kagent-adk 0.6.10__tar.gz → 0.6.12__tar.gz

{kagent_adk-0.6.10 → kagent_adk-0.6.12}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: kagent-adk
-Version: 0.6.10
+Version: 0.6.12
 Summary: kagent-adk is an sdk for integrating adk agents with kagent
 Requires-Python: >=3.12.11
 Requires-Dist: a2a-sdk>=0.3.1

{kagent_adk-0.6.10 → kagent_adk-0.6.12}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "kagent-adk"
-version = "0.6.10"
+version = "0.6.12"
 description = "kagent-adk is an sdk for integrating adk agents with kagent"
 readme = "README.md"
 requires-python = ">=3.12.11"

kagent_adk-0.6.12/src/kagent/adk/converters/error_mappings.py

@@ -0,0 +1,60 @@
+"""Error code to user-friendly message mappings for ADK events.
+
+This module provides mappings from Google GenAI finish reasons to user-friendly
+error messages, excluding STOP which is a normal completion reason.
+"""
+
+from typing import Dict, Optional
+
+from google.genai import types as genai_types
+
+# Error code to user-friendly message mappings
+# Based on Google GenAI types.py FinishReason enum (excluding STOP)
+ERROR_CODE_MESSAGES: Dict[str, str] = {
+    # Length and token limits
+    genai_types.FinishReason.MAX_TOKENS: "Response was truncated due to maximum token limit. Try asking a shorter question or breaking it into parts.",
+    # Safety and content filtering
+    genai_types.FinishReason.SAFETY: "Response was blocked due to safety concerns. Please rephrase your request to avoid potentially harmful content.",
+    genai_types.FinishReason.RECITATION: "Response was blocked due to unauthorized citations. Please rephrase your request.",
+    genai_types.FinishReason.BLOCKLIST: "Response was blocked due to restricted terminology. Please rephrase your request using different words.",
+    genai_types.FinishReason.PROHIBITED_CONTENT: "Response was blocked due to prohibited content. Please rephrase your request.",
+    genai_types.FinishReason.SPII: "Response was blocked due to sensitive personal information concerns. Please avoid including personal details.",
+    # Function calling errors
+    genai_types.FinishReason.MALFORMED_FUNCTION_CALL: "The agent generated an invalid function call. This may be due to complex input data. Try rephrasing your request or breaking it into simpler steps.",
+    # Generic fallback
+    genai_types.FinishReason.OTHER: "An unexpected error occurred during processing. Please try again or rephrase your request.",
+}
+
+# Normal completion reasons that should not be treated as errors
+NORMAL_COMPLETION_REASONS = {
+    genai_types.FinishReason.STOP,  # Normal completion
+}
+
+# Default error message when no specific mapping exists
+DEFAULT_ERROR_MESSAGE = "An error occurred during processing"
+
+
+def _get_error_message(error_code: Optional[str]) -> str:
+    """Get a user-friendly error message for the given error code.
+
+    Args:
+        error_code: The error code from the ADK event (e.g., finish_reason)
+
+    Returns:
+        User-friendly error message string
+    """
+
+    # Return mapped message or default
+    return ERROR_CODE_MESSAGES.get(error_code, DEFAULT_ERROR_MESSAGE)
+
+
+def _is_normal_completion(error_code: Optional[str]) -> bool:
+    """Check if the error code represents normal completion rather than an error.
+
+    Args:
+        error_code: The error code to check
+
+    Returns:
+        True if this is a normal completion reason, False otherwise
+    """
+    return error_code in NORMAL_COMPLETION_REASONS

{kagent_adk-0.6.10 → kagent_adk-0.6.12}/src/kagent/adk/converters/event_converter.py

@@ -20,6 +20,7 @@ from kagent.core.a2a import (
     get_kagent_metadata_key,
 )
 
+from .error_mappings import _get_error_message, _is_normal_completion
 from .part_converter import (
     convert_genai_part_to_a2a_part,
 )
@@ -27,7 +28,6 @@ from .part_converter import (
 # Constants
 
 ARTIFACT_ID_SEPARATOR = "-"
-DEFAULT_ERROR_MESSAGE = "An error occurred during processing"
 
