from unittest.mock import Mock, patch
import io
from types import SimpleNamespace

import pytest

from aisuite import Client
from aisuite.framework.message import ChatCompletionMessageToolCall, Function, Message
from aisuite.framework.message import TranscriptionResult
from aisuite.provider import ASRError


@pytest.fixture(scope="module")
def provider_configs():
    return {
        "openai": {"api_key": "test_openai_api_key"},
        "aws": {
            "aws_access_key": "test_aws_access_key",
            "aws_secret_key": "test_aws_secret_key",
            "aws_session_token": "test_aws_session_token",
            "aws_region": "us-west-2",
        },
        "azure": {
            "api_key": "azure-api-key",
            "base_url": "https://model.ai.azure.com",
        },
        "groq": {
            "api_key": "groq-api-key",
        },
        "mistral": {
            "api_key": "mistral-api-key",
        },
        "google": {
            "project_id": "test_google_project_id",
            "region": "us-west4",
            "application_credentials": "test_google_application_credentials",
        },
        "fireworks": {
            "api_key": "fireworks-api-key",
        },
        "nebius": {
            "api_key": "nebius-api-key",
        },
        "inception": {
            "api_key": "inception-api-key",
        },
        "deepgram": {
            "api_key": "deepgram-api-key",
        },
    }


@pytest.mark.parametrize(
    argnames=("provider", "model"),
    argvalues=[
        ("openai", "gpt-4o"),
        ("mistral", "mistral-model"),
        ("groq", "groq-model"),
        ("aws", "claude-v3"),
        ("azure", "azure-model"),
        ("anthropic", "anthropic-model"),
        ("google", "google-model"),
        ("fireworks", "fireworks-model"),
        ("nebius", "nebius-model"),
        ("inception", "mercury"),
    ],
)
def test_client_chat_completions(provider_configs: dict, provider: str, model: str):
    expected_response = f"{provider}_{model}"
    mock_provider = Mock()
    mock_provider.chat_completions_create.return_value = expected_response

    with patch("aisuite.provider.ProviderFactory.create_provider") as create_provider:
        create_provider.return_value = mock_provider
        client = Client()
        client.configure(provider_configs)
        messages = [
            {"role": "system", "content": "You are a helpful assistant."},
            {"role": "user", "content": "Who won the world series in 2020?"},
        ]

        model_str = f"{provider}:{model}"
        model_response = client.chat.completions.create(model_str, messages=messages)
        assert model_response == expected_response
        create_provider.assert_called_once_with(
            provider, provider_configs.get(provider, {})
        )
        mock_provider.chat_completions_create.assert_called_once_with(model, messages)


def test_invalid_provider_in_client_config():
    # Testing an invalid provider name in the configuration
    invalid_provider_configs = {
        "invalid_provider": {"api_key": "invalid_api_key"},
    }

    # With lazy loading, Client initialization should succeed
    client = Client()
    client.configure(invalid_provider_configs)

    messages = [
        {"role": "user", "content": "Hello"},
    ]

    # Expect ValueError when actually trying to use the invalid provider
    with pytest.raises(
        ValueError,
        match=r"Invalid provider key 'invalid_provider'. Supported providers: ",
    ):
        client.chat.completions.create("invalid_provider:some-model", messages=messages)


def test_invalid_model_format_in_create():
    # Valid provider configurations
    provider_configs = {
        "openai": {"api_key": "test_openai_api_key"},
    }

    # Initialize the client with valid provider
    client = Client()
    client.configure(provider_configs)

    messages = [
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "Tell me a joke."},
    ]

    # Invalid model format
    invalid_model = "invalidmodel"

    # Expect ValueError when calling create with invalid model format and verify message
    with pytest.raises(
        ValueError, match=r"Invalid model format. Expected 'provider:model'"
    ):
        client.chat.completions.create(invalid_model, messages=messages)


def _chat_response(content=None, tool_calls=None):
    message = Message(role="assistant", content=content, tool_calls=tool_calls)
    return SimpleNamespace(choices=[SimpleNamespace(message=message)])


def _tool_call(name, arguments, call_id="call_1"):
    return ChatCompletionMessageToolCall(
        id=call_id,
        type="function",
        function=Function(name=name, arguments=arguments),
    )


