deepeval 3.7.4__py3-none-any.whl → 3.7.6__py3-none-any.whl

deepeval/evaluate/execute.py

@@ -56,7 +56,6 @@ from deepeval.telemetry import capture_evaluation_run
 from deepeval.metrics import (
     BaseMetric,
     BaseConversationalMetric,
-    BaseMultimodalMetric,
     TaskCompletionMetric,
 )
 from deepeval.metrics.indicator import (
@@ -70,7 +69,6 @@ from deepeval.models.retry_policy import (
 from deepeval.test_case import (
     LLMTestCase,
     ConversationalTestCase,
-    MLLMTestCase,
 )
 from deepeval.test_case.api import create_api_test_case
 from deepeval.test_run import (
@@ -263,13 +261,10 @@ async def _await_with_outer_deadline(obj, *args, timeout: float, **kwargs):
 
 
 def execute_test_cases(
-    test_cases: Union[
-        List[LLMTestCase], List[ConversationalTestCase], List[MLLMTestCase]
-    ],
+    test_cases: Union[List[LLMTestCase], List[ConversationalTestCase]],
     metrics: Union[
         List[BaseMetric],
         List[BaseConversationalMetric],
-        List[BaseMultimodalMetric],
     ],
     error_config: Optional[ErrorConfig] = ErrorConfig(),
     display_config: Optional[DisplayConfig] = DisplayConfig(),
@@ -302,15 +297,12 @@ def execute_test_cases(
 
     conversational_metrics: List[BaseConversationalMetric] = []
     llm_metrics: List[BaseMetric] = []
-    mllm_metrics: List[BaseMultimodalMetric] = []
     for metric in metrics:
         metric.async_mode = False
         if isinstance(metric, BaseMetric):
             llm_metrics.append(metric)
         elif isinstance(metric, BaseConversationalMetric):
             conversational_metrics.append(metric)
-        elif isinstance(metric, BaseMultimodalMetric):
-            mllm_metrics.append(metric)
 
     test_results: List[TestResult] = []
 
@@ -318,7 +310,6 @@ def execute_test_cases(
         progress: Optional[Progress] = None, pbar_id: Optional[int] = None
     ):
         llm_test_case_count = -1
-        mllm_test_case_count = -1
         conversational_test_case_count = -1
         show_metric_indicator = (
             display_config.show_indicator and not _use_bar_indicator
@@ -330,11 +321,6 @@ def execute_test_cases(
                     update_pbar(progress, pbar_id)
                     continue
                 per_case_total = len(llm_metrics)
-            elif isinstance(test_case, MLLMTestCase):
-                if not mllm_metrics:
-                    update_pbar(progress, pbar_id)
-                    continue
-                per_case_total = len(mllm_metrics)
             elif isinstance(test_case, ConversationalTestCase):
                 if not conversational_metrics:
                     update_pbar(progress, pbar_id)
@@ -349,23 +335,15 @@ def execute_test_cases(
 
             metrics_for_case = (
                 llm_metrics
-                if isinstance(test_case, LLMTestCase)
-                else (
-                    mllm_metrics
-                    if isinstance(test_case, MLLMTestCase)
-                    else conversational_metrics
-                )
+                if (isinstance(test_case, LLMTestCase))
+                else conversational_metrics
             )
             api_test_case = create_api_test_case(
                 test_case=test_case,
                 index=(
                     llm_test_case_count + 1
-                    if isinstance(test_case, LLMTestCase)
-                    else (
-                        mllm_test_case_count + 1
-                        if isinstance(test_case, MLLMTestCase)
-                        else conversational_test_case_count + 1
-                    )
+                    if (isinstance(test_case, LLMTestCase))
+                    else (conversational_test_case_count + 1)
                 ),
             )
             emitted = [False] * len(metrics_for_case)
@@ -378,7 +356,7 @@ def execute_test_cases(
             try:
 
                 def _run_case():
-                    nonlocal new_cached_test_case, current_index, llm_test_case_count, mllm_test_case_count, conversational_test_case_count
+                    nonlocal new_cached_test_case, current_index, llm_test_case_count, conversational_test_case_count
                     with capture_evaluation_run("test case"):
                         for metric in metrics:
                             metric.error = None  # Reset metric error
@@ -435,26 +413,6 @@ def execute_test_cases(
                                     )
                                 update_pbar(progress, pbar_test_case_id)
 
