Merge pull request #448 from jlowin/openapi-def

Ensure openapi defs for structured objects are loaded properly
This commit is contained in:
Jeremiah Lowin 2025-05-14 14:29:10 -04:00 committed by GitHub
commit deb5e6cc04
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
8 changed files with 486 additions and 126 deletions

View file

@ -13,7 +13,7 @@ from mcp.types import PromptArgument as MCPPromptArgument
from pydantic import BaseModel, BeforeValidator, Field, TypeAdapter, validate_call
from fastmcp.server.dependencies import get_context
from fastmcp.utilities.json_schema import prune_params
from fastmcp.utilities.json_schema import compress_schema
from fastmcp.utilities.logging import get_logger
from fastmcp.utilities.types import (
_convert_set_defaults,
@ -115,7 +115,11 @@ class Prompt(BaseModel):
context_kwarg = find_kwarg_by_type(fn, kwarg_type=Context)
if context_kwarg:
parameters = prune_params(parameters, params=[context_kwarg])
prune_params = [context_kwarg]
else:
prune_params = None
parameters = compress_schema(parameters, prune_params=prune_params)
# Convert parameters to PromptArguments
arguments: list[PromptArgument] = []

View file

@ -21,6 +21,7 @@ from pydantic import (
from fastmcp.resources.types import FunctionResource, Resource
from fastmcp.server.dependencies import get_context
from fastmcp.utilities.json_schema import compress_schema
from fastmcp.utilities.types import (
_convert_set_defaults,
find_kwarg_by_type,
@ -150,6 +151,10 @@ class ResourceTemplate(BaseModel):
# Get schema from TypeAdapter - will fail if function isn't properly typed
parameters = TypeAdapter(fn).json_schema()
# compress the schema
prune_params = [context_kwarg] if context_kwarg else None
parameters = compress_schema(parameters, prune_params=prune_params)
# ensure the arguments are properly cast
fn = validate_call(fn)

View file

@ -12,7 +12,7 @@ from pydantic import BaseModel, BeforeValidator, Field
import fastmcp
from fastmcp.server.dependencies import get_context
from fastmcp.utilities.json_schema import prune_params
from fastmcp.utilities.json_schema import compress_schema
from fastmcp.utilities.logging import get_logger
from fastmcp.utilities.types import (
Image,
@ -81,7 +81,11 @@ class Tool(BaseModel):
context_kwarg = find_kwarg_by_type(fn, kwarg_type=Context)
if context_kwarg:
schema = prune_params(schema, params=[context_kwarg])
prune_params = [context_kwarg]
else:
prune_params = None
schema = compress_schema(schema, prune_params=prune_params)
return cls(
fn=fn,

View file

@ -14,6 +14,7 @@ def _prune_param(schema: dict, param: str) -> dict:
removed = props.pop(param, None)
if removed is None: # nothing to do
return schema
# Keep empty properties object rather than removing it entirely
schema["properties"] = props
if param in schema.get("required", []):
@ -21,7 +22,12 @@ def _prune_param(schema: dict, param: str) -> dict:
if not schema["required"]:
schema.pop("required")
# ── 2. collect all remaining local $ref targets ───────────────────
return schema
def _prune_unused_defs(schema: dict) -> dict:
"""Remove unused definitions from the schema."""
# collect all remaining local $ref targets
used_defs: set[str] = set()
def walk(node: object) -> None: # depth-first traversal
@ -37,7 +43,8 @@ def _prune_param(schema: dict, param: str) -> dict:
walk(schema)
# ── 3. remove orphaned definitions ────────────────────────────────
# remove orphaned definitions
defs = schema.get("$defs", {})
for def_name in list(defs):
if def_name not in used_defs:
@ -48,12 +55,28 @@ def _prune_param(schema: dict, param: str) -> dict:
return schema
def prune_params(schema: dict, params: list[str]) -> dict:
def _prune_additional_properties(schema: dict) -> dict:
"""Remove additionalProperties from the schema if it is False."""
if schema.get("additionalProperties", None) is False:
schema.pop("additionalProperties")
return schema
def compress_schema(
schema: dict,
prune_params: list[str] | None = None,
prune_defs: bool = True,
prune_additional_properties: bool = True,
) -> dict:
"""
Remove the given parameters from the schema.
"""
schema = copy.deepcopy(schema)
for param in params:
for param in prune_params or []:
schema = _prune_param(schema, param=param)
if prune_defs:
schema = _prune_unused_defs(schema)
if prune_additional_properties:
schema = _prune_additional_properties(schema)
return schema

