unique_toolkit 1.16.0__tar.gz → 1.16.1__tar.gz

{unique_toolkit-1.16.0 → unique_toolkit-1.16.1}/CHANGELOG.md

@@ -5,6 +5,9 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/), 
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [1.16.1] - 2025-10-16
+- Update debug info for better tool call tracking
+
 ## [1.16.0] - 2025-10-16
 - Add responses api support.
 - Add utilities for code execution.

{unique_toolkit-1.16.0 → unique_toolkit-1.16.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: unique_toolkit
-Version: 1.16.0
+Version: 1.16.1
 Summary: 
 License: Proprietary
 Author: Cedric Klinkert
@@ -118,6 +118,9 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/), 
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [1.16.1] - 2025-10-16
+- Update debug info for better tool call tracking
+
 ## [1.16.0] - 2025-10-16
 - Add responses api support.
 - Add utilities for code execution.

{unique_toolkit-1.16.0 → unique_toolkit-1.16.1}/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "unique_toolkit"
-version = "1.16.0"
+version = "1.16.1"
 description = ""
 authors = [
     "Cedric Klinkert <cedric.klinkert@unique.ch>",

unique_toolkit-1.16.1/unique_toolkit/agentic/debug_info_manager/debug_info_manager.py

@@ -0,0 +1,28 @@
+from typing import Any
+
+from unique_toolkit.agentic.tools.schemas import ToolCallResponse
+
+
+class DebugInfoManager:
+    def __init__(self):
+        self.debug_info = {"tools": []}
+
+    def extract_tool_debug_info(
+        self,
+        tool_call_responses: list[ToolCallResponse],
+        loop_iteration_index: int | None = None,
+    ):
+        for tool_call_response in tool_call_responses:
+            tool_info = {
+                "name": tool_call_response.name,
+                "info": tool_call_response.debug_info,
+            }
+            if loop_iteration_index is not None:
+                tool_info["info"]["loop_iteration"] = loop_iteration_index
+            self.debug_info["tools"].append(tool_info)
+
+    def add(self, key: str, value: Any) -> None:
+        self.debug_info = self.debug_info | {key: value}
+
+    def get(self) -> dict[str, Any]:
+        return self.debug_info

unique_toolkit-1.16.1/unique_toolkit/agentic/debug_info_manager/test/test_debug_info_manager.py