 # Logger
 logger = logging.getLogger("kagent_adk." + __name__)
@@ -191,13 +191,16 @@ def _create_error_status_event(
     Returns:
       A TaskStatusUpdateEvent with FAILED state.
     """
-    error_message = getattr(event, "error_message", None) or DEFAULT_ERROR_MESSAGE
+    error_message = getattr(event, "error_message", None)
 
     # Get context metadata and add error code
     event_metadata = _get_context_metadata(event, invocation_context)
     if event.error_code:
         event_metadata[get_kagent_metadata_key("error_code")] = str(event.error_code)
 
+        if not error_message:
+            error_message = _get_error_message(event.error_code)
+
     return TaskStatusUpdateEvent(
         task_id=task_id,
         context_id=context_id,
@@ -298,7 +301,7 @@ def convert_event_to_a2a_events(
 
     try:
         # Handle error scenarios
-        if event.error_code:
+        if event.error_code and not _is_normal_completion(event.error_code):
             error_event = _create_error_status_event(event, invocation_context, task_id, context_id)
             a2a_events.append(error_event)
 

{kagent_adk-0.6.10 → kagent_adk-0.6.12}/src/kagent/adk/models/_openai.py

@@ -9,6 +9,7 @@ from typing import TYPE_CHECKING, Any, AsyncGenerator, Iterable, Literal, Option
 from google.adk.models import BaseLlm
 from google.adk.models.llm_response import LlmResponse
 from google.genai import types
+from google.genai.types import FunctionCall, FunctionResponse
 from openai import AsyncAzureOpenAI, AsyncOpenAI
 from openai.types.chat import (
     ChatCompletion,
@@ -27,7 +28,7 @@ from openai.types.chat.chat_completion_message_tool_call_param import (
 from openai.types.chat.chat_completion_message_tool_call_param import (
     Function as ToolCallFunction,
 )
-from openai.types.shared_params import FunctionDefinition
+from openai.types.shared_params import FunctionDefinition, FunctionParameters
 from pydantic import Field
 
 if TYPE_CHECKING:
@@ -56,7 +57,7 @@ def _convert_content_to_openai_messages(
         messages.append(system_message)
 
     # First pass: collect all function responses to match with tool calls
-    all_function_responses = {}
+    all_function_responses: dict[str, FunctionResponse] = {}
     for content in contents:
         for part in content.parts or []:
             if part.function_response:
@@ -67,18 +68,18 @@ def _convert_content_to_openai_messages(
         role = _convert_role_to_openai(content.role)
 
         # Separate different types of parts
-        text_parts = []
-        function_calls = []
-        function_responses = []
+        text_parts: list[str] = []
+        function_calls: list[FunctionCall] = []
+        function_responses: list[FunctionResponse] = []
         image_parts = []
 
         for part in content.parts or []:
             if part.text:
                 text_parts.append(part.text)
             elif part.function_call:
-                function_calls.append(part)
+                function_calls.append(part.function_call)
             elif part.function_response:
-                function_responses.append(part)
+                function_responses.append(part.function_response)
             elif part.inline_data and part.inline_data.mime_type and part.inline_data.mime_type.startswith("image"):
                 if part.inline_data.data:
                     image_data = base64.b64encode(part.inline_data.data).decode()
@@ -98,43 +99,43 @@ def _convert_content_to_openai_messages(
 
             for func_call in function_calls:
                 tool_call_function: ToolCallFunction = {
-                    "name": func_call.function_call.name or "",
-                    "arguments": str(func_call.function_call.args) if func_call.function_call.args else "{}",
-                }
-                tool_call_id = func_call.function_call.id or "call_1"
-                tool_call: ChatCompletionMessageToolCallParam = {
-                    "id": tool_call_id,
-                    "type": "function",
-                    "function": tool_call_function,
+                    "name": func_call.name or "",
+                    "arguments": json.dumps(func_call.args) if func_call.args else "{}",
                 }
+                tool_call_id = func_call.id or "call_1"
+                tool_call = ChatCompletionMessageToolCallParam(
+                    id=tool_call_id,
+                    type="function",
+                    function=tool_call_function,
+                )
                 tool_calls.append(tool_call)
 