View file

@ -84,6 +84,9 @@ class HTTPRoute(BaseModel):
responses: dict[str, ResponseInfo] = Field(
default_factory=dict
) # Key: status code str
schema_definitions: dict[str, JsonSchema] = Field(
default_factory=dict
) # Store component schemas
# Export public symbols
@ -221,6 +224,27 @@ class OpenAPI31Parser(BaseOpenAPIParser):
logger.warning("OpenAPI schema has no paths defined.")
return []
# Extract component schemas to add to each route
schema_definitions = {}
if hasattr(self.openapi, "components") and self.openapi.components:
components = self.openapi.components
if hasattr(components, "schemas") and components.schemas:
for name, schema in components.schemas.items():
try:
if isinstance(schema, Reference):
resolved_schema = self._resolve_ref(schema)
schema_definitions[name] = self._extract_schema_as_dict(
resolved_schema
)
else:
schema_definitions[name] = self._extract_schema_as_dict(
schema
)
except Exception as e:
logger.warning(
f"Failed to extract schema definition '{name}': {e}"
)
for path_str, path_item_obj in self.openapi.paths.items():
if not isinstance(path_item_obj, PathItem):
logger.warning(
@ -269,6 +293,7 @@ class OpenAPI31Parser(BaseOpenAPIParser):
parameters=parameters,
request_body=request_body_info,
responses=responses,
schema_definitions=schema_definitions,
)
routes.append(route)
logger.info(
@ -386,16 +411,36 @@ class OpenAPI31Parser(BaseOpenAPIParser):
param_schema_dict = {}
if param_schema_obj: # Check if schema exists
# Resolve the schema if it's a reference
resolved_schema = self._resolve_ref(param_schema_obj)
param_schema_dict = self._extract_schema_as_dict(param_schema_obj)
# Ensure default value is preserved from resolved schema
if (
not isinstance(resolved_schema, Reference)
and hasattr(resolved_schema, "default")
and resolved_schema.default is not None
):
param_schema_dict["default"] = resolved_schema.default
elif parameter.content:
# Handle complex parameters with 'content'
first_media_type = next(iter(parameter.content.values()), None)
if (
first_media_type and first_media_type.media_type_schema
): # CORRECTED: Use 'media_type_schema'
param_schema_dict = self._extract_schema_as_dict(
first_media_type.media_type_schema
)
# Resolve the schema if it's a reference
media_schema = first_media_type.media_type_schema
resolved_media_schema = self._resolve_ref(media_schema)
param_schema_dict = self._extract_schema_as_dict(media_schema)
# Ensure default value is preserved from resolved schema
if (
not isinstance(resolved_media_schema, Reference)
and hasattr(resolved_media_schema, "default")
and resolved_media_schema.default is not None
):
param_schema_dict["default"] = resolved_media_schema.default
logger.debug(
f"Parameter '{parameter.name}' using schema from 'content' field."
)
@ -543,6 +588,27 @@ class OpenAPI30Parser(BaseOpenAPIParser):
logger.warning("OpenAPI schema has no paths defined.")
return []
# Extract component schemas to add to each route
schema_definitions = {}
if hasattr(self.openapi, "components") and self.openapi.components:
components = self.openapi.components
if hasattr(components, "schemas") and components.schemas:
for name, schema in components.schemas.items():
try:
if isinstance(schema, Reference_30):
resolved_schema = self._resolve_ref(schema)
schema_definitions[name] = self._extract_schema_as_dict(
resolved_schema
)
else:
schema_definitions[name] = self._extract_schema_as_dict(
schema
)
except Exception as e:
