feat: add end to end integration test

pyproject.toml

@@ -142,6 +142,10 @@ test = [
     "hatch test --cover --cov-report term --cov-report html --cov-report xml {args}"
 ]
 
+test-integ = [
+    "hatch test tests-integ {args}"
+]
+
 [tool.mypy]
 python_version = "3.10"
 warn_return_any = true
@@ -160,7 +164,7 @@ ignore_missing_imports = false
 
 [tool.ruff]
 line-length = 120
-include = ["src/**/*.py", "tests/**/*.py", "tools/**/*.py"]
+include = ["src/**/*.py", "tests/**/*.py", "tools/**/*.py","tests-integ/**/*.py"]
 
 [tool.ruff.lint]
 select = [

tests-integ/end_to_end_test.py

@@ -0,0 +1,70 @@
+import re
+
+import pytest
+from strands.agent import Agent
+from strands.models import BedrockModel
+
+
+@pytest.fixture
+def bedrock_model():
+    return BedrockModel(
+        model_id="us.anthropic.claude-3-7-sonnet-20250219-v1:0",
+        streaming=True,
+    )
+
+
+def extract_code_from_response(response):
+    """Extract Python code blocks from the agent's response."""
+    code_pattern = r"```python\s*(.*?)\s*```"
+    matches = re.findall(code_pattern, str(response), re.DOTALL)
+    if matches:
+        return matches[0].strip()
+    return None
+
+
+def test_agent_creates_calculator_tool(tmp_path, monkeypatch, bedrock_model):
+    """End-to-end test: Ask agent to create a tool, validate it, then delete it."""
+    tools_dir = tmp_path / "tools"
+    tools_dir.mkdir(parents=True)
+    monkeypatch.setenv("STRANDS_TOOLS_DIR", str(tools_dir))
+
+    tool_path = tools_dir / "calculator_tool.py"
+
+    try:
+        creator_agent = Agent(model=bedrock_model)
+        prompt = """
+        Create a Python tool named calculator_tool that can perform basic arithmetic operations.
+        The tool should:
+        1. Take two numbers and an operator ('+', '-', '*', '/')
+        2. Return the result of the operation
+        3. Include proper docstrings and error handling
+        4. Be compatible with the Strands SDK tool system
+        
+        Provide only the Python code with no additional explanation.
+        """
+        creation_response = creator_agent(prompt)
+
+        tool_code = extract_code_from_response(creation_response)
+        assert tool_code is not None, "Failed to extract tool code from agent response"
+
+        with open(tool_path, "w") as f:
+            f.write(tool_code)
+        assert tool_path.exists(), f"Tool file was not created at {tool_path}"
+
+        agent = Agent(model=bedrock_model, tools=None, load_tools_from_directory=True)
+
+        test_cases = [
+            ("What is 1 + 2?", "3"),
+            ("Calculate 2 - 1", "1"),
+            ("What is 1 multiplied by 5?", "5"),
+            ("Divide 20 by 4", "5"),
+        ]
+
+        for question, expected in test_cases:
+            response = agent(question)
+            response_str = str(response).lower()
+            assert expected.lower() in response_str, f"Failed to find '{expected}' in response to '{question}'"
+
+    finally:
+        if tool_path.exists():
+            tool_path.unlink()