import pytest
import datetime
from unittest.mock import MagicMock
from promptsmith.main import recommend_strategy, generate_meta_prompt
from promptsmith.domain.selector import TaskAnalysis, TaskCategory, ComplexityLevel
from promptsmith.domain.builder import GeneratedPrompt
# In a real E2E test, we might use a separate process or the fastmcp test client.
# Since fastmcp's test client details aren't fully exposed here, we will simulate
# the "User Scenario" by invoking the tools in sequence as an agent would,
# but still mocking the internal logic to keep it deterministic (Component-Level E2E).
class TestWorkflowScenario:
def test_complete_reasoning_workflow(self, override_deps):
"""
Scenario: User wants to generate a prompt for a complex coding task.
1. User asks for strategy recommendation.
2. User accepts recommendation and asks for the prompt.
3. User provides feedback on the result.
"""
# ------------------------------------------------------------------
# Step 1: Recommend Strategy
# ------------------------------------------------------------------
task_text = "Refactor the legacy authentication system to use OAuth2."
context_text = "Must support Google and GitHub login."
# Mock analysis result
override_deps.selector.analyze.return_value = TaskAnalysis(
task=task_text,
category=TaskCategory.CODE,
complexity_score=7.5,
complexity_level=ComplexityLevel.HIGH,
recommended_framework="chain_of_thought",
reasoning="Authenticaion is complex.",
alternative_frameworks=["tree_of_thoughts"]
)
strategy = recommend_strategy(task=task_text, context=context_text)
assert strategy["category"] == "code"
assert strategy["recommended_framework"] == "chain_of_thought"
# In a real scenario, the Agent would "read" this output.
# ------------------------------------------------------------------
# Step 2: Generate Meta-Prompt
# ------------------------------------------------------------------
# Mock builder result
override_deps.builder.build.return_value = GeneratedPrompt(
task=task_text,
framework_used="chain_of_thought",
meta_prompt="Context: ... Instructions: Think step-by-step..."
)
# Mock storage creation (simulate DB returning an ID)
mock_log = MagicMock()
mock_log.id = "task-uuid-1234"
override_deps.storage.create_log.return_value = mock_log
result = generate_meta_prompt(
task=task_text,
context=context_text,
framework="chain_of_thought" # User chooses the recommended one
)
assert result["task_id"] == "task-uuid-1234"
assert "Think step-by-step" in result["meta_prompt"]
# ------------------------------------------------------------------
# Step 3: Log Feedback (Post-Execution)
# ------------------------------------------------------------------
# Mock storage update
# update_feedback returns the log if successful, None if failed
override_deps.storage.update_feedback.return_value = MagicMock(execution_feedback="Great prompt!")
# We need to import this tool locally as it wasn't imported in top level
from promptsmith.main import log_execution_feedback
feedback_result = log_execution_feedback(
task_id="task-uuid-1234",
feedback="Great prompt! The agent handled the oauth flow perfectly."
)
assert feedback_result["success"] is True
# Verify the entire chain of calls
override_deps.selector.analyze.assert_called()
override_deps.builder.build.assert_called()
override_deps.storage.create_log.assert_called_once()
override_deps.storage.update_feedback.assert_called_with("task-uuid-1234", "Great prompt! The agent handled the oauth flow perfectly.")