@patch("aisuite.provider.ProviderFactory.create_provider")
def test_chat_completions_executes_tools_until_final_response(mock_create_provider):
    provider = Mock()
    provider.chat_completions_create.side_effect = [
        _chat_response(
            tool_calls=[
                _tool_call(
                    "get_weather",
                    '{"location": "San Francisco"}',
                )
            ]
        ),
        _chat_response(content="It is sunny in San Francisco."),
    ]
    mock_create_provider.return_value = provider

    def get_weather(location: str):
        """Get the weather for a location."""
        return {"location": location, "condition": "sunny"}

    client = Client()
    response = client.chat.completions.create(
        model="openai:gpt-4o",
        messages=[{"role": "user", "content": "What is the weather?"}],
        tools=[get_weather],
        max_turns=2,
    )

    assert response.choices[0].message.content == "It is sunny in San Francisco."
    assert len(response.intermediate_responses) == 1
    assert len(response.choices[0].intermediate_messages) == 3

    second_call_messages = provider.chat_completions_create.call_args_list[1].args[1]
    assert second_call_messages[-1] == {
        "role": "tool",
        "name": "get_weather",
        "content": '{"location": "San Francisco", "condition": "sunny"}',
        "tool_call_id": "call_1",
    }


@patch("aisuite.provider.ProviderFactory.create_provider")
def test_ollama_model_runs_full_tool_loop(mock_create_provider):
    """End-to-end: an ollama:<model> drives a real tool-execution turn.

    Exercises the real OllamaProvider (request conversion + OpenAI passthrough)
    with only its underlying OpenAI client mocked, proving tool calls survive.
    """
    from aisuite.providers.ollama_provider import OllamaProvider

    provider = OllamaProvider()
    provider.client.chat.completions.create = Mock(
        side_effect=[
            _chat_response(
                tool_calls=[_tool_call("get_weather", '{"location": "San Francisco"}')]
            ),
            _chat_response(content="It is sunny in San Francisco."),
        ]
    )
    mock_create_provider.return_value = provider

    def get_weather(location: str):
        """Get the weather for a location."""
        return {"location": location, "condition": "sunny"}

    client = Client()
    response = client.chat.completions.create(
        model="ollama:llama3.1",
        messages=[{"role": "user", "content": "What is the weather?"}],
        tools=[get_weather],
        max_turns=2,
    )

    assert response.choices[0].message.content == "It is sunny in San Francisco."
    # Two model calls: the tool-call turn and the final answer.
    assert provider.client.chat.completions.create.call_count == 2


@patch("aisuite.provider.ProviderFactory.create_provider")
def test_chat_completions_returns_last_response_when_max_turns_reached(
    mock_create_provider,
):
    provider = Mock()
    provider.chat_completions_create.return_value = _chat_response(
        tool_calls=[_tool_call("echo", '{"value": "hello"}')]
    )
    mock_create_provider.return_value = provider

    def echo(value: str):
        """Echo a value."""
        return value

    client = Client()
    response = client.chat.completions.create(
        model="openai:gpt-4o",
        messages=[{"role": "user", "content": "Echo hello forever"}],
        tools=[echo],
        max_turns=1,
    )

    assert provider.chat_completions_create.call_count == 1
    assert response.choices[0].message.tool_calls[0].function.name == "echo"
    assert response.intermediate_responses == []
    assert len(response.choices[0].intermediate_messages) == 2


@patch("aisuite.provider.ProviderFactory.create_provider")
def test_chat_completions_extracts_thinking_content(mock_create_provider):
    provider = Mock()
    provider.chat_completions_create.return_value = _chat_response(
        content="<think>private reasoning</think>\nFinal answer"
    )
    mock_create_provider.return_value = provider

    client = Client()
    response = client.chat.completions.create(
        model="openai:gpt-4o",
        messages=[{"role": "user", "content": "Answer"}],
    )

    assert response.choices[0].message.reasoning_content == "private reasoning"
    assert response.choices[0].message.content == "Final answer"


class TestClientASR:
    """Test suite for Client ASR functionality - essential tests only."""

    def test_audio_interface_initialization(self):
        """Test that Audio interface is properly initialized."""
        client = Client()
        assert hasattr(client, "audio")
        assert hasattr(client.audio, "transcriptions")

    @patch("aisuite.provider.ProviderFactory.create_provider")
    def test_transcriptions_create_success(
        self, mock_create_provider, provider_configs
    ):
        """Test successful audio transcription with OpenAI."""
        mock_result = TranscriptionResult(
            text="Hello, this is a test transcription.",
            language="en",
            confidence=0.95,
            task="transcribe",
        )

