Pydantic JSON Schema gen chokes on `BeforeValidator` with a

The Fix

pip install pydantic==2.10.4

Based on closed pydantic/pydantic issue #11033 · PR/commit linked

Production note: Most teams hit this during upgrades or environment changes. Roll out with a canary and smoke critical endpoints (health, OpenAPI/docs) before 100%.

Jump to Verify Open PR/Commit

@@ -4,8 +4,6 @@
 
 if TYPE_CHECKING:
-    from pydantic_core import CoreSchema
-
     from ..config import JsonDict, JsonSchemaExtraCallable

repro.py

from pydantic.plugin import TypeAlias
from typing_extensions import TypeAliasType
from typing import Annotated
from pydantic import BaseModel, BeforeValidator, AfterValidator

class TypeA(BaseModel):
    pass

class TypeB(BaseModel):
    pass

FieldType = Annotated[list[str], BeforeValidator(lambda x: x if isinstance(x, list) else [x], TypeA | TypeB)]

class Model(BaseModel):
    field: FieldType

print(Model.model_json_schema())

verify

Re-run the minimal reproduction on your broken version, then apply the fix and re-run.

fix.md

Option A — Upgrade to fixed release\npip install pydantic==2.10.4\nWhen NOT to use: This fix should not be applied if the schema relies on complex metadata patterns.\n\nOption C — Workaround\nis obviously to make the input schema.... a `TypeAliasType`! 😂\nWhen NOT to use: This fix should not be applied if the schema relies on complex metadata patterns.\n\n

Why This Fix Works in Production

Mechanism: The JSON Schema generation fails due to unresolved references in union types with BeforeValidator
Why the fix works: Partly fixes the issue with JSON Schema generation when using BeforeValidator with union types. (first fixed release: 2.10.4).

Production impact:

If left unfixed, the same config can fail only in production (env differences), causing startup failures or partial feature outages.

Why This Breaks in Prod

Shows up under Python 3.11 in real deployments (not just unit tests).
The JSON Schema generation fails due to unresolved references in union types with BeforeValidator
Surfaces as: Traceback (most recent call last):

Proof / Evidence

GitHub issue: #11033
Fix PR: https://github.com/pydantic/pydantic/pull/11085
First fixed release: 2.10.4
Reproduced locally: No (not executed)
Last verified: 2026-02-09
Confidence: 0.95
Did this fix it?: Yes (upstream fix exists)
Own content ratio: 0.57

Discussion

High-signal excerpts from the issue thread (symptoms, repros, edge-cases).

Jump to Sources Open on GitHub

“I suspect this may be related to https://github.com/pydantic/pydantic/blob/aee6057378ccfec02126bf9c984a9b6d6b411777/pydantic/functional_validators.py#L130 (maybe it doesn't handle more complex schemas?)”

@thejcannon · 2024-12-03 · source

“Oh yeah the input_schema is {'type': 'union', 'choices': [{'type': 'definition-ref', 'schema_ref': '__main__.TypeA:5517702256'}, {'type': 'definition-ref', 'schema_ref': '__main__.TypeB:5805303600'}]} so the resolve_ref_schema call should be recursive over the”

@thejcannon · 2024-12-04 · source

“Which means the workaround is obviously to make the input schema.... a TypeAliasType! 😂”

@thejcannon · 2024-12-04 · source

“@Viicos, if I understand your point correctly, we'd be migrating this info out of metadata and onto the core schema directly? I'm in favor of…”

@sydney-runkle · 2024-12-06 · source

Error Message

Stack trace

error.txt

Error Message
-------------
Traceback (most recent call last):
...
  File "/<...>/.venv/lib/python3.11/site-packages/pydantic/json_schema.py", line 2245, in _add_json_refs
    _add_json_refs(self.definitions[defs_ref])
                   ~~~~~~~~~~~~~~~~^^^^^^^^^^
KeyError: '__main____TypeA-Input__1'

Minimal Reproduction

repro.py

from pydantic.plugin import TypeAlias
from typing_extensions import TypeAliasType
from typing import Annotated
from pydantic import BaseModel, BeforeValidator, AfterValidator

class TypeA(BaseModel):
    pass

class TypeB(BaseModel):
    pass

FieldType = Annotated[list[str], BeforeValidator(lambda x: x if isinstance(x, list) else [x], TypeA | TypeB)]

class Model(BaseModel):
    field: FieldType

print(Model.model_json_schema())

Environment

Python: 3.11
Pydantic: 2

What Broke

JSON Schema generation results in a KeyError, causing application failures.

Why It Broke

The JSON Schema generation fails due to unresolved references in union types with BeforeValidator

Fix Options (Details)

Option A — Upgrade to fixed release Safe default (recommended)

pip install pydantic==2.10.4

When NOT to use: This fix should not be applied if the schema relies on complex metadata patterns.

Use when you can deploy the upstream fix. It is usually lower-risk than long-lived workarounds.

Option C — Workaround Temporary workaround

is obviously to make the input schema.... a `TypeAliasType`! 😂

When NOT to use: This fix should not be applied if the schema relies on complex metadata patterns.

Use only if you cannot change versions today. Treat this as a stopgap and remove once upgraded.

Fix reference: https://github.com/pydantic/pydantic/pull/11085

First fixed release: 2.10.4

Last verified: 2026-02-09. Validate in your environment.

When NOT to Use This Fix

This fix should not be applied if the schema relies on complex metadata patterns.

Verify Fix

verify

Re-run the minimal reproduction on your broken version, then apply the fix and re-run.

Did This Fix Work in Your Case?

Quick signal helps us prioritize which fixes to verify and improve.

Prevention

Add a CI check that diffs key outputs after upgrades (OpenAPI schema snapshots, JSON payload shapes, CLI output).
Upgrade behind a canary and run integration tests against the canary before 100% rollout.

Version Compatibility Table

Version	Status
2.10.4	Fixed

Related Issues

No related fixes found.

Cluster: pydantic:configuration-error Pydantic hub Pydantic best practices All hubs All clusters

Related clusters: Data consistency OpenAPI schema Serialization

Sources

We don’t republish the full GitHub discussion text. Use the links above for context.

Pydantic JSON Schema gen chokes on `BeforeValidator` with a union in the `json_schema_input_type` (Fix)