-                        # No caching and not sending test cases to Confident AI for multimodal metrics yet
-                        elif isinstance(test_case, MLLMTestCase):
-                            mllm_test_case_count += 1
-                            for metric in mllm_metrics:
-                                current_index = index_of[id(metric)]
-                                res = _execute_metric(
-                                    metric=metric,
-                                    test_case=test_case,
-                                    show_metric_indicator=show_metric_indicator,
-                                    in_component=False,
-                                    error_config=error_config,
-                                )
-                                if res == "skip":
-                                    continue
-
-                                metric_data = create_metric_data(metric)
-                                api_test_case.update_metric_data(metric_data)
-                                emitted[current_index] = True
-                                update_pbar(progress, pbar_test_case_id)
-
                         # No caching for conversational metrics yet
                         elif isinstance(test_case, ConversationalTestCase):
                             conversational_test_case_count += 1
@@ -560,13 +518,10 @@ def execute_test_cases(
 
 
 async def a_execute_test_cases(
-    test_cases: Union[
-        List[LLMTestCase], List[ConversationalTestCase], List[MLLMTestCase]
-    ],
+    test_cases: Union[List[LLMTestCase], List[ConversationalTestCase]],
     metrics: Union[
         List[BaseMetric],
         List[BaseConversationalMetric],
-        List[BaseMultimodalMetric],
     ],
     error_config: Optional[ErrorConfig] = ErrorConfig(),
     display_config: Optional[DisplayConfig] = DisplayConfig(),
@@ -600,20 +555,16 @@ async def a_execute_test_cases(
             metric.verbose_mode = display_config.verbose_mode
 
     llm_metrics: List[BaseMetric] = []
-    mllm_metrics: List[BaseMultimodalMetric] = []
     conversational_metrics: List[BaseConversationalMetric] = []
     for metric in metrics:
         if isinstance(metric, BaseMetric):
             llm_metrics.append(metric)
-        elif isinstance(metric, BaseMultimodalMetric):
-            mllm_metrics.append(metric)
         elif isinstance(metric, BaseConversationalMetric):
             conversational_metrics.append(metric)
 
     llm_test_case_counter = -1
-    mllm_test_case_counter = -1
     conversational_test_case_counter = -1
-    test_results: List[Union[TestResult, MLLMTestCase]] = []
+    test_results: List[Union[TestResult, LLMTestCase]] = []
     tasks = []
 
     if display_config.show_indicator and _use_bar_indicator:
@@ -660,28 +611,6 @@ async def a_execute_test_cases(
                         )
                         tasks.append(asyncio.create_task(task))
 
-                    elif isinstance(test_case, MLLMTestCase):
-                        mllm_test_case_counter += 1
-                        copied_multimodal_metrics: List[
-                            BaseMultimodalMetric
-                        ] = copy_metrics(mllm_metrics)
-                        task = execute_with_semaphore(
-                            func=_a_execute_mllm_test_cases,
-                            metrics=copied_multimodal_metrics,
-                            test_case=test_case,
-                            test_run_manager=test_run_manager,
-                            test_results=test_results,
-                            count=mllm_test_case_counter,
-                            ignore_errors=error_config.ignore_errors,
-                            skip_on_missing_params=error_config.skip_on_missing_params,
-                            show_indicator=display_config.show_indicator,
-                            _use_bar_indicator=_use_bar_indicator,
-                            _is_assert_test=_is_assert_test,
-                            progress=progress,
-                            pbar_id=pbar_id,
-                        )
-                        tasks.append(asyncio.create_task(task))
-
                     elif isinstance(test_case, ConversationalTestCase):
                         conversational_test_case_counter += 1
 
@@ -772,26 +701,6 @@ async def a_execute_test_cases(
                     )
                     tasks.append(asyncio.create_task((task)))
 
-                elif isinstance(test_case, MLLMTestCase):
-                    mllm_test_case_counter += 1
-                    copied_multimodal_metrics: List[BaseMultimodalMetric] = (
-                        copy_metrics(mllm_metrics)
-                    )
-                    task = execute_with_semaphore(
-                        func=_a_execute_mllm_test_cases,
-                        metrics=copied_multimodal_metrics,
-                        test_case=test_case,
-                        test_run_manager=test_run_manager,
-                        test_results=test_results,
-                        count=mllm_test_case_counter,
-                        ignore_errors=error_config.ignore_errors,
-                        skip_on_missing_params=error_config.skip_on_missing_params,
-                        _use_bar_indicator=_use_bar_indicator,
-                        _is_assert_test=_is_assert_test,
-                        show_indicator=display_config.show_indicator,
-                    )
-                    tasks.append(asyncio.create_task(task))
-
                 await asyncio.sleep(async_config.throttle_value)
 
         try:
@@ -815,7 +724,7 @@ async def _a_execute_llm_test_cases(
     metrics: List[BaseMetric],
     test_case: LLMTestCase,
     test_run_manager: TestRunManager,
-    test_results: List[Union[TestResult, MLLMTestCase]],
+    test_results: List[Union[TestResult, LLMTestCase]],
     count: int,
     test_run: TestRun,
     ignore_errors: bool,
@@ -932,88 +841,11 @@ async def _a_execute_llm_test_cases(
         update_pbar(progress, pbar_id)
 