@@ -0,0 +1,278 @@
+"""
+Test suite for DebugInfoManager class.
+
+This test suite validates the DebugInfoManager's ability to:
+1. Initialize with empty debug info
+2. Extract tool debug info from ToolCallResponse objects
+3. Handle loop iteration indices
+4. Add arbitrary key-value pairs to debug info
+5. Retrieve the complete debug info dictionary
+"""
+
+from unique_toolkit.agentic.debug_info_manager.debug_info_manager import (
+    DebugInfoManager,
+)
+from unique_toolkit.agentic.tools.schemas import ToolCallResponse
+
+
+class TestDebugInfoManager:
+    """Test suite for DebugInfoManager functionality."""
+
+    def test_init__initializes_empty_debug_info__on_creation(self):
+        """Test that DebugInfoManager initializes with empty tools list."""
+        manager = DebugInfoManager()
+
+        assert manager.debug_info == {"tools": []}
+        assert manager.get() == {"tools": []}
+
+    def test_extract_tool_debug_info__adds_single_tool__with_valid_response(self):
+        """Test extracting debug info from a single ToolCallResponse."""
+        manager = DebugInfoManager()
+        tool_call_response = ToolCallResponse(
+            id="tool_1",
+            name="TestTool",
+            debug_info={"execution_time": "100ms", "status": "success"},
+        )
+
+        manager.extract_tool_debug_info([tool_call_response])
+
+        debug_info = manager.get()
+        assert len(debug_info["tools"]) == 1
+        assert debug_info["tools"][0]["name"] == "TestTool"
+        assert debug_info["tools"][0]["info"]["execution_time"] == "100ms"
+        assert debug_info["tools"][0]["info"]["status"] == "success"
+
+    def test_extract_tool_debug_info__adds_multiple_tools__with_multiple_responses(
+        self,
+    ):
+        """Test extracting debug info from multiple ToolCallResponse objects."""
+        manager = DebugInfoManager()
+        tool_call_responses = [
+            ToolCallResponse(
+                id="tool_1",
+                name="SearchTool",
+                debug_info={"query": "test query", "results": 5},
+            ),
+            ToolCallResponse(
+                id="tool_2",
+                name="CalculatorTool",
+                debug_info={"operation": "add", "result": 42},
+            ),
+            ToolCallResponse(
+                id="tool_3",
+                name="WeatherTool",
+                debug_info={"location": "New York", "temperature": "72F"},
+            ),
+        ]
+
+        manager.extract_tool_debug_info(tool_call_responses)
+
+        debug_info = manager.get()
+        assert len(debug_info["tools"]) == 3
+        assert debug_info["tools"][0]["name"] == "SearchTool"
+        assert debug_info["tools"][1]["name"] == "CalculatorTool"
+        assert debug_info["tools"][2]["name"] == "WeatherTool"
+
+    def test_extract_tool_debug_info__preserves_order__with_sequential_calls(self):
+        """Test that multiple calls to extract_tool_debug_info preserve order."""
+        manager = DebugInfoManager()
+
+        # First call
+        manager.extract_tool_debug_info(
+            [ToolCallResponse(id="tool_1", name="Tool1", debug_info={"step": 1})]
+        )
+
+        # Second call
+        manager.extract_tool_debug_info(
+            [ToolCallResponse(id="tool_2", name="Tool2", debug_info={"step": 2})]
+        )
+
+        # Third call
+        manager.extract_tool_debug_info(
+            [ToolCallResponse(id="tool_3", name="Tool3", debug_info={"step": 3})]
+        )
+
+        debug_info = manager.get()
+        assert len(debug_info["tools"]) == 3
+        assert debug_info["tools"][0]["info"]["step"] == 1
+        assert debug_info["tools"][1]["info"]["step"] == 2
+        assert debug_info["tools"][2]["info"]["step"] == 3
+
+    def test_extract_tool_debug_info__adds_loop_iteration__when_index_provided(self):
+        """Test that loop_iteration_index is added to debug info when provided."""
+        manager = DebugInfoManager()
+        tool_call_response = ToolCallResponse(
+            id="tool_1", name="IterativeTool", debug_info={"status": "processing"}
+        )
+
+        manager.extract_tool_debug_info([tool_call_response], loop_iteration_index=3)
+
+        debug_info = manager.get()
+        assert debug_info["tools"][0]["info"]["loop_iteration"] == 3
+        assert debug_info["tools"][0]["info"]["status"] == "processing"
+
+    def test_extract_tool_debug_info__omits_loop_iteration__when_index_is_none(self):
+        """Test that loop_iteration is not added when index is None."""
+        manager = DebugInfoManager()
+        tool_call_response = ToolCallResponse(
+            id="tool_1", name="SingleRunTool", debug_info={"status": "complete"}
+        )
+
+        manager.extract_tool_debug_info([tool_call_response], loop_iteration_index=None)
+
+        debug_info = manager.get()
+        assert "loop_iteration" not in debug_info["tools"][0]["info"]
+        assert debug_info["tools"][0]["info"]["status"] == "complete"
+
+    def test_extract_tool_debug_info__handles_empty_debug_info__gracefully(self):
+        """Test extracting from ToolCallResponse with empty debug_info dict."""
+        manager = DebugInfoManager()
+        tool_call_response = ToolCallResponse(
+            id="tool_1", name="MinimalTool", debug_info={}
+        )
+
+        manager.extract_tool_debug_info([tool_call_response])
+
+        debug_info = manager.get()
