Bug: Optional[int] type hint incorrectly generates string type in JSON Schema #2354
Description
Describe the bug
Description
When using Optional[int] type hints in MCP tool functions, the awslabs.mcp_lambda_handler package incorrectly generates JSON Schema with type: "string" instead of type: "integer".
Environment
- Package:
awslabs.mcp_lambda_handler - Python version: 3.x
- Type hint:
Optional[int](which isUnion[int, None])
Expected Behavior
When a function parameter is annotated with Optional[int], the generated JSON Schema should be:
{
"type": "integer",
"description": "..."
}Or with nullable support:
{
"type": ["integer", "null"],
"description": "..."
}Current Behavior
The generated JSON Schema incorrectly uses string type:
{
"type": "string",
"description": "..."
}Reproduction Steps
from typing import Optional
from awslabs.mcp_lambda_handler import MCPLambdaHandler
mcp = MCPLambdaHandler(
name="test-mcp",
version="1.0.0",
)
@mcp.tool()
def test_int_param(days: Optional[int] = 30, status: Optional[str] = "new") -> str:
"""Test integer parameter
Args:
days: Number of days, default 30
status: Status, default "new"
Returns:
Test result
"""
return f"days={days} (type: {type(days).__name__}), status={status}"
# Check generated schema
import json
print(json.dumps(mcp.tools, indent=2))Output:
{
"test_int_param": {
"name": "test_int_param",
"description": "Test integer parameter",
"inputSchema": {
"type": "object",
"properties": {
"days": {
"type": "string", // ❌ Should be "integer"
"description": "Number of days, default 30"
},
"status": {
"type": "string", // ✅ Correct
"description": "Status, default \"new\""
}
},
"required": ["days", "status"]
}
}
}Function annotations (correct):
{
'days': typing.Optional[int],
'status': typing.Optional[str],
'return': <class 'str'>
}Possible Solution
Add handling for Union types before the default case:
from typing import Union, get_origin, get_args
def get_type_schema(type_hint: Any) -> Dict[str, Any]:
# Handle basic types
if type_hint is int:
return {'type': 'integer'}
elif type_hint is float:
return {'type': 'number'}
elif type_hint is bool:
return {'type': 'boolean'}
elif type_hint is str:
return {'type': 'string'}
# Get origin type
origin = get_origin(type_hint)
# Handle Union types (including Optional)
if origin is Union:
args = get_args(type_hint)
# Filter out NoneType
non_none_args = [arg for arg in args if arg is not type(None)]
if len(non_none_args) == 1:
# This is Optional[T], return schema for T
return get_type_schema(non_none_args[0])
else:
# Multiple non-None types in Union
# Could use anyOf, but for simplicity, use the first type
return get_type_schema(non_none_args[0])
if origin is None:
return {'type': 'string'} # Default for unknown types
# ... rest of the code ...Additional Information/Context
This issue was discovered when integrating with MCP clients that validate parameter types against the JSON Schema. The clients correctly send integer values, but the schema indicates they should be strings, causing confusion and requiring workarounds.
OS
Lambda
Server
lambda-mcp-server
Server Version
0.1.9
Region experiencing the issue
cn-north-1
Other information
No response
Service quota
- I have reviewed the service quotas for this construct