logger.warning(
f"Failed to extract schema definition '{name}': {e}"
)
for path_str, path_item_obj in self.openapi.paths.items():
if not isinstance(path_item_obj, PathItem_30):
logger.warning(
@ -593,6 +659,7 @@ class OpenAPI30Parser(BaseOpenAPIParser):
parameters=parameters,
request_body=request_body_info,
responses=responses,
schema_definitions=schema_definitions,
)
routes.append(route)
logger.info(
@ -711,14 +778,34 @@ class OpenAPI30Parser(BaseOpenAPIParser):
param_schema_dict = {}
if param_schema_obj: # Check if schema exists
# Resolve the schema if it's a reference
resolved_schema = self._resolve_ref(param_schema_obj)
param_schema_dict = self._extract_schema_as_dict(param_schema_obj)
# Ensure default value is preserved from resolved schema
if (
not isinstance(resolved_schema, Reference_30)
and hasattr(resolved_schema, "default")
and resolved_schema.default is not None
):
param_schema_dict["default"] = resolved_schema.default
elif parameter.content:
# Handle complex parameters with 'content'
first_media_type = next(iter(parameter.content.values()), None)
if first_media_type and first_media_type.media_type_schema:
param_schema_dict = self._extract_schema_as_dict(
first_media_type.media_type_schema
)
# Resolve the schema if it's a reference
media_schema = first_media_type.media_type_schema
resolved_media_schema = self._resolve_ref(media_schema)
param_schema_dict = self._extract_schema_as_dict(media_schema)
# Ensure default value is preserved from resolved schema
if (
not isinstance(resolved_media_schema, Reference_30)
and hasattr(resolved_media_schema, "default")
and resolved_media_schema.default is not None
):
param_schema_dict["default"] = resolved_media_schema.default
logger.debug(
f"Parameter '{parameter.name}' using schema from 'content' field."
)
@ -1173,6 +1260,23 @@ def _combine_schemas(route: openapi.HTTPRoute) -> dict[str, Any]:
# Copy the schema and add description if available
param_schema = param.schema_.copy() if isinstance(param.schema_, dict) else {}
# Convert #/components/schemas references to #/$defs references
if isinstance(param_schema, dict) and "$ref" in param_schema:
ref_path = param_schema["$ref"]
if ref_path.startswith("#/components/schemas/"):
schema_name = ref_path.split("/")[-1]
param_schema["$ref"] = f"#/$defs/{schema_name}"
# Also handle anyOf, allOf, oneOf references
for section in ["anyOf", "allOf", "oneOf"]:
if section in param_schema and isinstance(param_schema[section], list):
for i, item in enumerate(param_schema[section]):
if isinstance(item, dict) and "$ref" in item:
ref_path = item["$ref"]
if ref_path.startswith("#/components/schemas/"):
schema_name = ref_path.split("/")[-1]
param_schema[section][i]["$ref"] = f"#/$defs/{schema_name}"
# Add parameter description to schema if available and not already present
if param.description and not param_schema.get("description"):
param_schema["description"] = param.description
@ -1193,8 +1297,19 @@ def _combine_schemas(route: openapi.HTTPRoute) -> dict[str, Any]:
if route.request_body.required:
required.extend(body_schema.get("required", []))
return {
result = {
"type": "object",
"properties": properties,
"required": required,
}
# Add schema definitions if available
if route.schema_definitions:
result["$defs"] = route.schema_definitions
# Use compress_schema to remove unused definitions
from fastmcp.utilities.json_schema import compress_schema
result = compress_schema(result)
return result

