Text to SQL That Actually Works
Text to SQL with LLMs, Part 4: Production-Grade Systems: Semantic Layers, Agents, Guardrails, and Evals
How production text to SQL systems work: semantic layers, agentic self-correction, read-only guardrails, MCP integration, and the eval harness that proves accuracy.
Part 3 / 4Text to SQL That Actually Works
Text to SQL with LLMs, Part 3: Retrieval, Golden Queries, and Business Context
Static schema context has a ceiling. Learn how retrieval, verified question-SQL pairs, a business glossary, and value grounding push text to SQL accuracy higher.
Part 2 / 4Text to SQL That Actually Works
Text to SQL with LLMs, Part 2: Schema Context Engineering That Actually Moves Accuracy
The biggest accuracy lever in text to SQL is how you present your schema. Column descriptions, sample values, relationships, and curated views, with SQL to build them.
Part 1 / 4Text to SQL That Actually Works
Text to SQL with LLMs, Part 1: Why the Baseline Approach Fails on Real Databases
Basic schema-in-prompt text to SQL still fails on real databases in 2026. See the current benchmark numbers, the five failure modes, and what actually fixes them.