                 # Check if we have a response for this tool call
                 if tool_call_id in all_function_responses:
                     func_response = all_function_responses[tool_call_id]
-                    tool_message: ChatCompletionToolMessageParam = {
-                        "role": "tool",
-                        "tool_call_id": tool_call_id,
-                        "content": str(func_response.response.get("result", "")) if func_response.response else "",
-                    }
+                    tool_message = ChatCompletionToolMessageParam(
+                        role="tool",
+                        tool_call_id=tool_call_id,
+                        content=str(func_response.response.get("result", "")) if func_response.response else "",
+                    )
                     tool_response_messages.append(tool_message)
                 else:
                     # If no response is available, create a placeholder response
                     # This prevents the OpenAI API error
-                    tool_message: ChatCompletionToolMessageParam = {
-                        "role": "tool",
-                        "tool_call_id": tool_call_id,
-                        "content": "No response available for this function call.",
-                    }
+                    tool_message = ChatCompletionToolMessageParam(
+                        role="tool",
+                        tool_call_id=tool_call_id,
+                        content="No response available for this function call.",
+                    )
                     tool_response_messages.append(tool_message)
 
             # Create assistant message with tool calls
             text_content = "\n".join(text_parts) if text_parts else None
-            assistant_message: ChatCompletionAssistantMessageParam = {
-                "role": "assistant",
-                "content": text_content,
-                "tool_calls": tool_calls,
-            }
+            assistant_message = ChatCompletionAssistantMessageParam(
+                role="assistant",
+                content=text_content,
+                tool_calls=tool_calls,
+            )
             messages.append(assistant_message)
 
             # Add all tool response messages immediately after the assistant message
@@ -145,22 +146,22 @@ def _convert_content_to_openai_messages(
             if role == "user":
                 if image_parts and text_parts:
                     # Multi-modal content
-                    text_part: ChatCompletionContentPartTextParam = {"type": "text", "text": "\n".join(text_parts)}
+                    text_part = ChatCompletionContentPartTextParam(type="text", text="\n".join(text_parts))
                     content_parts = [text_part] + image_parts
-                    user_message: ChatCompletionUserMessageParam = {"role": "user", "content": content_parts}
+                    user_message = ChatCompletionUserMessageParam(role="user", content=content_parts)
                 elif image_parts:
                     # Image only
-                    user_message: ChatCompletionUserMessageParam = {"role": "user", "content": image_parts}
+                    user_message = ChatCompletionUserMessageParam(role="user", content=image_parts)
                 else:
                     # Text only
-                    user_message: ChatCompletionUserMessageParam = {"role": "user", "content": "\n".join(text_parts)}
+                    user_message = ChatCompletionUserMessageParam(role="user", content="\n".join(text_parts))
                 messages.append(user_message)
             elif role == "assistant":
                 # Assistant messages with text (no tool calls)
-                assistant_message: ChatCompletionAssistantMessageParam = {
-                    "role": "assistant",
-                    "content": "\n".join(text_parts),
-                }
+                assistant_message = ChatCompletionAssistantMessageParam(
+                    role="assistant",
+                    content="\n".join(text_parts),
+                )
                 messages.append(assistant_message)
 
     return messages
@@ -197,10 +198,10 @@ def _convert_tools_to_openai(tools: list[types.Tool]) -> list[ChatCompletionTool
         if tool.function_declarations:
             for func_decl in tool.function_declarations:
                 # Build function definition
-                function_def: FunctionDefinition = {
-                    "name": func_decl.name or "",
-                    "description": func_decl.description or "",
-                }
+                function_def = FunctionDefinition(
+                    name=func_decl.name or "",
+                    description=func_decl.description or "",
+                )
 
                 # Always include parameters field, even if empty
                 properties = {}
@@ -219,7 +220,7 @@ def _convert_tools_to_openai(tools: list[types.Tool]) -> list[ChatCompletionTool
                 function_def["parameters"] = {"type": "object", "properties": properties, "required": required}
 
