synth-ai 0.1.9__py3-none-any.whl → 0.2.1.dev0__py3-none-any.whl

synth_ai/environments/examples/minigrid/units/test_minigrid_environment.py

@@ -0,0 +1,238 @@
+"""Unit tests for MiniGrid environment."""
+
+import asyncio
+import pytest
+import json
+
+from synth_ai.environments.examples.minigrid.environment import (
+    MiniGridEnvironment,
+    MiniGridInteractTool,
+)
+from synth_ai.environments.examples.minigrid.taskset import DEFAULT_MINIGRID_TASK
+from synth_ai.environments.environment.tools import EnvToolCall
+
+
+@pytest.mark.asyncio
+async def test_environment_initialization():
+    """Test environment initialization."""
+    env = MiniGridEnvironment(DEFAULT_MINIGRID_TASK)
+
+    # Check initial setup
+    assert env.name == "MiniGridEnvironment"
+    assert env.task_instance == DEFAULT_MINIGRID_TASK
+
+    # Initialize
+    obs = await env.initialize()
+
+    # Check observation
+    assert isinstance(obs, dict)
+    assert "observation" in obs
+    assert "terminated" in obs
+    assert "total_reward" in obs
+    assert obs["terminated"] is False
+    assert obs["total_reward"] == 0.0
+
+    # Check observation text
+    obs_text = obs["observation"]
+    assert "Mission:" in obs_text
+    assert "Grid:" in obs_text
+    assert "Legend:" in obs_text
+
+
+@pytest.mark.asyncio
+async def test_environment_step():
+    """Test environment step functionality."""
+    env = MiniGridEnvironment(DEFAULT_MINIGRID_TASK)
+    await env.initialize()
+
+    # Test forward action
+    tool_call = {"name": "minigrid_act", "args": {"action": "forward"}}
+    obs = await env.step(tool_call)
+
+    assert isinstance(obs, dict)
+    assert "observation" in obs
+    assert "reward_last" in obs
+    assert obs["reward_last"] == -0.01  # Step penalty
+
+    # Test turn action
+    tool_call = {"name": "minigrid_act", "args": {"action": "right"}}
+    obs = await env.step(tool_call)
+    assert obs["reward_last"] == -0.01
+
+
+@pytest.mark.asyncio
+async def test_tool_validation():
+    """Test tool call validation."""
+    env = MiniGridEnvironment(DEFAULT_MINIGRID_TASK)
+
+    # Test different input formats
+    # Format 1: Dict with name and args
+    call1 = env.validate_tool_calls({"name": "minigrid_act", "args": {"action": "forward"}})
+    assert call1.tool == "minigrid_act"
+    assert call1.args == {"action": "forward"}
+
+    # Format 2: Dict with tool and args
+    call2 = env.validate_tool_calls({"tool": "minigrid_act", "args": {"action": "left"}})
+    assert call2.tool == "minigrid_act"
+    assert call2.args == {"action": "left"}
+
+    # Format 3: List of tool calls
+    call3 = env.validate_tool_calls([{"name": "minigrid_act", "args": {"action": "right"}}])
+    assert call3.tool == "minigrid_act"
+    assert call3.args == {"action": "right"}
+
+    # Format 4: With input field
+    call4 = env.validate_tool_calls({"name": "minigrid_act", "input": {"action": "pickup"}})
+    assert call4.tool == "minigrid_act"
+    assert call4.args == {"action": "pickup"}
+
+    # Format 5: With string input
+    call5 = env.validate_tool_calls({"name": "minigrid_act", "input": '{"action": "drop"}'})
+    assert call5.tool == "minigrid_act"
+    assert call5.args == {"action": "drop"}
+
+    # Test invalid tool name
+    with pytest.raises(ValueError, match="Unknown tool"):
+        env.validate_tool_calls({"name": "invalid_tool", "args": {"action": "forward"}})
+
+
+@pytest.mark.asyncio
+async def test_invalid_actions():
+    """Test invalid action handling."""
