Ensure LC tools work

pyproject.toml

@@ -51,6 +51,8 @@ tests = [
     "pytest>=7.0",
     "pytest-timeout",
     "pytest-xdist",
+    "langchain_community",
+    "duckduckgo-search",
 ]
 dev = [
     "controlflow[tests]",

src/controlflow/agents/agent.py

@@ -20,7 +20,12 @@
 from controlflow.instructions import get_instructions
 from controlflow.llm.messages import AIMessage, BaseMessage
 from controlflow.llm.rules import LLMRules
-from controlflow.tools.tools import handle_tool_call, handle_tool_call_async
+from controlflow.tools.tools import (
+    as_lc_tools,
+    as_tools,
+    handle_tool_call,
+    handle_tool_call_async,
+)
 from controlflow.utilities.context import ctx
 from controlflow.utilities.types import ControlFlowModel
 
@@ -156,7 +161,7 @@ def get_model(self, tools: list["Tool"] = None) -> BaseChatModel:
                 f"Agent {self.name}: No model provided and no default model could be loaded."
             )
         if tools:
-            model = model.bind_tools([t.to_lc_tool() for t in tools])
+            model = model.bind_tools(as_lc_tools(tools))
         return model
 
     def get_llm_rules(self) -> LLMRules:
@@ -165,7 +170,7 @@ def get_llm_rules(self) -> LLMRules:
         """
         return controlflow.llm.rules.rules_for_model(self.get_model())
 
-    def get_tools(self) -> list[Callable]:
+    def get_tools(self) -> list["Tool"]:
         from controlflow.tools.talk_to_user import talk_to_user
 
         tools = self.tools.copy()
@@ -174,7 +179,7 @@ def get_tools(self) -> list[Callable]:
         if self.memory is not None:
             tools.extend(self.memory.get_tools())
 
-        return tools
+        return as_tools(tools)
 
     @contextmanager
     def create_context(self):

src/controlflow/orchestration/agent_context.py

@@ -1,6 +1,6 @@
 from contextlib import ExitStack
 from functools import partial, wraps
-from typing import Callable, Optional
+from typing import Any, Callable, Optional
 
 from pydantic import Field
 
@@ -11,7 +11,7 @@
 from controlflow.llm.messages import BaseMessage
 from controlflow.orchestration.handler import Handler
 from controlflow.tasks.task import Task
-from controlflow.tools.tools import Tool
+from controlflow.tools.tools import Tool, as_tools
 from controlflow.utilities.context import ctx
 from controlflow.utilities.types import ControlFlowModel
 
@@ -30,7 +30,7 @@ class AgentContext(ControlFlowModel):
     model_config = dict(arbitrary_types_allowed=True)
     flow: Flow
     tasks: list[Task]
-    tools: list[Tool] = []
+    tools: list[Any] = []
     agents: list[BaseAgent] = Field(
         default_factory=list,
         description="Any other agents that are relevant to this operation, in order to properly load events",
@@ -59,7 +59,7 @@ def add_handlers(self, handlers: list[Handler]):
         self.handlers = self.handlers + handlers
 
     def add_tools(self, tools: list[Tool]):
-        self.tools = self.tools + tools
+        self.tools = self.tools + as_tools(tools)
 
     def add_instructions(self, instructions: list[str]):
         self.instructions = self.instructions + instructions

src/controlflow/settings.py

@@ -96,7 +96,7 @@ class Settings(ControlFlowSettings):
     )
 
     tools_verbose: bool = Field(
-        default=False, description="If True, tools will log additional information."
+        default=True, description="If True, tools will log additional information."
     )
 
     # ------------ Prefect settings ------------

src/controlflow/tools/tools.py

@@ -209,10 +209,27 @@ def as_tools(
         else:
             raise ValueError(f"Invalid tool: {t}")
 
-        if (t.name, t.description, t.fn) in seen:
+        if (t.name, t.description) in seen:
             continue
         new_tools.append(t)
-        seen.add((t.name, t.description, t.fn))
+        seen.add((t.name, t.description))
+    return new_tools
+
+
+def as_lc_tools(
+    tools: list[Union[Callable, langchain_core.tools.BaseTool, Tool]],
+) -> list[langchain_core.tools.BaseTool]:
+    new_tools = []
+    for t in tools:
+        if isinstance(t, langchain_core.tools.BaseTool):
+            pass
+        elif isinstance(t, Tool):
+            t = t.to_lc_tool()
+        elif inspect.isfunction(t):
+            t = langchain_core.tools.StructuredTool.from_function(t)
+        else:
+            raise ValueError(f"Invalid tool: {t}")
+        new_tools.append(t)
     return new_tools
 
 
@@ -266,6 +283,8 @@ def handle_tool_call(
                 fn_output = tool.run(input=fn_args)
             elif isinstance(tool, langchain_core.tools.BaseTool):
                 fn_output = tool.invoke(input=fn_args)
+            else:
+                raise ValueError(f"Invalid tool: {tool}")
         except Exception as exc:
             fn_output = f'Error calling function "{fn_name}": {exc}'
             is_error = True
@@ -307,6 +326,8 @@ async def handle_tool_call_async(tool_call: ToolCall, tools: list[Tool]) -> Any:
                 fn_output = await tool.run_async(input=fn_args)
             elif isinstance(tool, langchain_core.tools.BaseTool):
                 fn_output = await tool.ainvoke(input=fn_args)
+            else:
+                raise ValueError(f"Invalid tool: {tool}")
         except Exception as exc:
             fn_output = f'Error calling function "{fn_name}": {exc}'
             is_error = True

