langchain-robocorp: Fix parsing of Union types (such as Optional). (#22277)

libs/partners/robocorp/langchain_robocorp/_common.py

@@ -1,5 +1,6 @@
+import time
 from dataclasses import dataclass
-from typing import Any, Dict, List, Tuple, Union
+from typing import Any, Dict, List, Set, Tuple, Union
 
 from langchain_core.pydantic_v1 import BaseModel, Field, create_model
 from langchain_core.utils.json_schema import dereference_refs
@@ -93,26 +94,45 @@ def get_schema(endpoint_spec: dict) -> dict:
     )
 
 
-def create_field(schema: dict, required: bool) -> Tuple[Any, Any]:
+def create_field(
+    schema: dict, required: bool, created_model_names: Set[str]
+) -> Tuple[Any, Any]:
     """
     Creates a Pydantic field based on the schema definition.
     """
+    if "anyOf" in schema:
+        field_types = [
+            create_field(sub_schema, required, created_model_names)[0]
+            for sub_schema in schema["anyOf"]
+        ]
+        if len(field_types) == 1:
+            field_type = field_types[0]  # Simplified handling
+        else:
+            field_type = Union[tuple(field_types)]
+    else:
         field_type = type_mapping.get(schema.get("type", "string"), str)
+
     description = schema.get("description", "")
 
     # Handle nested objects
-    if schema["type"] == "object":
+    if schema.get("type") == "object":
         nested_fields = {
-            k: create_field(v, k in schema.get("required", []))
+            k: create_field(v, k in schema.get("required", []), created_model_names)
             for k, v in schema.get("properties", {}).items()
         }
-        model_name = schema.get("title", "NestedModel")
+        model_name = schema.get("title", f"NestedModel{time.time()}")
+        if model_name in created_model_names:
+            # needs to be unique
+            model_name = model_name + str(time.time())
         nested_model = create_model(model_name, **nested_fields)  # type: ignore
+        created_model_names.add(model_name)
         return nested_model, Field(... if required else None, description=description)
 
     # Handle arrays
-    elif schema["type"] == "array":
+    elif schema.get("type") == "array":
-        item_type, _ = create_field(schema["items"], required=True)
+        item_type, _ = create_field(
+            schema["items"], required=True, created_model_names=created_model_names
+        )
         return List[item_type], Field(  # type: ignore
             ... if required else None, description=description
         )
@@ -128,9 +148,10 @@ def get_param_fields(endpoint_spec: dict) -> dict:
     required_fields = schema.get("required", [])
 
     fields = {}
+    created_model_names: Set[str] = set()
     for key, value in properties.items():
         is_required = key in required_fields
-        field_info = create_field(value, is_required)
+        field_info = create_field(value, is_required, created_model_names)
         fields[key] = field_info
 
     return fields

libs/partners/robocorp/tests/unit_tests/test_toolkits.py

@@ -3,7 +3,10 @@ import json
 from pathlib import Path
 from unittest.mock import MagicMock, patch
 
-from langchain_core.utils.function_calling import convert_to_openai_function
+from langchain_core.utils.function_calling import (
+    convert_to_openai_function,
+    convert_to_openai_tool,
+)
 
 from langchain_robocorp.toolkits import ActionServerToolkit
 
@@ -118,3 +121,66 @@ Strictly adhere to the schema."""
             ],
         }
         assert params["properties"]["rows_to_add"] == expected
+
+
+def test_get_tools_with_complex_inputs() -> None:
+    toolkit_instance = ActionServerToolkit(
+        url="http://example.com", api_key="dummy_key"
+    )
+
+    fixture_path = Path(__file__).with_name("_openapi3.fixture.json")
+
+    with patch(
+        "langchain_robocorp.toolkits.requests.get"
+    ) as mocked_get, fixture_path.open("r") as f:
+        data = json.load(f)  # Using json.load directly on the file object
+        mocked_response = MagicMock()
+        mocked_response.json.return_value = data
+        mocked_response.status_code = 200
+        mocked_response.headers = {"Content-Type": "application/json"}
+        mocked_get.return_value = mocked_response
+
+        # Execute
+        tools = toolkit_instance.get_tools()
+        assert len(tools) == 4
+
+        tool = tools[0]
+        assert tool.name == "create_event"
+        assert tool.description == "Creates a new event in the specified calendar."
+
+        all_tools_as_openai_tools = [convert_to_openai_tool(t) for t in tools]
+        openai_tool_spec = all_tools_as_openai_tools[0]["function"]
+
+        assert isinstance(
+            openai_tool_spec, dict
+        ), "openai_func_spec should be a dictionary."
+        assert set(openai_tool_spec.keys()) == {
+            "description",
+            "name",
+            "parameters",
+        }, "Top-level keys mismatch."
+
+        assert openai_tool_spec["description"] == tool.description
+        assert openai_tool_spec["name"] == tool.name
+
+        assert isinstance(
+            openai_tool_spec["parameters"], dict
+        ), "Parameters should be a dictionary."
+
+        params = openai_tool_spec["parameters"]
+        assert set(params.keys()) == {
+            "type",
+            "properties",
+            "required",
+        }, "Parameters keys mismatch."
+        assert params["type"] == "object", "`type` in parameters should be 'object'."
+        assert isinstance(
+            params["properties"], dict
+        ), "`properties` should be a dictionary."
+        assert isinstance(params["required"], list), "`required` should be a list."
+
+        assert set(params["required"]) == {
+            "event",
+        }, "Required fields mismatch."
+
+        assert set(params["properties"].keys()) == {"calendar_id", "event"}