                 # Create the tool param
-                openai_tool: ChatCompletionToolParam = {"type": "function", "function": function_def}
+                openai_tool = ChatCompletionToolParam(type="function", function=function_def)
                 openai_tools.append(openai_tool)
 
     return openai_tools
@@ -389,6 +390,7 @@ class AzureOpenAI(BaseOpenAI):
     api_version: Optional[str] = None
     azure_endpoint: Optional[str] = None
     azure_deployment: Optional[str] = None
+    headers: Optional[dict[str, str]] = None
 
     @cached_property
     def _client(self) -> AsyncAzureOpenAI:
@@ -407,4 +409,8 @@ class AzureOpenAI(BaseOpenAI):
                 "API key must be provided either via api_key parameter or AZURE_OPENAI_API_KEY environment variable"
             )
 
-        return AsyncAzureOpenAI(api_version=api_version, azure_endpoint=azure_endpoint, api_key=api_key)
+        default_headers = self.headers or {}
+
+        return AsyncAzureOpenAI(
+            api_version=api_version, azure_endpoint=azure_endpoint, api_key=api_key, default_headers=default_headers
+        )

{kagent_adk-0.6.10 → kagent_adk-0.6.12}/src/kagent/adk/types.py

@@ -36,6 +36,7 @@ class RemoteAgentConfig(BaseModel):
 
 class BaseLLM(BaseModel):
     model: str
+    headers: dict[str, str] | None = None
 
 
 class OpenAI(BaseLLM):
@@ -97,20 +98,26 @@ class AgentConfig(BaseModel):
                     agent_card=f"{remote_agent.url}/{AGENT_CARD_WELL_KNOWN_PATH}",
                     description=remote_agent.description,
                 )
-                mcp_toolsets.append(AgentTool(agent=remote_agent, skip_summarization=True))
+                mcp_toolsets.append(
+                    AgentTool(agent=remote_agent, skip_summarization=True)
+                )  # Get headers from model config
+
+        extra_headers = self.model.headers or {}
 
         if self.model.type == "openai":
             model = OpenAINative(model=self.model.model, base_url=self.model.base_url, type="openai")
         elif self.model.type == "anthropic":
-            model = LiteLlm(model=f"anthropic/{self.model.model}", base_url=self.model.base_url)
+            model = LiteLlm(
+                model=f"anthropic/{self.model.model}", base_url=self.model.base_url, extra_headers=extra_headers
+            )
         elif self.model.type == "gemini_vertex_ai":
             model = GeminiLLM(model=self.model.model)
         elif self.model.type == "gemini_anthropic":
             model = ClaudeLLM(model=self.model.model)
         elif self.model.type == "ollama":
-            model = LiteLlm(model=f"ollama_chat/{self.model.model}")
+            model = LiteLlm(model=f"ollama_chat/{self.model.model}", extra_headers=extra_headers)
         elif self.model.type == "azure_openai":
-            model = OpenAIAzure(model=self.model.model, type="azure_openai")
+            model = OpenAIAzure(model=self.model.model, type="azure_openai", headers=extra_headers)
         elif self.model.type == "gemini":
             model = self.model.model
         else:

