backend/app/crud/validator_config.py (1)

1-1: Minor: typing.List is deprecated; use built-in list.

Per Ruff UP035. Lines 92 and 113 use List[dict] — replace with list[dict] for consistency with the rest of the file (e.g., Line 52).

backend/app/alembic/versions/003_added_validator_config.py (1)

58-61: Low-cardinality indexes on on_fail_action and is_enabled are unlikely to help.

These columns have very few distinct values (3 and 2 respectively). Single-column B-tree indexes on such columns provide negligible selectivity and add write overhead. Consider removing them unless they're part of specific composite query plans.

backend/app/tests/test_validator_configs_integration.py (1)

67-87: get_validator and list_validators_by_config_id are near-duplicates hitting the same endpoint.

Both methods construct the same URL pattern (BASE_URL + id + query_params). The only difference is the extra **query_params support in list_validators_by_config_id. Also, get_validator is now misleadingly named since the endpoint returns a list by config_id, not a single validator by row ID.

Consider consolidating into one method (e.g., list_validators_by_config_id) and updating all call sites, or at minimum have get_validator delegate to list_validators_by_config_id.

♻️ Proposed consolidation

-    def get_validator(self, client, validator_id):
-        """Helper to get a specific validator."""
-        return client.get(f"{BASE_URL}{validator_id}{DEFAULT_QUERY_PARAMS}")
-
     def list_validators(self, client, **query_params):
         """Helper to list validators with optional filters."""
         params_str = (
@@ -80,11 +76,11 @@
 
     def list_validators_by_config_id(self, client, config_id, **query_params):
-        """Helper to list validators by config_id."""
+        """Helper to list validators by config_id (also used for single-validator retrieval)."""
         params_str = (
             f"?organization_id={TEST_ORGANIZATION_ID}&project_id={TEST_PROJECT_ID}"
         )
         if query_params:
             params_str += "&" + "&".join(f"{k}={v}" for k, v in query_params.items())
         return client.get(f"{BASE_URL}{config_id}{params_str}")

backend/app/tests/test_guardrails_api.py (1)

5-6: Unused imports: GuardrailOnFail and GuardrailRequest are not referenced in any test.

These imports appear to have been added in preparation for new tests that aren't included yet. If tests exercising the normalizer and on-fail behavior are planned, consider adding them in this PR; otherwise these will be flagged by linters.

backend/app/alembic/versions/003_added_validator_config.py (1)

55-58: Indexes on low-cardinality columns may not be beneficial.

on_fail_action has only 3 possible enum values and is_enabled is a boolean. Standalone B-tree indexes on such low-cardinality columns rarely improve query performance and add write overhead. These are typically only useful as part of a composite index or if the value distribution is heavily skewed (e.g., very few disabled rows).

Consider whether these indexes are actually needed for your query patterns, or if a composite index would serve better.

backend/app/tests/test_validator_configs_integration.py (1)

89-147: Batch endpoint tests cover the happy path — consider adding error-case coverage.

The tests validate successful batch create and batch fetch, which is good. Consider adding tests for edge cases in a follow-up:

• Batch create with duplicate type+stage combinations (should trigger IntegrityError / 400).
• Batch fetch with non-existent IDs (should return partial or empty results).
• Empty validators list in batch create.

backend/app/crud/validator_config.py (1)

2-2: Use built-in list instead of typing.List.

typing.List is deprecated since Python 3.9. Use list directly for type hints, which is already done elsewhere in this file (e.g., line 58).

Proposed fix

-from typing import Any, List, Optional
+from typing import Any, Optional

Then replace List[dict] with list[dict] and List[ValidatorBatchFetchItem] with list[ValidatorBatchFetchItem] on lines 102, 122, 123, 140.

backend/app/schemas/validator_config.py (1)

23-24: Consider adding a minimum-length constraint on validators.

An empty list is currently valid, which would result in a no-op batch creation (committing zero rows). A min_length=1 constraint would give the caller an immediate 422 instead of a silent success with an empty result.

Proposed fix

+from pydantic import ConfigDict, field_validator
+
 class ValidatorBatchCreate(SQLModel):
-    validators: list[ValidatorCreate]
+    validators: list[ValidatorCreate] = Field(min_length=1)

Or using Pydantic's annotated form:

validators: list[ValidatorCreate] = Field(min_length=1)

backend/app/crud/validator_config.py (2)

2-2: Use list instead of typing.List (deprecated since Python 3.9).

typing.List is used on lines 102, 122, 123, and 137 while list (lowercase) is already used elsewhere in the file (e.g., line 58). Standardize to the built-in list for consistency and to clear the Ruff UP035 warning.