View file

@ -1,6 +1,7 @@
import base64
import json
import re
from enum import Enum
import httpx
import pytest
@ -1817,3 +1818,75 @@ class TestReprMethods:
assert f"name={template.name!r}" in template_repr
assert "uri_template=" in template_repr
assert "path=" in template_repr
class TestEnumHandling:
"""Tests for handling enum parameters in OpenAPI schemas."""
async def test_enum_parameter_schema(self):
"""Test that enum parameters are properly handled in tool parameter schemas."""
# Define an enum just like in example.py
class QueryEnum(str, Enum):
foo = "foo"
bar = "bar"
baz = "baz"
# Create a minimal FastAPI app with an endpoint using the enum
app = FastAPI()
@app.post("/items/{item_id}")
def read_item(
item_id: int,
query: QueryEnum | None = None,
):
return {"item_id": item_id, "query": query}
# Create a client for the app
client = AsyncClient(transport=ASGITransport(app=app), base_url="http://test")
# Create the FastMCPOpenAPI server from the app
openapi_spec = app.openapi()
server = FastMCPOpenAPI(
openapi_spec=openapi_spec,
client=client,
name="Enum Test",
)
# Get the tools from the server
tools = server._tool_manager.list_tools()
# Find the read_item tool
read_item_tool = next(
(t for t in tools if t.name == "read_item_items__item_id__post"), None
)
# Verify the tool exists
assert read_item_tool is not None, "read_item tool wasn't created"
# Check that the parameters include the enum reference
assert "properties" in read_item_tool.parameters
assert "query" in read_item_tool.parameters["properties"]
# Check for the anyOf with $ref to the enum definition
query_param = read_item_tool.parameters["properties"]["query"]
assert "anyOf" in query_param
# Find the ref in the anyOf list
ref_found = False
for option in query_param["anyOf"]:
if "$ref" in option and option["$ref"].startswith("#/$defs/QueryEnum"):
ref_found = True
break
assert ref_found, "Reference to enum definition not found in query parameter"
# Check that the $defs section exists and contains the enum definition
assert "$defs" in read_item_tool.parameters
assert "QueryEnum" in read_item_tool.parameters["$defs"]
# Verify the enum definition
enum_def = read_item_tool.parameters["$defs"]["QueryEnum"]
assert "enum" in enum_def
assert enum_def["enum"] == ["foo", "bar", "baz"]
assert enum_def["type"] == "string"

