AI agents

Choosing and using JSON schemas for LLM output

• Field Order as a Contract: Testing the First Keys
• JSON Schema Patterns for LLM Outputs
• JSON Schema Patterns That Stream Well
• Schema Tests for Approval Requests: Fixtures That Catch Edits
• A Confidence Schema for Support Answers: Score and Source
• Choosing Nullable, Enum or Union by First-Failure Rate
• Field-Level Extraction Before Merging
• Incremental JSON Parsers Compared: Which Field Arrives First
• Saving a Partial Object: Staging Rows and Commit Rules
• Schema Design for Reliability: Nullable Fields, Enums and Unions
• Schema-Driven Routing: Picking the Next Step
• Versioning Enum Values in LLM Outputs