tests/test_env.py

"""Tests for the FinePrint environment."""

import sys
from pathlib import Path

sys.path.insert(0, str(Path(__file__).resolve().parent.parent))

import pytest
import numpy as np
from fineprint.env import FinePrintEnv, ACTION_TYPES


@pytest.fixture
def env():
    policies_path = str(Path(__file__).resolve().parent.parent / "policies")
    e = FinePrintEnv(policies_dir=policies_path)
    yield e
    e.close()


class TestReset:
    def test_reset_returns_obs_and_info(self, env):
        obs, info = env.reset(seed=42)
        assert isinstance(obs, dict)
        assert isinstance(info, dict)

    def test_reset_obs_keys(self, env):
        obs, _ = env.reset(seed=42)
        expected_keys = {
            "current_workflow",
            "current_step",
            "workflow_progress",
            "user_message",
            "conversation_history",
            "user_satisfaction",
            "agent_believed_version",
            "cached_policies",
            "steps_since_last_verify",
            "system_notification",
            "contradiction_detected",
            "user_expressed_confusion",
            "last_action_compliant",
            "last_compliance_note",
        }
        assert set(obs.keys()) == expected_keys

    def test_reset_initial_values(self, env):
        obs, info = env.reset(seed=42)
        assert obs["agent_believed_version"] == "v1_base"
        assert obs["steps_since_last_verify"][0] == 0
        assert obs["user_satisfaction"][0] == 1.0
        assert obs["contradiction_detected"] == 0
        assert obs["user_expressed_confusion"] == 0

    def test_reset_info_contains_version(self, env):
        _, info = env.reset(seed=42)
        assert info["active_version"] == "v1_base"
        assert info["total_versions"] == 8

    def test_reset_reproducibility(self, env):
        obs1, _ = env.reset(seed=42)
        obs2, _ = env.reset(seed=42)
        assert obs1["current_workflow"] == obs2["current_workflow"]


class TestStep:
    def test_step_returns_5_tuple(self, env):
        env.reset(seed=42)
        result = env.step({"action_type": 5, "message": "Hello"})
        assert len(result) == 5
        obs, reward, terminated, truncated, info = result
        assert isinstance(obs, dict)
        assert isinstance(reward, float)
        assert isinstance(terminated, bool)
        assert isinstance(truncated, bool)
        assert isinstance(info, dict)

    def test_step_without_reset_raises(self, env):
        with pytest.raises(RuntimeError):
            env.step({"action_type": 5})

    def test_request_verification(self, env):
        env.reset(seed=42)
        obs, reward, _, _, info = env.step({"action_type": 0})
        assert info["compliance"]["compliant"] is True
        assert "Verified" in info["compliance"]["reason"]

    def test_quote_policy_correct(self, env):
        env.reset(seed=42)
        # First verify to ensure cache is fresh
        env.step({"action_type": 0})
        # Quote a known v1_base value
        obs, reward, _, _, info = env.step({
            "action_type": 1,
            "policy_field": "return.window_days",
            "quoted_value": "30",
        })
        assert info["compliance"]["compliant"] is True
        assert reward > 0

    def test_quote_policy_incorrect(self, env):
        env.reset(seed=42)
        obs, reward, _, _, info = env.step({
            "action_type": 1,
            "policy_field": "return.window_days",
            "quoted_value": "999",
        })
        assert info["compliance"]["compliant"] is False

    def test_episode_terminates(self, env):
        env.reset(seed=42)
        done = False
        steps = 0
        while not done and steps < 100:
            _, _, terminated, truncated, _ = env.step(
                {"action_type": 5, "message": "OK"}
            )
            done = terminated or truncated
            steps += 1
        assert done

    def test_info_contains_metrics(self, env):
        env.reset(seed=42)
        _, _, _, _, info = env.step({"action_type": 5, "message": "Hi"})
        assert "compliance_failures" in info
        assert "drift_detections" in info
        assert "user_satisfaction" in info
        assert "active_version" in info
        assert "agent_version" in info


class TestRender:
    def test_render_ansi(self, env):
        env.reset(seed=42)
        output = env.render(mode="ansi")
        assert isinstance(output, str)
        assert "FINEPRINT" in output

    def test_render_before_reset(self, env):
        result = env.render(mode="ansi")
        assert result is None


class TestClose:
    def test_close(self, env):
        env.reset(seed=42)
        env.close()
        assert env.state is None