View file

@ -1,110 +1,246 @@
from fastmcp.utilities.json_schema import _prune_param, prune_params
from fastmcp.utilities.json_schema import (
_prune_additional_properties,
_prune_param,
_prune_unused_defs,
compress_schema,
)
def test_prune_param_nonexistent():
"""Test pruning a parameter that doesn't exist."""
schema = {"properties": {"foo": {"type": "string"}}}
result = _prune_param(schema, "bar")
assert result == schema # Schema should be unchanged
class TestPruneParam:
"""Tests for the _prune_param function."""
def test_nonexistent(self):
"""Test pruning a parameter that doesn't exist."""
schema = {"properties": {"foo": {"type": "string"}}}
result = _prune_param(schema, "bar")
assert result == schema # Schema should be unchanged
def test_exists(self):
"""Test pruning a parameter that exists."""
schema = {"properties": {"foo": {"type": "string"}, "bar": {"type": "integer"}}}
result = _prune_param(schema, "bar")
assert result["properties"] == {"foo": {"type": "string"}}
def test_last_property(self):
"""Test pruning the only/last parameter, should leave empty properties object."""
schema = {"properties": {"foo": {"type": "string"}}}
result = _prune_param(schema, "foo")
assert "properties" in result
assert result["properties"] == {}
def test_from_required(self):
"""Test pruning a parameter that's in the required list."""
schema = {
"properties": {"foo": {"type": "string"}, "bar": {"type": "integer"}},
"required": ["foo", "bar"],
}
result = _prune_param(schema, "bar")
assert result["required"] == ["foo"]
def test_last_required(self):
"""Test pruning the last required parameter, should remove required field."""
schema = {
"properties": {"foo": {"type": "string"}, "bar": {"type": "integer"}},
"required": ["foo"],
}
result = _prune_param(schema, "foo")
assert "required" not in result
def test_prune_param_exists():
"""Test pruning a parameter that exists."""
schema = {"properties": {"foo": {"type": "string"}, "bar": {"type": "integer"}}}
result = _prune_param(schema, "bar")
assert result["properties"] == {"foo": {"type": "string"}}
class TestPruneUnusedDefs:
"""Tests for the _prune_unused_defs function."""
def test_prune_param_last_property():
"""Test pruning the only/last parameter, should leave empty properties object."""
schema = {"properties": {"foo": {"type": "string"}}}
result = _prune_param(schema, "foo")
assert "properties" in result
assert result["properties"] == {}
def test_prune_param_from_required():
"""Test pruning a parameter that's in the required list."""
schema = {
"properties": {"foo": {"type": "string"}, "bar": {"type": "integer"}},
"required": ["foo", "bar"],
}
result = _prune_param(schema, "bar")
assert result["required"] == ["foo"]
def test_prune_param_last_required():
"""Test pruning the last required parameter, should remove required field."""
schema = {
"properties": {"foo": {"type": "string"}, "bar": {"type": "integer"}},
"required": ["foo"],
}
result = _prune_param(schema, "foo")
assert "required" not in result
def test_prune_param_with_refs():
"""Test pruning a parameter that has references in $defs."""
schema = {
"properties": {
"foo": {"$ref": "#/$defs/foo_def"},
"bar": {"$ref": "#/$defs/bar_def"},
},
"$defs": {
"foo_def": {"type": "string"},
"bar_def": {"type": "integer"},
},
}
result = _prune_param(schema, "bar")
assert "bar_def" not in result["$defs"]
assert "foo_def" in result["$defs"]
def test_prune_param_all_refs():
"""Test pruning all parameters with refs, should remove $defs."""
schema = {
"properties": {
"foo": {"$ref": "#/$defs/foo_def"},
},
"$defs": {
"foo_def": {"type": "string"},
},
}
result = _prune_param(schema, "foo")
assert "$defs" not in result
def test_prune_params_multiple():
"""Test pruning multiple parameters at once."""
schema = {
"properties": {
"foo": {"type": "string"},
"bar": {"type": "integer"},
"baz": {"type": "boolean"},
},
"required": ["foo", "bar"],
}
result = prune_params(schema, ["foo", "baz"])
assert result["properties"] == {"bar": {"type": "integer"}}
assert result["required"] == ["bar"]
def test_prune_params_nested_refs():
"""Test pruning with nested references."""
schema = {
"properties": {
"foo": {
"type": "object",
"properties": {"nested": {"$ref": "#/$defs/nested_def"}},
def test_removes_unreferenced_defs(self):
"""Test that unreferenced definitions are removed."""
schema = {
"properties": {
"foo": {"$ref": "#/$defs/foo_def"},
},
"bar": {"$ref": "#/$defs/bar_def"},
},
"$defs": {
"nested_def": {"type": "string"},
"bar_def": {"type": "integer"},
},
}
# Removing foo should keep nested_def as it's not referenced anymore
result = _prune_param(schema, "foo")
assert "nested_def" not in result["$defs"]
assert "bar_def" in result["$defs"]
"$defs": {
"foo_def": {"type": "string"},
"unused_def": {"type": "integer"},
},
}
result = _prune_unused_defs(schema)
assert "foo_def" in result["$defs"]
assert "unused_def" not in result["$defs"]
def test_nested_references_kept(self):
"""Test that definitions referenced via nesting are kept."""
schema = {
"properties": {
"foo": {"$ref": "#/$defs/foo_def"},
},
"$defs": {
"foo_def": {
"type": "object",
"properties": {"nested": {"$ref": "#/$defs/nested_def"}},
},
"nested_def": {"type": "string"},
"unused_def": {"type": "integer"},
},
}
result = _prune_unused_defs(schema)
assert "foo_def" in result["$defs"]
assert "nested_def" in result["$defs"]
assert "unused_def" not in result["$defs"]
def test_array_references_kept(self):
"""Test that definitions referenced in array items are kept."""
schema = {
"properties": {
"items": {"type": "array", "items": {"$ref": "#/$defs/item_def"}},
},
"$defs": {
"item_def": {"type": "string"},
"unused_def": {"type": "integer"},
},
}
result = _prune_unused_defs(schema)
assert "item_def" in result["$defs"]
assert "unused_def" not in result["$defs"]
def test_removes_defs_field_when_empty(self):
"""Test that $defs field is removed when all definitions are unused."""
schema = {
"properties": {
"foo": {"type": "string"},
},
"$defs": {
"unused_def": {"type": "integer"},
},
}
result = _prune_unused_defs(schema)
assert "$defs" not in result
class TestPruneAdditionalProperties:
"""Tests for the _prune_additional_properties function."""
def test_removes_when_false(self):
"""Test that additionalProperties is removed when it's false."""
schema = {
"type": "object",
"properties": {"foo": {"type": "string"}},
"additionalProperties": False,
}
result = _prune_additional_properties(schema)
assert "additionalProperties" not in result
def test_keeps_when_true(self):
"""Test that additionalProperties is kept when it's true."""
schema = {
"type": "object",
"properties": {"foo": {"type": "string"}},
"additionalProperties": True,
}
result = _prune_additional_properties(schema)
assert "additionalProperties" in result
assert result["additionalProperties"] is True
def test_keeps_when_object(self):
"""Test that additionalProperties is kept when it's an object schema."""
schema = {
"type": "object",
"properties": {"foo": {"type": "string"}},
"additionalProperties": {"type": "string"},
}
result = _prune_additional_properties(schema)
assert "additionalProperties" in result
assert result["additionalProperties"] == {"type": "string"}
class TestCompressSchema:
"""Tests for the compress_schema function."""
def test_prune_params(self):
"""Test pruning parameters with compress_schema."""
schema = {
"properties": {
"foo": {"type": "string"},
"bar": {"type": "integer"},
"baz": {"type": "boolean"},
},
"required": ["foo", "bar"],
}
result = compress_schema(schema, prune_params=["foo", "baz"])
assert result["properties"] == {"bar": {"type": "integer"}}
assert result["required"] == ["bar"]
def test_prune_defs(self):
"""Test pruning unused definitions with compress_schema."""
schema = {
"properties": {
"foo": {"$ref": "#/$defs/foo_def"},
"bar": {"type": "integer"},
},
"$defs": {
"foo_def": {"type": "string"},
"unused_def": {"type": "number"},
},
}
result = compress_schema(schema)
assert "foo_def" in result["$defs"]
assert "unused_def" not in result["$defs"]
def test_disable_prune_defs(self):
"""Test disabling pruning of unused definitions."""
schema = {
"properties": {
"foo": {"$ref": "#/$defs/foo_def"},
"bar": {"type": "integer"},
},
"$defs": {
"foo_def": {"type": "string"},
"unused_def": {"type": "number"},
},
}
result = compress_schema(schema, prune_defs=False)
assert "foo_def" in result["$defs"]
assert "unused_def" in result["$defs"]
def test_pruning_additional_properties(self):
"""Test pruning additionalProperties when False."""
schema = {
"type": "object",
"properties": {"foo": {"type": "string"}},
"additionalProperties": False,
}
result = compress_schema(schema)
assert "additionalProperties" not in result
def test_disable_pruning_additional_properties(self):
"""Test disabling pruning of additionalProperties."""
schema = {
"type": "object",
"properties": {"foo": {"type": "string"}},
"additionalProperties": False,
}
result = compress_schema(schema, prune_additional_properties=False)
assert "additionalProperties" in result
assert result["additionalProperties"] is False
def test_combined_operations(self):
"""Test all pruning operations together."""
schema = {
"type": "object",
"properties": {
"keep": {"type": "string"},
"remove": {"$ref": "#/$defs/remove_def"},
},
"required": ["keep", "remove"],
"additionalProperties": False,
"$defs": {
"remove_def": {"type": "string"},
"unused_def": {"type": "number"},
},
}
result = compress_schema(schema, prune_params=["remove"])
# Check that parameter was removed
assert "remove" not in result["properties"]
# Check that required list was updated
assert result["required"] == ["keep"]
# Check that unused definitions were removed
assert "$defs" not in result # Both defs should be gone
# Check that additionalProperties was removed
assert "additionalProperties" not in result

View file

@ -13,7 +13,7 @@ import annotated_types
import pytest
from pydantic import BaseModel, Field
from fastmcp.utilities.json_schema import prune_params
from fastmcp.utilities.json_schema import compress_schema
from fastmcp.utilities.types import get_cached_typeadapter
@ -175,7 +175,7 @@ def test_skip_names():
# Get schema and prune parameters
type_adapter = get_cached_typeadapter(func_with_many_params)
schema = type_adapter.json_schema()
pruned_schema = prune_params(schema, params=["skip_this", "also_skip"])
pruned_schema = compress_schema(schema, prune_params=["skip_this", "also_skip"])
# Check that only the desired parameters remain
assert "keep_this" in pruned_schema["properties"]