tests/test_settings.py

@@ -8,9 +8,10 @@
 
 
 def test_defaults():
-    # ensure that debug settings etc. are not left on by default
+    # ensure that debug settings etc. are not misconfigured during development
+    # change these settings to match whatever the default should be
     assert controlflow.settings.tools_raise_on_error is False
-    assert controlflow.settings.tools_verbose is False
+    assert controlflow.settings.tools_verbose is True
 
 
 def test_temporary_settings():
@@ -35,65 +36,84 @@ def test_prefect_settings_apply_at_runtime(caplog):
 
 
 def test_import_without_default_api_key_warns_but_does_not_fail(monkeypatch, caplog):
-    # remove the OPENAI_API_KEY environment variable
-    monkeypatch.delenv("OPENAI_API_KEY", raising=False)
-
-    # Clear any previous logs
-    caplog.clear()
-
-    # Import the library
-    with caplog.at_level("WARNING"):
-        # Reload the library to apply changes
+    try:
+        with monkeypatch.context() as m:
+            # remove the OPENAI_API_KEY environment variable
+            m.delenv("OPENAI_API_KEY", raising=False)
+
+            # Clear any previous logs
+            caplog.clear()
+
+            # Import the library
+            with caplog.at_level("WARNING"):
+                # Reload the library to apply changes
+                defaults_module = importlib.import_module("controlflow.defaults")
+                importlib.reload(defaults_module)
+                importlib.reload(controlflow)
+
+            # Check if the warning was logged
+            assert any(
+                record.levelname == "WARNING"
+                and "The default LLM model could not be created" in record.message
+                for record in caplog.records
+            ), "The expected warning was not logged"
+    finally:
         defaults_module = importlib.import_module("controlflow.defaults")
-        importlib.reload(controlflow)
         importlib.reload(defaults_module)
-
-    # Check if the warning was logged
-    assert any(
-        record.levelname == "WARNING"
-        and "The default LLM model could not be created" in record.message
-        for record in caplog.records
-    ), "The expected warning was not logged"
+        importlib.reload(controlflow)
 
 
 def test_import_without_default_api_key_errors_when_loading_model(monkeypatch):
-    # remove the OPENAI_API_KEY environment variable
-    monkeypatch.delenv("OPENAI_API_KEY", raising=False)
-
-    # Reload the library to apply changes
-    defaults_module = importlib.import_module("controlflow.defaults")
-    importlib.reload(controlflow)
-    importlib.reload(defaults_module)
-
-    with pytest.raises(ValueError, match="Did not find openai_api_key"):
-        controlflow.llm.models.get_default_model()
-
-    with pytest.raises(
-        ValueError, match="No model provided and no default model could be loaded"
-    ):
-        controlflow.Agent().get_model()
+    try:
+        with monkeypatch.context() as m:
+            # remove the OPENAI_API_KEY environment variable
+            m.delenv("OPENAI_API_KEY", raising=False)
+
+            # Reload the library to apply changes
+            defaults_module = importlib.import_module("controlflow.defaults")
+            importlib.reload(defaults_module)
+            importlib.reload(controlflow)
+
+            with pytest.raises(ValueError, match="Did not find openai_api_key"):
+                controlflow.llm.models.get_default_model()
+
+            with pytest.raises(
+                ValueError,
+                match="No model provided and no default model could be loaded",
+            ):
+                controlflow.Agent().get_model()
+    finally:
+        defaults_module = importlib.import_module("controlflow.defaults")
+        importlib.reload(defaults_module)
+        importlib.reload(controlflow)
 
 
 def test_import_without_api_key_for_non_default_model_warns_but_does_not_fail(
     monkeypatch, caplog
 ):
-    # remove the OPENAI_API_KEY environment variable
-    monkeypatch.delenv("OPENAI_API_KEY", raising=False)
-    monkeypatch.setenv("CONTROLFLOW_LLM_MODEL", "anthropic/not-a-model")
-
-    # Clear any previous logs
-    caplog.clear()
-
-    # Import the library
-    with caplog.at_level("WARNING"):
-        # Reload the library to apply changes
+    try:
+        with monkeypatch.context() as m:
+            # remove the OPENAI_API_KEY environment variable
+            m.delenv("OPENAI_API_KEY", raising=False)
+            m.setenv("CONTROLFLOW_LLM_MODEL", "anthropic/not-a-model")
+
+            # Clear any previous logs
+            caplog.clear()
+
+            # Import the library
+            with caplog.at_level("WARNING"):
+                # Reload the library to apply changes
+                defaults_module = importlib.import_module("controlflow.defaults")
+                importlib.reload(defaults_module)
+                importlib.reload(controlflow)
+
+            # Check if the warning was logged
+            assert any(
+                record.levelname == "WARNING"
+                and "The default LLM model could not be created" in record.message
+                for record in caplog.records
+            ), "The expected warning was not logged"
+    finally:
         defaults_module = importlib.import_module("controlflow.defaults")
-        importlib.reload(controlflow)
         importlib.reload(defaults_module)
-
-    # Check if the warning was logged
-    assert any(
-        record.levelname == "WARNING"
-        and "The default LLM model could not be created" in record.message
-        for record in caplog.records
-    ), "The expected warning was not logged"
+        importlib.reload(controlflow)

tests/tools/test_lc_tools.py

@@ -0,0 +1,13 @@
+import controlflow
+from langchain_community.tools import DuckDuckGoSearchRun
+
+
+def test_ddg_tool():
+    task = controlflow.Task(
+        "Retrieve and summarize today's two top business headlines",
+        tools=[DuckDuckGoSearchRun()],
+        # agent=summarizer,
+        result_type=list[str],
+    )
+    task.run()
+    assert task.is_successful()