        # Create a mock provider with audio support
        mock_provider = Mock()
        mock_provider.audio.transcriptions.create.return_value = mock_result
        mock_create_provider.return_value = mock_provider

        client = Client()
        client.configure(provider_configs)

        audio_data = io.BytesIO(b"fake audio data")
        result = client.audio.transcriptions.create(
            model="openai:whisper-1", file=audio_data, language="en"
        )

        assert isinstance(result, TranscriptionResult)
        assert result.text == "Hello, this is a test transcription."
        mock_provider.audio.transcriptions.create.assert_called_once()

    @patch("aisuite.provider.ProviderFactory.create_provider")
    def test_transcriptions_create_deepgram(
        self, mock_create_provider, provider_configs
    ):
        """Test audio transcription with Deepgram provider."""
        mock_result = TranscriptionResult(
            text="Deepgram transcription result.",
            language="en",
            confidence=0.92,
            task="transcribe",
        )

        # Create a mock provider with audio support
        mock_provider = Mock()
        mock_provider.audio.transcriptions.create.return_value = mock_result
        mock_create_provider.return_value = mock_provider

        client = Client()
        client.configure(provider_configs)

        result = client.audio.transcriptions.create(
            model="deepgram:nova-2", file="test_audio.wav", language="en"
        )

        assert isinstance(result, TranscriptionResult)
        assert result.text == "Deepgram transcription result."
        mock_provider.audio.transcriptions.create.assert_called_once()

    def test_transcriptions_invalid_model_format(self, provider_configs):
        """Test that invalid model format raises ValueError."""
        client = Client()
        client.configure(provider_configs)

        with pytest.raises(ValueError, match="Invalid model format"):
            client.audio.transcriptions.create(
                model="invalid-format", file="test.wav", language="en"
            )

    def test_transcriptions_unsupported_provider(self, provider_configs):
        """Test error handling for unsupported ASR provider."""
        client = Client()
        client.configure(provider_configs)

        with pytest.raises(ValueError, match="Invalid provider key"):
            client.audio.transcriptions.create(
                model="unsupported:model", file="test.wav", language="en"
            )


class TestClientASRParameterValidation:
    """Test suite for Client-level ASR parameter validation."""

    def test_client_initialization_strict_mode(self):
        """Test Client initialization with strict extra_param_mode."""
        client = Client(extra_param_mode="strict")
        assert client.extra_param_mode == "strict"
        assert client.param_validator.extra_param_mode == "strict"

    def test_client_initialization_warn_mode(self):
        """Test Client initialization with warn extra_param_mode (default)."""
        client = Client()
        assert client.extra_param_mode == "warn"
        assert client.param_validator.extra_param_mode == "warn"

    def test_client_initialization_permissive_mode(self):
        """Test Client initialization with permissive extra_param_mode."""
        client = Client(extra_param_mode="permissive")
        assert client.extra_param_mode == "permissive"
        assert client.param_validator.extra_param_mode == "permissive"

    @patch("aisuite.provider.ProviderFactory.create_provider")
    def test_strict_mode_rejects_unknown_param(self, mock_create_provider):
        """Test that strict mode raises ValueError for unknown parameters."""
        client = Client(
            provider_configs={"openai": {"api_key": "test"}}, extra_param_mode="strict"
        )

        # Mock provider shouldn't be called due to validation error
        mock_provider = Mock()
        mock_create_provider.return_value = mock_provider

        with pytest.raises(ValueError, match="Unknown parameters for openai"):
            client.audio.transcriptions.create(
                model="openai:whisper-1",
                file=io.BytesIO(b"audio"),
                language="en",
                invalid_param=True,  # Unknown param
            )

        # Provider should not have been called (validation failed first)
        mock_provider.audio.transcriptions.create.assert_not_called()

    @patch("aisuite.provider.ProviderFactory.create_provider")
    def test_strict_mode_typo_detection(self, mock_create_provider):
        """Test that strict mode catches typos in parameter names."""
        client = Client(
            provider_configs={"openai": {"api_key": "test"}}, extra_param_mode="strict"
        )

        mock_provider = Mock()
        mock_create_provider.return_value = mock_provider

        with pytest.raises(
            ValueError, match="Unknown parameters for openai: \\['langauge'\\]"
        ):
            client.audio.transcriptions.create(
                model="openai:whisper-1",
                file=io.BytesIO(b"audio"),
                langauge="en",  # TYPO: should be "language"
            )