+    env = MiniGridEnvironment(DEFAULT_MINIGRID_TASK)
+    await env.initialize()
+
+    # Test invalid action
+    tool_call = {"name": "minigrid_act", "args": {"action": "invalid_action"}}
+    obs = await env.step(tool_call)
+
+    # Should have error in observation
+    assert "error" in obs
+    assert "Invalid action" in obs["error"]
+
+
+@pytest.mark.asyncio
+async def test_checkpoint_and_terminate():
+    """Test checkpoint and terminate functionality."""
+    env = MiniGridEnvironment(DEFAULT_MINIGRID_TASK)
+    await env.initialize()
+
+    # Take some steps
+    await env.step({"name": "minigrid_act", "args": {"action": "forward"}})
+    await env.step({"name": "minigrid_act", "args": {"action": "right"}})
+
+    # Checkpoint
+    checkpoint = await env.checkpoint()
+    assert isinstance(checkpoint, dict)
+    assert "mission" in checkpoint
+    assert "total_steps" in checkpoint
+    assert "total_reward" in checkpoint
+
+    # Terminate
+    final_obs = await env.terminate()
+    assert isinstance(final_obs, dict)
+    assert "mission" in final_obs
+    assert "final_position" in final_obs
+
+
+@pytest.mark.asyncio
+async def test_interact_tool():
+    """Test the interact tool directly."""
+    from synth_ai.environments.examples.minigrid.engine import MiniGridEngine
+
+    engine = MiniGridEngine(DEFAULT_MINIGRID_TASK)
+    await engine._reset_engine()
+
+    tool = MiniGridInteractTool(engine)
+
+    # Test valid action
+    call = EnvToolCall(tool="minigrid_act", args={"action": "forward"})
+    result = await tool(call)
+
+    assert result.ok is True
+    assert "message" in result.payload
+    assert "public_state" in result.payload
+    assert "private_state" in result.payload
+
+    # Test invalid action
+    call = EnvToolCall(tool="minigrid_act", args={"action": "invalid"})
+    result = await tool(call)
+
+    assert result.ok is False
+    assert "Invalid action" in result.error
+
+
+@pytest.mark.asyncio
+async def test_observation_callables():
+    """Test observation callable functionality."""
+    from synth_ai.environments.examples.minigrid.engine import (
+        MiniGridObservationCallable,
+        MiniGridCheckpointObservationCallable,
+    )
+
+    env = MiniGridEnvironment(DEFAULT_MINIGRID_TASK)
+    await env.initialize()
+
+    # Get states
+    priv, pub = env.engine.get_current_states_for_observation()
+
+    # Test step observation
+    step_obs_callable = MiniGridObservationCallable()
+    step_obs = await step_obs_callable.get_observation(pub, priv)
+
+    assert "observation" in step_obs
+    assert "terminated" in step_obs
+    assert "reward_last" in step_obs
+
+    # Test checkpoint observation
+    ckpt_obs_callable = MiniGridCheckpointObservationCallable()
+    ckpt_obs = await ckpt_obs_callable.get_observation(pub, priv)
+
+    assert "mission" in ckpt_obs
+    assert "final_position" in ckpt_obs
+    assert "total_steps" in ckpt_obs
+
+
+@pytest.mark.asyncio
+async def test_full_episode():
+    """Test a full episode from start to finish."""
+    env = MiniGridEnvironment(DEFAULT_MINIGRID_TASK)
+
+    # Initialize
+    obs = await env.initialize()
+    assert obs["terminated"] is False
+
+    # Take multiple steps
+    actions = ["forward", "forward", "right", "forward", "forward"]
+    for action in actions:
+        obs = await env.step({"name": "minigrid_act", "args": {"action": action}})
+        if obs["terminated"]:
+            break
+
+    # Final observation should have accumulated rewards
+    assert obs["total_reward"] < 0  # Should have step penalties
+
+    # Terminate
+    final = await env.terminate()
+    assert "success" in final
+
+
+@pytest.mark.asyncio
+async def test_serialization():
+    """Test environment serialization."""
+    env = MiniGridEnvironment(DEFAULT_MINIGRID_TASK)
+    await env.initialize()
+
+    # Take some steps
+    await env.step({"name": "minigrid_act", "args": {"action": "forward"}})
+
+    # Serialize
+    serialized = await env._serialize_engine()
+
+    assert "task_instance_dict" in serialized
+    assert "engine_snapshot" in serialized
+    assert serialized["engine_snapshot"]["env_name"] == "MiniGrid-Empty-5x5-v0"
+
+
+if __name__ == "__main__":
+    asyncio.run(pytest.main([__file__, "-v"]))

synth_ai/environments/examples/minigrid/units/test_minigrid_environment_mapping.py

@@ -0,0 +1,301 @@
+"""Unit tests for MiniGrid environment mapping functionality."""
+
+import pytest
+from synth_ai.environments.examples.minigrid.environment_mapping import (
+    get_environment_from_seed,
+    get_difficulty_from_seed,
+    get_minigrid_environment,
+    get_environment_by_difficulty,
+    get_curriculum_environment,
+    validate_environment_name,
+    get_all_environments,
+    get_environments_by_difficulty,
+    ENVIRONMENT_MAPPING,
+    DIFFICULTY_MAPPING,
+)
+
+
+class TestEnvironmentMapping:
+    """Test the environment mapping functionality."""
+
+    def test_seed_to_environment_mapping(self):
+        """Test basic seed to environment mapping."""
+        # Test specific known mappings
+        assert get_environment_from_seed(0) == "MiniGrid-Empty-5x5-v0"
+        assert get_environment_from_seed(1) == "MiniGrid-Empty-6x6-v0"
+        assert get_environment_from_seed(7) == "MiniGrid-DoorKey-5x5-v0"
+        assert get_environment_from_seed(42) == "MiniGrid-KeyCorridorS6R3-v0"
+        assert get_environment_from_seed(59) == "MiniGrid-Empty-16x16-v0"
+
+        # Test modulo behavior
+        assert get_environment_from_seed(60) == get_environment_from_seed(0)
+        assert get_environment_from_seed(61) == get_environment_from_seed(1)
+        assert get_environment_from_seed(120) == get_environment_from_seed(0)
+
+        # Test negative seeds (hash behavior, not simple modulo)
+        assert get_environment_from_seed(-1) in ENVIRONMENT_MAPPING.values()
+        assert get_environment_from_seed(-5) in ENVIRONMENT_MAPPING.values()
+
+    def test_difficulty_mapping(self):
+        """Test difficulty level mapping."""
+        # Test ultra-easy (0-4)
+        assert get_difficulty_from_seed(0) == "ultra-easy"
+        assert get_difficulty_from_seed(4) == "ultra-easy"
+
+        # Test easy (5-14)
+        assert get_difficulty_from_seed(5) == "easy"
+        assert get_difficulty_from_seed(14) == "easy"
+
+        # Test medium (15-29)
+        assert get_difficulty_from_seed(15) == "medium"
+        assert get_difficulty_from_seed(29) == "medium"
+
+        # Test hard (30-44)
+        assert get_difficulty_from_seed(30) == "hard"
+        assert get_difficulty_from_seed(44) == "hard"
+
+        # Test ultra-hard (45-54)
+        assert get_difficulty_from_seed(45) == "ultra-hard"
+        assert get_difficulty_from_seed(54) == "ultra-hard"
+
+        # Test specialized (55-59)
+        assert get_difficulty_from_seed(55) == "specialized"
+        assert get_difficulty_from_seed(59) == "specialized"
+
+    def test_combined_mapping(self):
+        """Test the combined environment and difficulty mapping."""
+        test_seeds = [0, 7, 15, 30, 45, 55, 42, 1337]
+
+        for seed in test_seeds:
+            env_name = get_environment_from_seed(seed)
+            difficulty = get_difficulty_from_seed(seed)
+
+            combined_env, combined_diff = get_minigrid_environment(seed)
+
+            assert env_name == combined_env
+            assert difficulty == combined_diff
+
+    def test_environment_by_difficulty(self):
+        """Test selecting environments by difficulty level."""
+        # Test each difficulty level
+        for difficulty in DIFFICULTY_MAPPING:
+            env_name = get_environment_by_difficulty(difficulty, 0)
+            assert env_name in ENVIRONMENT_MAPPING.values()
+
+            # Test that different seeds give different environments (within range)
+            env_name2 = get_environment_by_difficulty(difficulty, 1)
+            assert env_name2 in ENVIRONMENT_MAPPING.values()
+
+    def test_environment_by_difficulty_invalid(self):
+        """Test invalid difficulty levels."""
+        with pytest.raises(ValueError, match="Unknown difficulty"):
+            get_environment_by_difficulty("invalid", 0)
+
+        with pytest.raises(ValueError, match="Unknown difficulty"):
+            get_environment_by_difficulty("super-easy", 0)
+
+    def test_curriculum_environment(self):
+        """Test curriculum-based environment selection."""
+        # Test early stage (0.0-0.2) -> ultra-easy
+        env_name, difficulty = get_curriculum_environment(0.0, 0)
+        assert difficulty == "ultra-easy"
+
+        env_name, difficulty = get_curriculum_environment(0.1, 0)
+        assert difficulty == "ultra-easy"
+
+        # Test beginning (0.2-0.4) -> easy
+        env_name, difficulty = get_curriculum_environment(0.3, 0)
+        assert difficulty == "easy"
+
+        # Test intermediate (0.4-0.6) -> medium
+        env_name, difficulty = get_curriculum_environment(0.5, 0)
+        assert difficulty == "medium"
+
+        # Test advanced (0.6-0.8) -> hard
+        env_name, difficulty = get_curriculum_environment(0.7, 0)
+        assert difficulty == "hard"
+
+        # Test expert (0.8-0.9) -> ultra-hard
+        env_name, difficulty = get_curriculum_environment(0.85, 0)
+        assert difficulty == "ultra-hard"
+
+        # Test master (0.9-1.0) -> specialized
+        env_name, difficulty = get_curriculum_environment(0.95, 0)
+        assert difficulty == "specialized"
+
+    def test_environment_validation(self):
+        """Test environment name validation."""
+        # Valid environments
+        assert validate_environment_name("MiniGrid-Empty-5x5-v0") is True
+        assert validate_environment_name("MiniGrid-FourRooms-v0") is True
+        assert validate_environment_name("MiniGrid-DoorKey-5x5-v0") is True
+
+        # Invalid environments
+        assert validate_environment_name("MiniGrid-NonExistent-v0") is False
+        assert validate_environment_name("InvalidEnv") is False
+        assert validate_environment_name("") is False
+
+    def test_get_all_environments(self):
+        """Test getting all environments."""
+        all_envs = get_all_environments()
+
+        assert len(all_envs) == 60
+        assert "MiniGrid-Empty-5x5-v0" in all_envs
+        assert "MiniGrid-FourRooms-v0" in all_envs
+        assert "MiniGrid-Empty-16x16-v0" in all_envs
+
+        # Check no duplicates
+        assert len(all_envs) == len(set(all_envs))
+
+    def test_get_environments_by_difficulty(self):
+        """Test getting environments by difficulty level."""
+        # Test each difficulty level
+        ultra_easy = get_environments_by_difficulty("ultra-easy")
+        assert len(ultra_easy) == 5  # Seeds 0-4
+
+        easy = get_environments_by_difficulty("easy")
+        assert len(easy) == 10  # Seeds 5-14
+
+        medium = get_environments_by_difficulty("medium")
+        assert len(medium) == 15  # Seeds 15-29
+
+        hard = get_environments_by_difficulty("hard")
+        assert len(hard) == 15  # Seeds 30-44
+
+        ultra_hard = get_environments_by_difficulty("ultra-hard")
+        assert len(ultra_hard) == 10  # Seeds 45-54
+
+        specialized = get_environments_by_difficulty("specialized")
+        assert len(specialized) == 5  # Seeds 55-59
+
+        # Test invalid difficulty
+        with pytest.raises(ValueError, match="Unknown difficulty"):
+            get_environments_by_difficulty("invalid")
+
+    def test_hash_seed_parameter(self):
+        """Test the hash_seed parameter for better distribution."""
+        # Test with hash_seed=True (default)
+        env1 = get_environment_from_seed(100, hash_seed=True)
+        diff1 = get_difficulty_from_seed(100, hash_seed=True)
+
+        # Test with hash_seed=False
+        env2 = get_environment_from_seed(100, hash_seed=False)
+        diff2 = get_difficulty_from_seed(100, hash_seed=False)
+
+        # Should be different due to hash distribution
+        # (Note: This might occasionally fail due to hash collisions, but very unlikely)
+        assert env1 in ENVIRONMENT_MAPPING.values()
+        assert env2 in ENVIRONMENT_MAPPING.values()
+
+        # Test consistency
+        assert get_environment_from_seed(100, hash_seed=True) == env1
+        assert get_environment_from_seed(100, hash_seed=False) == env2
+
+    def test_environment_mapping_completeness(self):
+        """Test that the environment mapping is complete."""
