langchain/libs/model-profiles/tests/unit_tests/test_cli.py

"""Tests for CLI functionality."""

import importlib.util
import warnings
from pathlib import Path
from typing import Any, get_type_hints
from unittest.mock import Mock, patch

import pytest
from langchain_core.language_models.model_profile import ModelProfile

from langchain_model_profiles.cli import (
    _model_data_to_profile,
    _warn_undeclared_profile_keys,
    refresh,
)


@pytest.fixture
def mock_models_dev_response() -> dict:
    """Create a mock response from models.dev API."""
    return {
        "anthropic": {
            "id": "anthropic",
            "name": "Anthropic",
            "models": {
                "claude-3-opus": {
                    "id": "claude-3-opus",
                    "name": "Claude 3 Opus",
                    "tool_call": True,
                    "limit": {"context": 200000, "output": 4096},
                    "modalities": {"input": ["text", "image"], "output": ["text"]},
                },
                "claude-3-sonnet": {
                    "id": "claude-3-sonnet",
                    "name": "Claude 3 Sonnet",
                    "tool_call": True,
                    "limit": {"context": 200000, "output": 4096},
                    "modalities": {"input": ["text", "image"], "output": ["text"]},
                },
            },
        },
        "openai": {
            "id": "openai",
            "name": "OpenAI",
            "models": {
                "gpt-4": {
                    "id": "gpt-4",
                    "name": "GPT-4",
                    "tool_call": True,
                    "limit": {"context": 8192, "output": 4096},
                    "modalities": {"input": ["text"], "output": ["text"]},
                }
            },
        },
    }


def test_refresh_generates_profiles_file(
    tmp_path: Path, mock_models_dev_response: dict
) -> None:
    """Test that refresh command generates _profiles.py with merged data."""
    data_dir = tmp_path / "data"
    data_dir.mkdir()

    # Create augmentations file
    aug_file = data_dir / "profile_augmentations.toml"
    aug_file.write_text("""
provider = "anthropic"

[overrides]
image_url_inputs = true
pdf_inputs = true
""")

    # Mock the httpx.get call
    mock_response = Mock()
    mock_response.json.return_value = mock_models_dev_response
    mock_response.raise_for_status = Mock()

    with (
        patch("langchain_model_profiles.cli.httpx.get", return_value=mock_response),
        patch("builtins.input", return_value="y"),
    ):
        refresh("anthropic", data_dir)

    # Verify _profiles.py was created
    profiles_file = data_dir / "_profiles.py"
    assert profiles_file.exists()

    # Import and verify content
    profiles_content = profiles_file.read_text()
    assert "DO NOT EDIT THIS FILE MANUALLY" in profiles_content
    assert "PROFILES:" in profiles_content
    assert "claude-3-opus" in profiles_content
    assert "claude-3-sonnet" in profiles_content

    # Check that augmentations were applied
    assert "image_url_inputs" in profiles_content
    assert "pdf_inputs" in profiles_content


def test_refresh_raises_error_for_missing_provider(
    tmp_path: Path, mock_models_dev_response: dict
) -> None:
    """Test that refresh exits with error for non-existent provider."""
    data_dir = tmp_path / "data"
    data_dir.mkdir()

    # Mock the httpx.get call
    mock_response = Mock()
    mock_response.json.return_value = mock_models_dev_response
    mock_response.raise_for_status = Mock()

    with (
        patch("langchain_model_profiles.cli.httpx.get", return_value=mock_response),
        patch("builtins.input", return_value="y"),
    ):
        with pytest.raises(SystemExit) as exc_info:
            refresh("nonexistent-provider", data_dir)

        assert exc_info.value.code == 1

    # Output file should not be created
    profiles_file = data_dir / "_profiles.py"
    assert not profiles_file.exists()


def test_refresh_works_without_augmentations(
    tmp_path: Path, mock_models_dev_response: dict
) -> None:
    """Test that refresh works even without augmentations file."""
    data_dir = tmp_path / "data"
    data_dir.mkdir()

    # Mock the httpx.get call
    mock_response = Mock()
    mock_response.json.return_value = mock_models_dev_response
    mock_response.raise_for_status = Mock()

    with (
        patch("langchain_model_profiles.cli.httpx.get", return_value=mock_response),
        patch("builtins.input", return_value="y"),
    ):
        refresh("anthropic", data_dir)

    # Verify _profiles.py was created
    profiles_file = data_dir / "_profiles.py"
    assert profiles_file.exists()
    assert profiles_file.stat().st_size > 0


def test_refresh_aborts_when_user_declines_external_directory(
    tmp_path: Path, mock_models_dev_response: dict
) -> None:
    """Test that refresh aborts when user declines writing to external directory."""
    data_dir = tmp_path / "data"
    data_dir.mkdir()

    # Mock the httpx.get call
    mock_response = Mock()
    mock_response.json.return_value = mock_models_dev_response
    mock_response.raise_for_status = Mock()

    with (
        patch("langchain_model_profiles.cli.httpx.get", return_value=mock_response),
        patch("builtins.input", return_value="n"),  # User declines
    ):
        with pytest.raises(SystemExit) as exc_info:
            refresh("anthropic", data_dir)

        assert exc_info.value.code == 1

    # Verify _profiles.py was NOT created
    profiles_file = data_dir / "_profiles.py"
    assert not profiles_file.exists()


def test_refresh_includes_models_defined_only_in_augmentations(
    tmp_path: Path, mock_models_dev_response: dict
) -> None:
    """Ensure models that only exist in augmentations are emitted."""
    data_dir = tmp_path / "data"
    data_dir.mkdir()

    aug_file = data_dir / "profile_augmentations.toml"
    aug_file.write_text("""
provider = "anthropic"

[overrides."custom-offline-model"]
structured_output = true
pdf_inputs = true
max_input_tokens = 123
""")

    mock_response = Mock()
    mock_response.json.return_value = mock_models_dev_response
    mock_response.raise_for_status = Mock()

    with (
        patch("langchain_model_profiles.cli.httpx.get", return_value=mock_response),
        patch("builtins.input", return_value="y"),
    ):
        refresh("anthropic", data_dir)

    profiles_file = data_dir / "_profiles.py"
    assert profiles_file.exists()

    spec = importlib.util.spec_from_file_location(
        "generated_profiles_aug_only", profiles_file
    )
    assert spec
    assert spec.loader
    module = importlib.util.module_from_spec(spec)
    spec.loader.exec_module(module)  # type: ignore[union-attr]

    assert "custom-offline-model" in module._PROFILES  # type: ignore[attr-defined]
    assert (
        module._PROFILES["custom-offline-model"]["structured_output"] is True  # type: ignore[index]
    )
    assert (
        module._PROFILES["custom-offline-model"]["max_input_tokens"] == 123  # type: ignore[index]
    )


def test_refresh_generates_sorted_profiles(
    tmp_path: Path, mock_models_dev_response: dict
) -> None:
    """Test that profiles are sorted alphabetically by model ID."""
    data_dir = tmp_path / "data"
    data_dir.mkdir()

    # Inject models in reverse-alphabetical order so the API response
    # is NOT already sorted.
    mock_models_dev_response["anthropic"]["models"] = {
        "z-model": {
            "id": "z-model",
            "name": "Z Model",
            "tool_call": True,
            "limit": {"context": 100000, "output": 2048},
            "modalities": {"input": ["text"], "output": ["text"]},
        },
        "a-model": {
            "id": "a-model",
            "name": "A Model",
            "tool_call": True,
            "limit": {"context": 100000, "output": 2048},
            "modalities": {"input": ["text"], "output": ["text"]},
        },
        "m-model": {
            "id": "m-model",
            "name": "M Model",
            "tool_call": True,
            "limit": {"context": 100000, "output": 2048},
            "modalities": {"input": ["text"], "output": ["text"]},
        },
    }

    mock_response = Mock()
    mock_response.json.return_value = mock_models_dev_response
    mock_response.raise_for_status = Mock()

    with (
        patch("langchain_model_profiles.cli.httpx.get", return_value=mock_response),
        patch("builtins.input", return_value="y"),
    ):
        refresh("anthropic", data_dir)

    profiles_file = data_dir / "_profiles.py"
    spec = importlib.util.spec_from_file_location(
        "generated_profiles_sorted", profiles_file
    )
    assert spec
    assert spec.loader
    module = importlib.util.module_from_spec(spec)
    spec.loader.exec_module(module)  # type: ignore[union-attr]

    model_ids = list(module._PROFILES.keys())  # type: ignore[attr-defined]
    assert model_ids == sorted(model_ids), f"Profile keys are not sorted: {model_ids}"


def test_model_data_to_profile_captures_all_models_dev_fields() -> None:
    """Test that all models.dev fields are captured in the profile."""
    model_data = {
        "id": "claude-opus-4-6",
        "name": "Claude Opus 4.6",
        "status": "deprecated",
        "release_date": "2025-06-01",
        "last_updated": "2025-07-01",
        "open_weights": False,
        "reasoning": True,
        "tool_call": True,
        "tool_choice": True,
        "structured_output": True,
        "attachment": True,
        "temperature": True,
        "limit": {"context": 200000, "output": 64000},
        "modalities": {
            "input": ["text", "image", "pdf"],
            "output": ["text"],
        },
    }
    profile = _model_data_to_profile(model_data)

    # Metadata
    assert profile["name"] == "Claude Opus 4.6"
    assert profile["status"] == "deprecated"
    assert profile["release_date"] == "2025-06-01"
    assert profile["last_updated"] == "2025-07-01"
    assert profile["open_weights"] is False

    # Limits
    assert profile["max_input_tokens"] == 200000
    assert profile["max_output_tokens"] == 64000

    # Capabilities
    assert profile["reasoning_output"] is True
    assert profile["tool_calling"] is True
    assert profile["tool_choice"] is True
    assert profile["structured_output"] is True
    assert profile["attachment"] is True

    # Modalities
    assert profile["text_inputs"] is True
    assert profile["image_inputs"] is True
    assert profile["pdf_inputs"] is True
    assert profile["text_outputs"] is True


def test_model_data_to_profile_omits_absent_fields() -> None:
    """Test that fields not present in source data are omitted (not None)."""
    minimal = {
        "modalities": {"input": ["text"], "output": ["text"]},
        "limit": {"context": 8192, "output": 4096},
    }
    profile = _model_data_to_profile(minimal)

    assert "status" not in profile
    assert "family" not in profile
    assert "knowledge_cutoff" not in profile
    assert "cost_input" not in profile
    assert "interleaved" not in profile
    assert None not in profile.values()


def test_model_data_to_profile_text_modalities() -> None:
    """Test that text input/output modalities are correctly mapped."""
    # Model with text in both input and output
    model_with_text = {
        "modalities": {"input": ["text", "image"], "output": ["text"]},
        "limit": {"context": 128000, "output": 4096},
    }
    profile = _model_data_to_profile(model_with_text)
    assert profile["text_inputs"] is True
    assert profile["text_outputs"] is True

    # Model without text input (e.g., Whisper-like audio model)
    audio_only_model = {
        "modalities": {"input": ["audio"], "output": ["text"]},
        "limit": {"context": 0, "output": 0},
    }
    profile = _model_data_to_profile(audio_only_model)
    assert profile["text_inputs"] is False
    assert profile["text_outputs"] is True

    # Model without text output (e.g., image generator)
    image_gen_model = {
        "modalities": {"input": ["text"], "output": ["image"]},
        "limit": {},
    }
    profile = _model_data_to_profile(image_gen_model)
    assert profile["text_inputs"] is True
    assert profile["text_outputs"] is False


def test_model_data_to_profile_keys_subset_of_model_profile() -> None:
    """All CLI-emitted profile keys must be declared in `ModelProfile`."""
    # Build a model_data dict with every possible field populated so
    # _model_data_to_profile includes all keys it can emit.
    model_data = {
        "id": "test-model",
        "name": "Test Model",
        "status": "active",
        "release_date": "2025-01-01",
        "last_updated": "2025-01-01",
        "open_weights": True,
        "reasoning": True,
        "tool_call": True,
        "tool_choice": True,
        "structured_output": True,
        "attachment": True,
        "temperature": True,
        "image_url_inputs": True,
        "image_tool_message": True,
        "pdf_tool_message": True,
        "pdf_inputs": True,
        "limit": {"context": 100000, "output": 4096},
        "modalities": {
            "input": ["text", "image", "audio", "video", "pdf"],
            "output": ["text", "image", "audio", "video"],
        },
    }

    profile = _model_data_to_profile(model_data)
    declared_fields = set(get_type_hints(ModelProfile).keys())
    emitted_fields = set(profile.keys())
    extra = emitted_fields - declared_fields

    assert not extra, (
        f"CLI emits profile keys not declared in ModelProfile: {sorted(extra)}. "
        f"Add these fields to langchain_core.language_models.model_profile."
        f"ModelProfile and release langchain-core before refreshing partner "
        f"profiles."
    )


class TestWarnUndeclaredProfileKeys:
    """Tests for _warn_undeclared_profile_keys."""

    def test_warns_on_undeclared_keys(self) -> None:
        """Extra keys across profiles trigger a single warning."""
        profiles: dict[str, dict[str, Any]] = {
            "model-a": {"max_input_tokens": 100, "future_key": True},
            "model-b": {"another_key": "val"},
        }
        with warnings.catch_warnings(record=True) as w:
            warnings.simplefilter("always")
            _warn_undeclared_profile_keys(profiles)

        assert len(w) == 1
        assert "another_key" in str(w[0].message)
        assert "future_key" in str(w[0].message)

    def test_silent_on_declared_keys_only(self) -> None:
        """No warning when all keys are declared in ModelProfile."""
        profiles: dict[str, dict[str, Any]] = {
            "model-a": {"max_input_tokens": 100, "tool_calling": True},
        }
        with warnings.catch_warnings(record=True) as w:
            warnings.simplefilter("always")
            _warn_undeclared_profile_keys(profiles)

        assert len(w) == 0

    def test_silent_when_langchain_core_not_installed(self) -> None:
        """Gracefully skips when langchain-core is not importable."""
        import sys

        profiles: dict[str, dict[str, Any]] = {
            "model-a": {"unknown": True},
        }
        with (
            patch.dict(
                sys.modules,
                {"langchain_core.language_models.model_profile": None},
            ),
            warnings.catch_warnings(record=True) as w,
        ):
            warnings.simplefilter("always")
            _warn_undeclared_profile_keys(profiles)

        undeclared_warnings = [x for x in w if "not declared" in str(x.message)]
        assert len(undeclared_warnings) == 0

    def test_survives_get_type_hints_failure(self) -> None:
        """Gracefully handles TypeError from get_type_hints."""
        profiles: dict[str, dict[str, Any]] = {
            "model-a": {"unknown": True},
        }
        with patch(
            "langchain_model_profiles.cli.get_type_hints",
            side_effect=TypeError("broken"),
        ):
            _warn_undeclared_profile_keys(profiles)