motief/tests/agent_tools/test_parity.py

"""Parity tests: verify agent tools can achieve what humans can.

These tests ensure the agent-native architecture satisfies the parity principle:
"Whatever the user can do through the UI/scripts, the agent can achieve through tools."
"""

import os
import pytest

pytest.importorskip("duckdb")


class TestDatabaseParity:
    """Agent database queries vs human SQL queries."""

    def test_agent_query_motions_matches_raw_sql(self, tmp_duckdb_path):
        """Human: SELECT * FROM motions LIMIT 10
        Agent: query_motions(db_path, limit=10)
        """
        import duckdb
        from agent_tools.database import query_motions

        # Human approach — handle empty DB gracefully
        con = duckdb.connect(tmp_duckdb_path)
        try:
            human_result = con.execute("SELECT * FROM motions LIMIT 10").fetchdf().to_dict("records")
        except Exception:
            human_result = []
        con.close()

        # Agent approach
        agent_result = query_motions(tmp_duckdb_path, limit=10)

        # Both should return lists
        assert isinstance(human_result, list)
        assert isinstance(agent_result, list)
        assert len(agent_result) == len(human_result)

    def test_agent_pipeline_status_matches_raw_query(self, tmp_duckdb_path):
        """Human: SELECT COUNT(*) FROM motions
        Agent: query_pipeline_status(db_path)
        """
        import duckdb
        from agent_tools.database import query_pipeline_status

        con = duckdb.connect(tmp_duckdb_path)
        try:
            human_count = con.execute("SELECT COUNT(*) FROM motions").fetchone()[0]
        except Exception:
            human_count = 0
        con.close()

        agent_status = query_pipeline_status(tmp_duckdb_path)

        assert agent_status["motion_count"] == human_count


class TestHealthCheckParity:
    """Agent health check vs human script execution."""

    def test_agent_health_check_matches_script(self, tmp_duckdb_path):
        """Human: python scripts/health_check.py
        Agent: pipeline_check_health(db_path)
        """
        from agent_tools.pipeline import pipeline_check_health

        # Agent approach
        agent_result = pipeline_check_health(tmp_duckdb_path)

        assert isinstance(agent_result, dict)
        assert "healthy" in agent_result
        assert "checks" in agent_result


class TestReportGenerationParity:
    """Agent report generation vs human manual analysis."""

    def test_agent_generates_summary_report(self, tmp_duckdb_path, tmp_path):
        """Human: Write a summary of pipeline state
        Agent: generate_report(db_path, "summary", ...)
        """
        from agent_tools.reports import generate_report

        output_path = str(tmp_path / "summary.md")
        result = generate_report(
            tmp_duckdb_path,
            report_type="summary",
            parameters={},
            output_path=output_path,
        )

        assert result["status"] == "written"
        assert os.path.exists(output_path)

        # Should contain key sections
        content = open(output_path).read()
        assert "Pipeline Summary" in content
        assert "Motions in database" in content


class TestAnalysisParity:
    """Agent analysis vs human analytical queries."""

    def test_agent_party_shift_analysis(self, tmp_duckdb_path):
        """Human: Write SQL to compare party positions across windows
        Agent: analyze_party_shift(db_path, ...)
        """
        from agent_tools.analysis import analyze_party_shift

        result = analyze_party_shift(
            tmp_duckdb_path,
            party="VVD",
            window_start="2020",
            window_end="2024",
        )

        # Should return structured result (or error if no data)
        assert isinstance(result, dict)
        assert "party" in result
        # Either shift data or error (empty DB is fine)
        assert "shift" in result or "error" in result


class TestIntegrationAgentDiagnosticLoop:
    """Integration: Agent performs full diagnostic loop."""

    def test_agent_diagnoses_stale_data(self, tmp_duckdb_path):
        """Agent loop:
        1. Check health
        2. Query pipeline status
        3. Identify issue (empty DB = no data)
        4. Suggest remediation
        """
        from agent_tools.pipeline import pipeline_check_health
        from agent_tools.database import query_pipeline_status

        # Step 1: Check health
        health = pipeline_check_health(tmp_duckdb_path)

        # Step 2: Query status
        status = query_pipeline_status(tmp_duckdb_path)

        # Step 3: Agent reasoning (simulated)
        issues = []
        if status["motion_count"] == 0:
            issues.append("No motions in database")
        if status["svd_window_count"] == 0:
            issues.append("No SVD windows computed")

        # Step 4: Suggest remediation
        suggestions = []
        if "No motions in database" in issues:
            suggestions.append("Run pipeline ingestion stage")
        if "No SVD windows computed" in issues:
            suggestions.append("Run SVD computation after ingestion")

        assert isinstance(issues, list)
        assert isinstance(suggestions, list)
        # Empty DB should produce actionable suggestions
        assert len(suggestions) > 0