+        assert len(debug_info["tools"]) == 1
+        assert debug_info["tools"][0]["name"] == "MinimalTool"
+        assert debug_info["tools"][0]["info"] == {}
+
+    def test_extract_tool_debug_info__handles_empty_list__without_error(self):
+        """Test that passing an empty list doesn't cause errors."""
+        manager = DebugInfoManager()
+
+        manager.extract_tool_debug_info([])
+
+        debug_info = manager.get()
+        assert debug_info["tools"] == []
+
+    def test_add__adds_new_key_value_pair__to_debug_info(self):
+        """Test adding a new key-value pair to debug_info."""
+        manager = DebugInfoManager()
+
+        manager.add("execution_summary", {"total_time": "500ms", "total_calls": 5})
+
+        debug_info = manager.get()
+        assert "execution_summary" in debug_info
+        assert debug_info["execution_summary"]["total_time"] == "500ms"
+        assert debug_info["execution_summary"]["total_calls"] == 5
+
+    def test_add__preserves_tools_list__when_adding_new_keys(self):
+        """Test that add() preserves the tools list."""
+        manager = DebugInfoManager()
+        manager.extract_tool_debug_info(
+            [
+                ToolCallResponse(
+                    id="tool_1", name="TestTool", debug_info={"test": "data"}
+                )
+            ]
+        )
+
+        manager.add("metadata", {"version": "1.0"})
+
+        debug_info = manager.get()
+        assert len(debug_info["tools"]) == 1
+        assert debug_info["tools"][0]["name"] == "TestTool"
+        assert debug_info["metadata"]["version"] == "1.0"
+
+    def test_add__overwrites_existing_key__when_key_exists(self):
+        """Test that add() overwrites an existing key."""
+        manager = DebugInfoManager()
+        manager.add("status", "in_progress")
+        manager.add("status", "completed")
+
+        debug_info = manager.get()
+        assert debug_info["status"] == "completed"
+
+    def test_add__adds_multiple_keys__with_sequential_calls(self):
+        """Test adding multiple key-value pairs with sequential calls."""
+        manager = DebugInfoManager()
+
+        manager.add("key1", "value1")
+        manager.add("key2", {"nested": "value2"})
+        manager.add("key3", [1, 2, 3])
+
+        debug_info = manager.get()
+        assert debug_info["key1"] == "value1"
+        assert debug_info["key2"]["nested"] == "value2"
+        assert debug_info["key3"] == [1, 2, 3]
+
+    def test_get__returns_complete_debug_info__with_mixed_data(self):
+        """Test get() returns complete debug info with tools and custom keys."""
+        manager = DebugInfoManager()
+
+        # Add tool debug info
+        manager.extract_tool_debug_info(
+            [ToolCallResponse(id="tool_1", name="Tool1", debug_info={"data": "test"})],
+            loop_iteration_index=0,
+        )
+
+        # Add custom keys
+        manager.add("start_time", "2025-10-16T10:00:00")
+        manager.add("end_time", "2025-10-16T10:01:00")
+
+        debug_info = manager.get()
+
+        assert "tools" in debug_info
+        assert "start_time" in debug_info
+        assert "end_time" in debug_info
+        assert len(debug_info["tools"]) == 1
+        assert debug_info["start_time"] == "2025-10-16T10:00:00"
+
+    def test_integration__complete_workflow__with_all_operations(self):
+        """Integration test: complete workflow using all DebugInfoManager methods."""
+        manager = DebugInfoManager()
+
+        # Initial state
+        assert manager.get() == {"tools": []}
+
+        # Add some metadata
+        manager.add("session_id", "abc-123")
+        manager.add("user_id", "user-456")
+
+        # First tool call (loop iteration 0)
+        manager.extract_tool_debug_info(
+            [
+                ToolCallResponse(
+                    id="tool_1",
+                    name="SearchTool",
+                    debug_info={"query": "AI research", "hits": 100},
+                )
+            ],
+            loop_iteration_index=0,
+        )
+
+        # Second tool call (loop iteration 1)
+        manager.extract_tool_debug_info(
+            [
+                ToolCallResponse(
+                    id="tool_2",
+                    name="AnalysisTool",
+                    debug_info={"processed": 50, "relevant": 10},
+                ),
+                ToolCallResponse(
+                    id="tool_3",
+                    name="SummaryTool",
+                    debug_info={"paragraphs": 3, "words": 250},
+                ),
+            ],
+            loop_iteration_index=1,
+        )
+
+        # Add final summary
+        manager.add("summary", {"total_tools": 3, "total_iterations": 2})
+
+        # Verify complete debug info
+        debug_info = manager.get()
+
+        assert debug_info["session_id"] == "abc-123"
+        assert debug_info["user_id"] == "user-456"
+        assert len(debug_info["tools"]) == 3
+        assert debug_info["tools"][0]["name"] == "SearchTool"
+        assert debug_info["tools"][0]["info"]["loop_iteration"] == 0
+        assert debug_info["tools"][1]["name"] == "AnalysisTool"
+        assert debug_info["tools"][1]["info"]["loop_iteration"] == 1
+        assert debug_info["tools"][2]["name"] == "SummaryTool"
+        assert debug_info["tools"][2]["info"]["loop_iteration"] == 1
+        assert debug_info["summary"]["total_tools"] == 3