 
-async def _a_execute_mllm_test_cases(
-    metrics: List[BaseMultimodalMetric],
-    test_case: MLLMTestCase,
-    test_run_manager: TestRunManager,
-    test_results: List[Union[TestResult, MLLMTestCase]],
-    count: int,
-    ignore_errors: bool,
-    skip_on_missing_params: bool,
-    show_indicator: bool,
-    _use_bar_indicator: bool,
-    _is_assert_test: bool,
-    progress: Optional[Progress] = None,
-    pbar_id: Optional[int] = None,
-):
-    show_metrics_indicator = show_indicator and not _use_bar_indicator
-    pbar_test_case_id = add_pbar(
-        progress,
-        f"    🎯 Evaluating test case #{count}",
-        total=len(metrics),
-    )
-
-    for metric in metrics:
-        metric.skipped = False
-        metric.error = None  # Reset metric error
-
-    api_test_case: LLMApiTestCase = create_api_test_case(
-        test_case=test_case, index=count if not _is_assert_test else None
-    )
-    test_start_time = time.perf_counter()
-    try:
-        await measure_metrics_with_indicator(
-            metrics=metrics,
-            test_case=test_case,
-            cached_test_case=None,
-            skip_on_missing_params=skip_on_missing_params,
-            ignore_errors=ignore_errors,
-            show_indicator=show_metrics_indicator,
-            pbar_eval_id=pbar_test_case_id,
-            progress=progress,
-        )
-    except asyncio.CancelledError:
-        msg = (
-            "Timed out/cancelled while evaluating metric. "
-            "Increase DEEPEVAL_PER_TASK_TIMEOUT_SECONDS_OVERRIDE or set "
-            "DEEPEVAL_LOG_STACK_TRACES=1 for full traceback."
-        )
-        for m in metrics:
-            if getattr(m, "skipped", False):
-                continue
-            # If the task never finished and didn't set a terminal state, mark it now
-            if getattr(m, "success", None) is None and not getattr(
-                m, "error", None
-            ):
-                m.success = False
-                m.error = msg
-        if not ignore_errors:
-            raise
-    finally:
-        for metric in metrics:
-            if metric.skipped:
-                continue
-
-            metric_data = create_metric_data(metric)
-            api_test_case.update_metric_data(metric_data)
-
-        test_end_time = time.perf_counter()
-        run_duration = test_end_time - test_start_time
-        api_test_case.update_run_duration(run_duration)
-
-        ### Update Test Run ###
-        test_run_manager.update_test_run(api_test_case, test_case)
-        test_results.append(create_test_result(api_test_case))
-        update_pbar(progress, pbar_id)
-
-
 async def _a_execute_conversational_test_cases(
-    metrics: List[
-        Union[BaseMetric, BaseMultimodalMetric, BaseConversationalMetric]
-    ],
+    metrics: List[Union[BaseMetric, BaseConversationalMetric]],
     test_case: ConversationalTestCase,
     test_run_manager: TestRunManager,
-    test_results: List[Union[TestResult, MLLMTestCase]],
+    test_results: List[Union[TestResult, LLMTestCase]],
     count: int,
     ignore_errors: bool,
     skip_on_missing_params: bool,
@@ -1776,7 +1608,7 @@ async def a_execute_agentic_test_cases(
 async def _a_execute_agentic_test_case(
     golden: Golden,
     test_run_manager: TestRunManager,
-    test_results: List[Union[TestResult, MLLMTestCase]],
+    test_results: List[Union[TestResult, LLMTestCase]],
     count: int,
     verbose_mode: Optional[bool],
     ignore_errors: bool,
@@ -3205,7 +3037,7 @@ async def _evaluate_test_case_pairs(
 
 def _execute_metric(
     metric: BaseMetric,
-    test_case: Union[LLMTestCase, ConversationalTestCase, MLLMTestCase],
+    test_case: Union[LLMTestCase, ConversationalTestCase],
     show_metric_indicator: bool,
     in_component: bool,
     error_config: ErrorConfig,

