convex-return-validators

Use return validators in Convex functions only when you need an exact runtime contract, such as for components codegen, OpenAPI generation, or handling unvalidated external data. Otherwise, rely on TypeScript inference by default to avoid unnecessary verbosity.

To enforce runtime validation for external data in Convex, apply return validators to your queries, mutations, or actions. This ensures the returned data matches an exact runtime contract, preventing unvalidated external data from causing runtime errors.

Prefer TypeScript inference by default in Convex to avoid unnecessary validators that hinder development. Only add return validators when you require an exact runtime contract for specific scenarios like static codegen or external data integration.

To reduce hallucinations and verbosity in AI-generated Convex code, use specific prompts and rules that enforce safe patterns, such as defaulting to TypeScript inference and only applying return validators when an exact runtime contract is explicitly required.

Yes, you can reuse return validators across Convex functions by sharing them. You can also derive new shapes from existing validators using methods like .pick, .omit, or .extend to minimize duplication and maintain type safety.

Adding unnecessary return validators to Convex mutations can hinder development by creating unnecessary code verbosity. The updated guidance encourages relying on TypeScript inference by default and only defining validators when a strict runtime contract is needed.