 50:21Semantido in application
- 55:56What Semantido is and isn’t designed for
- 1:00:09Semantido annotations
- 1:05:57Agentic workflows with Semantido
- 1:15:12Metadata and business context
- 1:27:15Q&A
related talks
34mLet's talk Databases in Python: SQLAlchemy and Alembic
30mThe magic of SQLAlchemy
32mDataEngBytes 2023 - PER-T1-11 - Mark Kitson
28mThe Ultimate Python Database Toolkit: SQLAlchemy
01h 28mSQLAlchemy and You - Making SQL the Best Thing Since Sliced Bread
10mDIY: Helping People Assess the Correctness of Natural Language to SQL Systems
30mThe Future of Data is Words
35mUsing Semantic Metadata to Create an Automated Microservice Data Mesh