Proposed fix

-from typing import Any, List, Optional
+from typing import Any, Optional

-    ) -> List[dict]:
+    ) -> list[dict]:

-        payload: List[ValidatorBatchFetchItem],
-    ) -> List[dict]:
+        payload: list[ValidatorBatchFetchItem],
+    ) -> list[dict]:

-        response: List[dict] = []
+        response: list[dict] = []

Also applies to: 102-102, 122-123, 137-137

---

196-199: Config keys could shadow base model fields in flatten.

{**base, **config} lets config keys overwrite base fields (e.g., id, type, stage). While split_validator_payload prevents this at creation time, data inserted via migrations or direct DB access could contain conflicting keys. Consider giving base fields precedence or logging on collision.

Defensive option: base wins

-        return {**base, **config}
+        return {**config, **base}

backend/app/alembic/versions/003_added_validator_config.py (1)

55-58: Low-cardinality indexes may not provide query benefits.

on_fail_action has only 3 enum values and is_enabled is a boolean. Single-column indexes on such low-cardinality columns rarely improve query performance and add write overhead. These columns are typically more useful as part of composite indexes (e.g., combined with organization_id + project_id) or as partial indexes (e.g., WHERE is_enabled = true).

Consider whether these standalone indexes are actually needed for your query patterns.

backend/app/api/routes/validator_configs.py (2)

9-9: Unused import: ValidatorBatchCreate.

ValidatorBatchCreate is imported but not used anywhere in this file.

Proposed fix

 from app.schemas.validator_config import (
-    ValidatorBatchCreate,
     ValidatorBatchFetchItem,
     ValidatorCreate,
     ValidatorResponse,
     ValidatorUpdate,
 )

---

54-65: Consider adding a batch size limit to prevent abuse.

The endpoint accepts an unbounded list[ValidatorBatchFetchItem]. A very large payload could generate a massive IN (...) clause. Consider adding a reasonable upper bound (e.g., 100 or 500 items) and returning a 400 if exceeded.

backend/app/crud/validator_config.py (2)

11-15: Unused import: ValidatorBatchCreate.

ValidatorBatchCreate is imported but not used in this file.

Proposed fix

 from app.schemas.validator_config import (
-    ValidatorBatchCreate,
     ValidatorBatchFetchItem,
     ValidatorCreate,
 )

---

2-2: Use built-in list instead of typing.List.

Per Ruff UP035, typing.List is deprecated in favor of the built-in list type hint (Python 3.9+). This applies to all usages in this file (lines 60, 75, 81, 95).

Proposed fix

-from typing import Any, List, Optional
+from typing import Any, Optional

Then replace List[dict] → list[dict] and List[ValidatorBatchFetchItem] → list[ValidatorBatchFetchItem] throughout.

backend/app/crud/validator_config.py (1)

2-2: Use built-in list instead of typing.List.

typing.List is deprecated since Python 3.9. Line 60 should use list[dict] directly.

Proposed fix

-from typing import Any, List, Optional
+from typing import Any, Optional

And on line 60:

-    ) -> List[dict]:
+    ) -> list[dict]:

backend/app/schemas/guardrail_config.py (1)

40-84: Model validator applies to all GuardrailRequest instantiations, not just config-API payloads.

The mode="before" validator fires on every GuardrailRequest construction — including the normal guardrails runtime path. While the logic is safe (it only strips fields that don't belong to the runtime schema and remaps on_fail_action → on_fail only when on_fail is absent), this coupling is worth being intentional about. If a future validator config introduces a field that collides with drop_fields (e.g. stage used in a different sense), it would be silently dropped.

Consider making the normalization an explicit classmethod (e.g. GuardrailRequest.from_config_api(data)) rather than an implicit pre-validator, so callers opt in.

backend/app/alembic/versions/003_added_validator_config.py (1)

55-58: Indexes on low-cardinality columns add write overhead with minimal query benefit.

on_fail_action has only 3 possible values and is_enabled is a boolean. Single-column B-tree indexes on such columns are rarely selective enough to be used by the query planner — queries already filter by organization_id + project_id first, which narrows the result set sufficiently.

Consider removing these indexes or, if you do need to query by these columns frequently, using a composite index (e.g., (organization_id, project_id, is_enabled)).

backend/app/crud/validator_config.py (1)

2-2: Use built-in list instead of typing.List (deprecated per Ruff UP035).

Proposed fix

-from typing import List, Optional
+from typing import Optional

Then on line 57:

-    ) -> List[dict]:
+    ) -> list[dict]: