Module

schema

Function signature → JSON Schema for MCP tool compatibility.

8Classes9Functions

Classes

MinLen 1 ▼
Minimum length for strings (minLength) or items for arrays (minItems).

Minimum length for strings (minLength) or items for arrays (minItems).

Attributes

Name Type Description
value int —
MaxLen 1 ▼
Maximum length for strings (maxLength) or items for arrays (maxItems).

Maximum length for strings (maxLength) or items for arrays (maxItems).

Attributes

Name Type Description
value int —
Gt 1 ▼
Exclusive minimum constraint for numbers.

Exclusive minimum constraint for numbers.

Attributes

Name Type Description
value int | float —
Lt 1 ▼
Exclusive maximum constraint for numbers.

Exclusive maximum constraint for numbers.

Attributes

Name Type Description
value int | float —
Ge 1 ▼
Inclusive minimum constraint for numbers.

Inclusive minimum constraint for numbers.

Attributes

Name Type Description
value int | float —
Le 1 ▼
Inclusive maximum constraint for numbers.

Inclusive maximum constraint for numbers.

Attributes

Name Type Description
value int | float —
Pattern 1 ▼
Regex pattern constraint for strings.

Regex pattern constraint for strings.

Attributes

Name Type Description
value str —
Description 1 ▼
Override or supplement the parameter description.

Override or supplement the parameter description.

Attributes

Name Type Description
value str —

Functions

function_to_schema 3 dict[str, Any] ▼
Generate MCP-compatible JSON Schema from function type annotations. Parameters…
def function_to_schema(func: Callable[..., Any], *, strict: bool = False, warn_missing_docs: bool = False) -> dict[str, Any]

Generate MCP-compatible JSON Schema from function type annotations.

Parameters with defaults are optional (not in required). X | None unions unwrapped to base type. Supports: str, int, float, bool, list[X], dict[str, X], X | None, Enum, Literal, dataclass, TypedDict, Union.

Parameter descriptions are extracted from the function's docstring (Google, NumPy, or Sphinx style) and included as"description" fields in the schema properties.

Context-typed parameters (and any parameter named ctx) are intentionally omitted from the returned schema. The CLI dispatcher injects them at call time, so they are invisible to MCP clients and should not appear intools/listdescriptors. See _is_context_type() for the exact detection rules.

When strict is True, unrecognized type annotations raise TypeError instead of silently falling back to "string".

When warn_missing_docs is True, every schema parameter without a description (noArgs: entry and no Annotated[..., Description(...)]) emits aUserWarning. Default Falseso production schema generation stays silent;milo verifyopts in.

Parameters
Name Type Description
func Callable[..., Any]
strict bool Default:False
warn_missing_docs bool Default:False
Returns
dict[str, Any]
_function_to_schema_cached 2 tuple[dict[str, Any], tu… ▼
Compute (schema, undocumented_param_names). Cached; warnings live in the wrappe…
def _function_to_schema_cached(func: Callable[..., Any], *, strict: bool = False) -> tuple[dict[str, Any], tuple[str, ...]]

Compute (schema, undocumented_param_names). Cached; warnings live in the wrapper.

Parameters
Name Type Description
func Callable[..., Any]
strict bool Default:False
Returns
tuple[dict[str, Any], tuple[str, ...]]
_type_to_schema 4 dict[str, Any] ▼
Convert Python type annotation to JSON Schema fragment. *_defs* accumulates ``…
def _type_to_schema(annotation: Any, _seen: set[int] | None = None, _defs: dict[str, dict[str, Any]] | None = None, *, _strict: bool = False) -> dict[str, Any]

Convert Python type annotation to JSON Schema fragment.

_defs accumulates$defsentries for recursive dataclasses so they can be emitted as{"$ref": "#/$defs/ClassName"}.

When _strict is True, unrecognized types raise TypeError.

Parameters
Name Type Description
annotation Any
_seen set[int] | None Default:None
_defs dict[str, dict[str, Any]] | None Default:None
_strict bool Default:False
Returns
dict[str, Any]
_is_typed_dict 1 bool ▼
Check if annotation is a TypedDict subclass.
def _is_typed_dict(annotation: Any) -> bool
Parameters
Name Type Description
annotation Any
Returns
bool
_is_optional 1 bool ▼
Check if annotation is X | None.
def _is_optional(annotation: Any) -> bool
Parameters
Name Type Description
annotation Any
Returns
bool
_unwrap_optional 1 Any ▼
Extract non-None type from X | None.
def _unwrap_optional(annotation: Any) -> Any
Parameters
Name Type Description
annotation Any
Returns
Any
return_to_schema 1 dict[str, Any] | None ▼
Generate JSON Schema from function return type annotation. Returns None if the…
def return_to_schema(func: Callable[..., Any]) -> dict[str, Any] | None

Generate JSON Schema from function return type annotation.

Returns None if the function has no return annotation or returns None.

Parameters
Name Type Description
func Callable[..., Any]
Returns
dict[str, Any] | None
_is_context_type 2 bool ▼
Check if an annotation refers to milo's Context type.
def _is_context_type(annotation: Any, name: str) -> bool
Parameters
Name Type Description
annotation Any
name str
Returns
bool
_parse_param_docs 1 dict[str, str] ▼
Extract parameter descriptions from a docstring. Supports Google, Sphinx, and …
def _parse_param_docs(docstring: str) -> dict[str, str]

Extract parameter descriptions from a docstring.

Supports Google, Sphinx, and NumPy docstring styles::

# Google style
Parameters
Name Type Description
docstring str
Returns
dict[str, str]