    @patch("aisuite.provider.ProviderFactory.create_provider")
    def test_warn_mode_continues_execution(self, mock_create_provider):
        """Test that warn mode continues execution after warning."""
        import warnings

        client = Client(
            provider_configs={"openai": {"api_key": "test"}}, extra_param_mode="warn"
        )

        mock_result = TranscriptionResult(text="Test", language="en")
        mock_provider = Mock()
        mock_provider.audio.transcriptions.create.return_value = mock_result
        mock_create_provider.return_value = mock_provider

        # Should warn but continue
        with warnings.catch_warnings(record=True) as w:
            warnings.simplefilter("always")
            result = client.audio.transcriptions.create(
                model="openai:whisper-1",
                file=io.BytesIO(b"audio"),
                language="en",
                invalid_param=True,  # Unknown param
            )

            # Should have issued a warning
            assert len(w) == 1
            assert "Unknown parameters" in str(w[0].message)

            # But execution should continue
            assert result.text == "Test"
            mock_provider.audio.transcriptions.create.assert_called_once()

    @patch("aisuite.provider.ProviderFactory.create_provider")
    def test_permissive_mode_allows_unknown_params(self, mock_create_provider):
        """Test that permissive mode allows unknown parameters."""
        import warnings

        client = Client(
            provider_configs={"openai": {"api_key": "test"}},
            extra_param_mode="permissive",
        )

        mock_result = TranscriptionResult(text="Test", language="en")
        mock_provider = Mock()
        mock_provider.audio.transcriptions.create.return_value = mock_result
        mock_create_provider.return_value = mock_provider

        # Should not warn or raise
        with warnings.catch_warnings(record=True) as w:
            warnings.simplefilter("always")
            result = client.audio.transcriptions.create(
                model="openai:whisper-1",
                file=io.BytesIO(b"audio"),
                experimental_feature=True,  # Unknown param
            )

            # Should not have issued any warnings
            assert len(w) == 0

            # Execution should succeed
            assert result.text == "Test"
            mock_provider.audio.transcriptions.create.assert_called_once()

            # Unknown param should be passed through
            call_kwargs = mock_provider.audio.transcriptions.create.call_args.kwargs
            assert call_kwargs.get("experimental_feature") is True

    @patch("aisuite.provider.ProviderFactory.create_provider")
    def test_common_param_mapping_at_client_level(self, mock_create_provider):
        """Test that common parameters are mapped correctly at Client level."""
        client = Client(
            provider_configs={"google": {"project_id": "test", "region": "us"}},
            extra_param_mode="strict",
        )

        mock_result = TranscriptionResult(text="Test", language="en")
        mock_provider = Mock()
        mock_provider.audio.transcriptions.create.return_value = mock_result
        mock_create_provider.return_value = mock_provider

        # Use common param "language" which should map to "language_code" for Google
        result = client.audio.transcriptions.create(
            model="google:latest_long",
            file=io.BytesIO(b"audio"),
            language="en",  # Common param
        )

        assert result.text == "Test"
        mock_provider.audio.transcriptions.create.assert_called_once()

        # Verify parameter was mapped to language_code
        call_kwargs = mock_provider.audio.transcriptions.create.call_args.kwargs
        assert "language_code" in call_kwargs
        assert call_kwargs["language_code"] == "en-US"  # Expanded
        assert "language" not in call_kwargs  # Original key should be mapped

    @patch("aisuite.provider.ProviderFactory.create_provider")
    def test_provider_specific_params_passthrough(self, mock_create_provider):
        """Test that provider-specific parameters pass through correctly."""
        client = Client(
            provider_configs={"deepgram": {"api_key": "test"}},
            extra_param_mode="strict",
        )

        mock_result = TranscriptionResult(text="Test", language="en")
        mock_provider = Mock()
        mock_provider.audio.transcriptions.create.return_value = mock_result
        mock_create_provider.return_value = mock_provider

        result = client.audio.transcriptions.create(
            model="deepgram:nova-2",
            file=io.BytesIO(b"audio"),
            punctuate=True,
            diarize=True,
        )

        assert result.text == "Test"

        # Verify provider-specific params passed through
        call_kwargs = mock_provider.audio.transcriptions.create.call_args.kwargs
        assert call_kwargs["punctuate"] is True
        assert call_kwargs["diarize"] is True

