Struggling with complex nested JSON schema validation errors in my microservice API gateway

• Initially, I used Ajv for JSON Schema validation. While powerful, I found it quite cumbersome to manage and express highly dynamic or conditional schemas effectively, especially as they grew in complexity.
• I then switched to Zod, which is TypeScript-first and generally offers a much better developer experience. However, I'm really struggling with how to elegantly define recursive types and, more critically, complex conditional logic in deeply nested objects. It feels like I'm fighting the type system or ending up with overly verbose schemas.
• I've also tried implementing custom validation functions, but these quickly become unwieldy, hard to test, and even harder to maintain as the schemas for different endpoints evolve.

Imagine a payload where a field's presence or type depends on another field's value, all nested within an array of objects. For instance, if an item.type is 'product', then item.details.productId is required, but if item.type is 'service', then item.details.serviceId is required, and both have different validation rules. When a specific condition isn't met, the validator either passes invalid data or, more often, throws a very generic error that doesn't pinpoint the exact sub-field that failed. This makes debugging a nightmare.

// Simplified example of a validation error I'm seeing:
// Expected:
// {
//   "status": "error",
//   "issues": [
//     { "path": ["items", 1, "details", "reason"], "message": "String must contain at least 1 character(s)" }
//   ]
// }
//
// Actual (when using current Zod schema with a complex conditional):
// {
//   "status": "error",
//   "issues": [
//     { "path": [], "message": "Invalid input" } // Too generic, doesn't help debug
//   ]
// }