+        # Test all 60 environments are mapped
+        assert len(ENVIRONMENT_MAPPING) == 60
+
+        # Test all indices 0-59 are present
+        for i in range(60):
+            assert i in ENVIRONMENT_MAPPING
+
+        # Test all environments are unique
+        env_names = list(ENVIRONMENT_MAPPING.values())
+        assert len(env_names) == len(set(env_names))
+
+    def test_difficulty_mapping_completeness(self):
+        """Test that the difficulty mapping is complete."""
+        # Test all difficulty levels are mapped
+        expected_difficulties = [
+            "ultra-easy",
+            "easy",
+            "medium",
+            "hard",
+            "ultra-hard",
+            "specialized",
+        ]
+        assert set(DIFFICULTY_MAPPING.keys()) == set(expected_difficulties)
+
+        # Test ranges are correct
+        assert DIFFICULTY_MAPPING["ultra-easy"] == (0, 4)
+        assert DIFFICULTY_MAPPING["easy"] == (5, 14)
+        assert DIFFICULTY_MAPPING["medium"] == (15, 29)
+        assert DIFFICULTY_MAPPING["hard"] == (30, 44)
+        assert DIFFICULTY_MAPPING["ultra-hard"] == (45, 54)
+        assert DIFFICULTY_MAPPING["specialized"] == (55, 59)
+
+        # Test ranges are contiguous and complete
+        ranges = list(DIFFICULTY_MAPPING.values())
+        ranges.sort()
+
+        current_end = -1
+        for start, end in ranges:
+            assert start == current_end + 1
+            current_end = end
+
+        assert current_end == 59
+
+    def test_reproducibility(self):
+        """Test that seed mapping is reproducible."""
+        test_seeds = [0, 42, 1337, -5, 999]
+
+        for seed in test_seeds:
+            # Multiple calls should return same result
+            env1 = get_environment_from_seed(seed)
+            env2 = get_environment_from_seed(seed)
+            assert env1 == env2
+
+            diff1 = get_difficulty_from_seed(seed)
+            diff2 = get_difficulty_from_seed(seed)
+            assert diff1 == diff2
+
+    def test_extreme_seeds(self):
+        """Test with extreme seed values."""
+        extreme_seeds = [
+            0,
+            1,
+            -1,
+            999999,
+            -999999,
+            2**31 - 1,
+            -(2**31),  # 32-bit int limits
+            2**63 - 1,
+            -(2**63),  # 64-bit int limits
+        ]
+
+        for seed in extreme_seeds:
+            env_name = get_environment_from_seed(seed)
+            difficulty = get_difficulty_from_seed(seed)
+
+            assert env_name in ENVIRONMENT_MAPPING.values()
+            assert difficulty in DIFFICULTY_MAPPING.keys()
+
+    def test_comprehensive_mapping_coverage(self):
+        """Test that all environments in the mapping are properly categorized."""
+        # Test that every environment has proper features
+        feature_keywords = {
+            "key": ["DoorKey", "Unlock", "KeyCorridor"],
+            "door": ["Door", "Room", "Unlock"],
+            "lava": ["Lava"],
+            "memory": ["Memory"],
+            "fetch": ["Fetch"],
+            "multi_room": ["MultiRoom"],
+            "crossing": ["Crossing"],
+            "maze": ["Maze"],
+            "empty": ["Empty"],
+        }
+
+        for env_name in ENVIRONMENT_MAPPING.values():
+            # Each environment should have at least one recognizable feature
+            has_feature = False
+            for feature, keywords in feature_keywords.items():
+                if any(keyword in env_name for keyword in keywords):
+                    has_feature = True
+                    break
+
+            # Some environments might not match these patterns, that's OK
+            # This is just a sanity check, not a strict requirement
+            assert isinstance(env_name, str)
+            assert env_name.startswith("MiniGrid-")
+            assert env_name.endswith("-v0")