agentsop-structured-output-picker

What problem does it solve?

It resolves the cross-library decision of how to enforce structured LM outputs when downstream code requires typed/validated objects, selecting both the enforcement mechanism and the failure stance to use when outputs are malformed.

Core Features & Use Cases

• Decode vs validate enforcement choice: picks grammar/token-masking (Outlines/Guidance/provider strict-mode) versus parse-and-retry validation (Instructor/DSPy) based on how costly malformed outputs are and whether decoder access is available.
• Failure stance selection (Assert vs Suggest): standardizes what happens on violations—hard-fail after bounded retries or soft nudge with graceful degradation.
• Integration guidance for typed pipelines: assumes output shape is already decided for typed objects, then focuses specifically on enforcement for API vs local models.

Quick Start

Tell an AI assistant to use agentsop-structured-output-picker when your LM output is parsed into a Pydantic model and you must decide which enforcement library to apply and whether malformed outputs should fail fast or degrade gracefully.