kagent_adk-0.6.12/tests/unittests/converters/test_event_converter.py

@@ -0,0 +1,119 @@
+from unittest.mock import Mock
+
+import pytest
+from a2a.types import TaskState, TaskStatusUpdateEvent
+from google.genai import types as genai_types
+
+from kagent.adk.converters.event_converter import convert_event_to_a2a_events
+from kagent.core.a2a import get_kagent_metadata_key
+
+
+def _create_mock_invocation_context():
+    """Create a mock invocation context for testing."""
+    context = Mock()
+    context.app_name = "test_app"
+    context.user_id = "test_user"
+    context.session.id = "test_session"
+    return context
+
+
+def _create_mock_event(error_code=None, content=None, invocation_id="test_invocation", author="test_author"):
+    """Create a mock event for testing."""
+    event = Mock()
+    event.error_code = error_code
+    event.content = content
+    event.invocation_id = invocation_id
+    event.author = author
+    event.branch = None
+    event.grounding_metadata = None
+    event.custom_metadata = None
+    event.usage_metadata = None
+    event.error_message = None
+    return event
+
+
+class TestEventConverter:
+    """Test cases for event converter functions."""
+
+    def test_convert_event_to_a2a_events(self):
+        """Test that STOP error codes with empty content don't create any events, while actual error codes create error events."""
+
+        invocation_context = _create_mock_invocation_context()
+
+        # Test case 1: Empty content with STOP error code
+        event1 = _create_mock_event(
+            error_code=genai_types.FinishReason.STOP, content=None, invocation_id="test_invocation_1"
+        )
+        result1 = convert_event_to_a2a_events(
+            event1, invocation_context, task_id="test_task_1", context_id="test_context_1"
+        )
+        error_events1 = [
+            e for e in result1 if isinstance(e, TaskStatusUpdateEvent) and e.status.state == TaskState.failed
+        ]
+        working_events1 = [
+            e for e in result1 if isinstance(e, TaskStatusUpdateEvent) and e.status.state == TaskState.working
+        ]
+        assert len(error_events1) == 0, (
+            f"Expected no error events for STOP with empty content, got {len(error_events1)}"
+        )
+        assert len(working_events1) == 0, (
+            f"Expected no working events for STOP with empty content (no content to convert), got {len(working_events1)}"
+        )
+
+        # Test case 2: Empty parts with STOP error code
+        content_mock = Mock()
+        content_mock.parts = []
+        event2 = _create_mock_event(
+            error_code=genai_types.FinishReason.STOP, content=content_mock, invocation_id="test_invocation_2"
+        )
+        result2 = convert_event_to_a2a_events(
+            event2, invocation_context, task_id="test_task_2", context_id="test_context_2"
+        )
+        error_events2 = [
+            e for e in result2 if isinstance(e, TaskStatusUpdateEvent) and e.status.state == TaskState.failed
+        ]
+        working_events2 = [
+            e for e in result2 if isinstance(e, TaskStatusUpdateEvent) and e.status.state == TaskState.working
+        ]
+        assert len(error_events2) == 0, f"Expected no error events for STOP with empty parts, got {len(error_events2)}"
+        assert len(working_events2) == 0, (
+            f"Expected no working events for STOP with empty parts (no content to convert), got {len(working_events2)}"
+        )
+
+        # Test case 3: Missing content with STOP error code
+        event3 = _create_mock_event(
+            error_code=genai_types.FinishReason.STOP, content=None, invocation_id="test_invocation_3"
+        )
+        result3 = convert_event_to_a2a_events(
+            event3, invocation_context, task_id="test_task_3", context_id="test_context_3"
+        )
+        error_events3 = [
+            e for e in result3 if isinstance(e, TaskStatusUpdateEvent) and e.status.state == TaskState.failed
+        ]
+        working_events3 = [
+            e for e in result3 if isinstance(e, TaskStatusUpdateEvent) and e.status.state == TaskState.working
+        ]
+        assert len(error_events3) == 0, (
+            f"Expected no error events for STOP with missing content, got {len(error_events3)}"
+        )
+        assert len(working_events3) == 0, (
+            f"Expected no working events for STOP with missing content (no content to convert), got {len(working_events3)}"
+        )
+
+        # Test case 4: Actual error code should create error event
+        event4 = _create_mock_event(
+            error_code=genai_types.FinishReason.MALFORMED_FUNCTION_CALL, content=None, invocation_id="test_invocation_4"
+        )
+        result4 = convert_event_to_a2a_events(
+            event4, invocation_context, task_id="test_task_4", context_id="test_context_4"
+        )
+        error_events4 = [
+            e for e in result4 if isinstance(e, TaskStatusUpdateEvent) and e.status.state == TaskState.failed
+        ]
+        assert len(error_events4) == 1, f"Expected 1 error event for MALFORMED_FUNCTION_CALL, got {len(error_events4)}"
+
+        # Check that the error event has the correct error code in metadata
+        error_event = error_events4[0]
+        error_code_key = get_kagent_metadata_key("error_code")
+        assert error_code_key in error_event.metadata
+        assert error_event.metadata[error_code_key] == str(genai_types.FinishReason.MALFORMED_FUNCTION_CALL)