{unique_toolkit-1.16.0 → unique_toolkit-1.16.1}/unique_toolkit/agentic/tools/tool_manager.py

@@ -62,6 +62,14 @@ class BaseToolManager(ABC):
     def get_tool_by_name(self, name: str) -> Tool | None:
         raise NotImplementedError()
 
+    @abstractmethod
+    def get_tool_choices(self) -> list[str]:
+        raise NotImplementedError()
+
+    @abstractmethod
+    def get_exclusive_tools(self) -> list[str]:
+        raise NotImplementedError()
+
     def does_a_tool_take_control(self, tool_calls: list[LanguageModelFunction]) -> bool:
         for tool_call in tool_calls:
             tool_instance = self.get_tool_by_name(tool_call.name)
@@ -121,6 +129,14 @@ class BaseToolManager(ABC):
             unpacked_tool_call_result = self._create_tool_call_response(
                 result, tool_calls[i]
             )
+            if unpacked_tool_call_result.debug_info is None:
+                unpacked_tool_call_result.debug_info = {}
+            unpacked_tool_call_result.debug_info["is_exclusive"] = (
+                tool_calls[i].name in self.get_exclusive_tools()
+            )
+            unpacked_tool_call_result.debug_info["is_forced"] = (
+                tool_calls[i].name in self.get_tool_choices()
+            )
             tool_call_results_unpacked.append(unpacked_tool_call_result)
 
         return tool_call_results_unpacked
@@ -230,6 +246,9 @@ class ToolManager(BaseToolManager):
         self._tools = []
         self._tool_choices = event.payload.tool_choices
         self._disabled_tools = event.payload.disabled_tools
+        self._exclusive_tools = [
+            tool.name for tool in self._config.tools if tool.is_exclusive
+        ]
         # this needs to be a set of strings to avoid duplicates
         self._tool_evaluation_check_list: set[EvaluationMetricName] = set()
         self._mcp_manager = mcp_manager
@@ -299,6 +318,14 @@ class ToolManager(BaseToolManager):
                 return tool
         return None
 
+    @override
+    def get_tool_choices(self) -> list[str]:
+        return self._tool_choices
+
+    @override
+    def get_exclusive_tools(self) -> list[str]:
+        return self._exclusive_tools
+
     def get_tools(self) -> list[Tool]:
         return self._tools  # type: ignore
 
@@ -393,6 +420,14 @@ class ResponsesApiToolManager(BaseToolManager):
     def get_tool_by_name(self, name: str) -> Tool | None:
         return self._tool_manager.get_tool_by_name(name)
 
+    @override
+    def get_tool_choices(self) -> list[str]:
+        return self._tool_manager._tool_choices
+
+    @override
+    def get_exclusive_tools(self) -> list[str]:
+        return self._tool_manager._exclusive_tools
+
     @property
     def sub_agents(self) -> list[SubAgentTool]:
         return self._tool_manager.sub_agents

unique_toolkit-1.16.0/unique_toolkit/agentic/debug_info_manager/debug_info_manager.py

@@ -1,18 +0,0 @@
-from unique_toolkit.agentic.tools.schemas import ToolCallResponse
-
-
-class DebugInfoManager:
-    def __init__(self):
-        self.debug_info = {"tools": []}
-
-    def extract_tool_debug_info(self, tool_call_responses: list[ToolCallResponse]):
-        for tool_call_response in tool_call_responses:
-            self.debug_info["tools"].append(
-                {"name": tool_call_response.name, "data": tool_call_response.debug_info}
-            )
-
-    def add(self, key, value):
-        self.debug_info = self.debug_info | {key: value}
-
-    def get(self):
-        return self.debug_info