models: Change TypeError to ValueError in validation logic

- Updated `_internal.py` to raise `ValueError` instead of `TypeError`
  for credential validation errors, ensuring compliance with Pydantic
  error handling best practices.
- Documented this change in AGENTS.md to guide contributors on proper
  error types for validators.

Assisted-by: Codex

Author: datalogics-kam

AGENTS.md

@@ -71,6 +71,9 @@
 - Avoid `@field_validator` on payload models. Prefer existing `BeforeValidator`
   helpers (e.g., `_allowed_mime_types`) so validation remains declarative and
   consistent across schemas.
+- In Pydantic validators, raise `ValueError`/`AssertionError` (or
+  `PydanticCustomError` when needed), not `TypeError`, so callers consistently
+  receive `ValidationError` surfaces.
 - Keep user-facing `PdfRestClient` and `AsyncPdfRestClient` endpoint helpers
   thin: they should primarily assemble payload dicts and delegate validation to
   payload models (`model_validate`). Avoid duplicating payload validation in

src/pdfrest/models/_internal.py

@@ -1633,7 +1633,7 @@ def _normalize_credentials(cls, data: Any) -> Any:
                 "credentials must be a mapping with either pfx/passphrase or "
                 "certificate/private_key."
             )
-            raise TypeError(msg)
+            raise ValueError(msg)  # noqa: TRY004
 
         normalized: dict[str, Any] = {str(key): value for key, value in payload.items()}
         credential_map = cast(Mapping[object, Any], credentials)