PR feedback

haystack/agents/base.py

@@ -206,6 +206,19 @@ def extract_tool_name_and_tool_input(self, llm_response: str) -> Tuple[Optional[
         return None, None
 
 
+def react_parameter_resolver(query: str, agent: Agent, agent_step: AgentStep, **kwargs) -> Dict[str, Any]:
+    """
+    A parameter resolver for ReAct based agents that returns the query, the tool names, the tool names
+    with descriptions, and the transcript (internal monologue).
+    """
+    return {
+        "query": query,
+        "tool_names": agent.tm.get_tool_names(),
+        "tool_names_with_descriptions": agent.tm.get_tool_names_with_descriptions(),
+        "transcript": agent_step.transcript,
+    }
+
+
 class Agent:
     """
     An Agent answers queries using the tools you give to it. The tools are pipelines or nodes. The Agent uses a large
@@ -268,14 +281,6 @@ def __init__(
                 f"Prompt template '{prompt_template}' not found. Please check the spelling of the template name."
             )
         self.prompt_template = resolved_prompt_template
-        react_parameter_resolver: Callable[
-            [str, Agent, AgentStep, Dict[str, Any]], Dict[str, Any]
-        ] = lambda query, agent, agent_step, **kwargs: {
-            "query": query,
-            "tool_names": agent.tm.get_tool_names(),
-            "tool_names_with_descriptions": agent.tm.get_tool_names_with_descriptions(),
-            "transcript": agent_step.transcript,
-        }
         self.prompt_parameters_resolver = (
             prompt_parameters_resolver if prompt_parameters_resolver else react_parameter_resolver
         )

haystack/agents/conversational.py

@@ -1,7 +1,7 @@
 from typing import Optional, Callable, Union
 
 from haystack.agents import Agent
-from haystack.agents.base import ToolsManager
+from haystack.agents.base import ToolsManager, react_parameter_resolver
 from haystack.agents.memory import Memory, ConversationMemory, ConversationSummaryMemory
 from haystack.nodes import PromptNode, PromptTemplate
 
@@ -130,11 +130,6 @@ def __init__(
             memory=memory if memory else ConversationSummaryMemory(prompt_node),
             prompt_parameters_resolver=prompt_parameters_resolver
             if prompt_parameters_resolver
-            else lambda query, agent, agent_step, **kwargs: {
-                "query": query,
-                "tool_names": agent.tm.get_tool_names(),
-                "tool_names_with_descriptions": agent.tm.get_tool_names_with_descriptions(),
-                "transcript": agent_step.transcript,
-            },
+            else react_parameter_resolver,
             final_answer_pattern=final_answer_pattern,
         )

test/agents/test_conversational_agent_with_tools.py

@@ -1,7 +1,7 @@
 import pytest
-from unittest.mock import MagicMock
+from unittest.mock import MagicMock, Mock, call
 
-from haystack.agents import Tool
+from haystack.agents import Tool, Agent, AgentStep
 from haystack.agents.conversational import ConversationalAgentWithTools
 from haystack.agents.memory import ConversationSummaryMemory, NoMemory
 from haystack.nodes import PromptNode
@@ -49,9 +49,33 @@ def test_agent_with_memory(prompt_node):
 
 @pytest.mark.unit
 def test_run(prompt_node):
+    """
+    Test that the invocation of ConversationalAgentWithTools run method in turn invokes _step of the Agent superclass
+    Make sure that the agent is starting from the correct step 1, and max_steps is 5
+    """
+    mock_step = Mock(spec=Agent._step)
+
+    # Replace the original _step method with the mock
+    Agent._step = mock_step
+
+    # Initialize agent
+    prompt_node = PromptNode()
     agent = ConversationalAgentWithTools(prompt_node)
 
-    # Mock the Agent run method
-    agent.run = MagicMock(return_value="Hello")
-    assert agent.run("query") == "Hello"
-    agent.run.assert_called_once_with("query")
+    # Run agent
+    agent.run(query="query")
+
+    assert mock_step.call_count == 1
+
+    # Check the parameters passed to _step method
+    assert mock_step.call_args[0][0] == "query"
+    agent_step = mock_step.call_args[0][1]
+    expected_agent_step = AgentStep(
+        current_step=1,
+        max_steps=5,
+        prompt_node_response="",
+        final_answer_pattern=r"Final Answer\s*:\s*(.*)",
+        transcript="",
+    )
+    # compare the string representation of the objects
+    assert str(agent_step) == str(expected_agent_step)