    @patch("aisuite.provider.ProviderFactory.create_provider")
    def test_mixed_common_and_provider_params(self, mock_create_provider):
        """Test mixing common and provider-specific parameters."""
        client = Client(
            provider_configs={"deepgram": {"api_key": "test"}},
            extra_param_mode="strict",
        )

        mock_result = TranscriptionResult(text="Test", language="en")
        mock_provider = Mock()
        mock_provider.audio.transcriptions.create.return_value = mock_result
        mock_create_provider.return_value = mock_provider

        result = client.audio.transcriptions.create(
            model="deepgram:nova-2",
            file=io.BytesIO(b"audio"),
            language="en",  # Common param
            prompt="meeting",  # Common param that maps to keywords
            punctuate=True,  # Deepgram-specific
            diarize=True,  # Deepgram-specific
        )

        assert result.text == "Test"

        # Verify both common and provider params processed correctly
        call_kwargs = mock_provider.audio.transcriptions.create.call_args.kwargs
        assert call_kwargs["language"] == "en"
        assert call_kwargs["keywords"] == ["meeting"]  # prompt mapped to keywords
        assert call_kwargs["punctuate"] is True
        assert call_kwargs["diarize"] is True

    @patch("aisuite.provider.ProviderFactory.create_provider")
    def test_validation_happens_before_provider_call(self, mock_create_provider):
        """Test that validation occurs before provider SDK is called."""
        client = Client(
            provider_configs={"openai": {"api_key": "test"}}, extra_param_mode="strict"
        )

        mock_provider = Mock()
        mock_create_provider.return_value = mock_provider

        # Validation should fail before provider is even initialized
        with pytest.raises(ValueError, match="Unknown parameters"):
            client.audio.transcriptions.create(
                model="openai:whisper-1",
                file=io.BytesIO(b"audio"),
                completely_invalid_param=True,
            )

        # Provider create method should still have been called to initialize
        # but the transcription method should never be called
        mock_provider.audio.transcriptions.create.assert_not_called()

    @patch("aisuite.provider.ProviderFactory.create_provider")
    def test_unsupported_common_param_ignored(self, mock_create_provider):
        """Test that unsupported common params are gracefully ignored."""
        client = Client(
            provider_configs={"deepgram": {"api_key": "test"}},
            extra_param_mode="strict",
        )

        mock_result = TranscriptionResult(text="Test", language="en")
        mock_provider = Mock()
        mock_provider.audio.transcriptions.create.return_value = mock_result
        mock_create_provider.return_value = mock_provider

        # temperature is not supported by Deepgram (should be ignored)
        result = client.audio.transcriptions.create(
            model="deepgram:nova-2",
            file=io.BytesIO(b"audio"),
            language="en",
            temperature=0.5,  # Not supported by Deepgram
        )

        assert result.text == "Test"

        # Verify temperature was not passed to provider
        call_kwargs = mock_provider.audio.transcriptions.create.call_args.kwargs
        assert "temperature" not in call_kwargs
        assert call_kwargs["language"] == "en"

    @patch("aisuite.provider.ProviderFactory.create_provider")
    def test_multiple_providers_with_same_client(self, mock_create_provider):
        """Test that the same client can handle multiple providers with different validation."""
        client = Client(
            provider_configs={
                "openai": {"api_key": "test1"},
                "deepgram": {"api_key": "test2"},
            },
            extra_param_mode="strict",
        )

        mock_result = TranscriptionResult(text="Test", language="en")
        mock_provider = Mock()
        mock_provider.audio.transcriptions.create.return_value = mock_result
        mock_create_provider.return_value = mock_provider

        # Test OpenAI with temperature (supported)
        result1 = client.audio.transcriptions.create(
            model="openai:whisper-1", file=io.BytesIO(b"audio"), temperature=0.5
        )
        assert result1.text == "Test"
        call_kwargs1 = mock_provider.audio.transcriptions.create.call_args.kwargs
        assert call_kwargs1.get("temperature") == 0.5

        # Reset mock
        mock_provider.reset_mock()

        # Test Deepgram with temperature (not supported, should be ignored)
        result2 = client.audio.transcriptions.create(
            model="deepgram:nova-2", file=io.BytesIO(b"audio"), temperature=0.5
        )
        assert result2.text == "Test"
        call_kwargs2 = mock_provider.audio.transcriptions.create.call_args.kwargs
        assert "temperature" not in call_kwargs2