deepeval/evaluate/utils.py

@@ -11,12 +11,10 @@ from deepeval.metrics import (
     ArenaGEval,
     BaseMetric,
     BaseConversationalMetric,
-    BaseMultimodalMetric,
 )
 from deepeval.test_case import (
     LLMTestCase,
     ConversationalTestCase,
-    MLLMTestCase,
 )
 from deepeval.test_run import (
     LLMApiTestCase,
@@ -129,17 +127,14 @@ def create_test_result(
             turns=api_test_case.turns,
         )
     else:
-        multimodal = (
-            api_test_case.multimodal_input is not None
-            and api_test_case.multimodal_input_actual_output is not None
-        )
+        multimodal = api_test_case.images_mapping
         if multimodal:
             return TestResult(
                 name=name,
                 success=api_test_case.success,
                 metrics_data=api_test_case.metrics_data,
-                input=api_test_case.multimodal_input,
-                actual_output=api_test_case.multimodal_input_actual_output,
+                input=api_test_case.input,
+                actual_output=api_test_case.actual_output,
                 conversational=False,
                 multimodal=True,
                 additional_metadata=api_test_case.additional_metadata,
@@ -222,7 +217,7 @@ def validate_assert_test_inputs(
         )
 
     if test_case and metrics:
-        if isinstance(test_case, LLMTestCase) and not all(
+        if (isinstance(test_case, LLMTestCase)) and not all(
             isinstance(metric, BaseMetric) for metric in metrics
         ):
             raise ValueError(
@@ -234,12 +229,6 @@ def validate_assert_test_inputs(
             raise ValueError(
                 "All 'metrics' for an 'ConversationalTestCase' must be instances of 'BaseConversationalMetric' only."
             )
-        if isinstance(test_case, MLLMTestCase) and not all(
-            isinstance(metric, BaseMultimodalMetric) for metric in metrics
-        ):
-            raise ValueError(
-                "All 'metrics' for an 'MLLMTestCase' must be instances of 'BaseMultimodalMetric' only."
-            )
 
     if not ((golden and observed_callback) or (test_case and metrics)):
         raise ValueError(
@@ -251,15 +240,12 @@ def validate_evaluate_inputs(
     goldens: Optional[List] = None,
     observed_callback: Optional[Callable] = None,
     test_cases: Optional[
-        Union[
-            List[LLMTestCase], List[ConversationalTestCase], List[MLLMTestCase]
-        ]
+        Union[List[LLMTestCase], List[ConversationalTestCase]]
     ] = None,
     metrics: Optional[
         Union[
             List[BaseMetric],
             List[BaseConversationalMetric],
-            List[BaseMultimodalMetric],
         ]
     ] = None,
     metric_collection: Optional[str] = None,
@@ -292,7 +278,7 @@ def validate_evaluate_inputs(
     if test_cases and metrics:
         for test_case in test_cases:
             for metric in metrics:
-                if isinstance(test_case, LLMTestCase) and not isinstance(
+                if (isinstance(test_case, LLMTestCase)) and not isinstance(
                     metric, BaseMetric
                 ):
                     raise ValueError(
@@ -305,12 +291,6 @@ def validate_evaluate_inputs(
                     raise ValueError(
                         f"Metric {metric.__name__} is not a valid metric for ConversationalTestCase."
                     )
-                if isinstance(test_case, MLLMTestCase) and not isinstance(
-                    metric, BaseMultimodalMetric
-                ):
-                    raise ValueError(
-                        f"Metric {metric.__name__} is not a valid metric for MLLMTestCase."
-                    )
 
 
 def print_test_result(test_result: TestResult, display: TestRunResultDisplay):

deepeval/integrations/pydantic_ai/agent.py

@@ -1,12 +1,29 @@
 import warnings
+from typing import TYPE_CHECKING, Any
 
 try:
-    from pydantic_ai.agent import Agent
+    from pydantic_ai.agent import Agent as _BaseAgent
 
     is_pydantic_ai_installed = True
-except:
+except ImportError:
     is_pydantic_ai_installed = False
 
+    class _BaseAgent:
+        """Dummy fallback so imports don't crash when pydantic-ai is missing."""
+
+        def __init__(self, *args: Any, **kwargs: Any) -> None:
+            # No-op: for compatibility
+            pass
+
+
+if TYPE_CHECKING:
+    # For type checkers: use the real Agent if available.
+    from pydantic_ai.agent import Agent  # type: ignore[unused-ignore]
+else:
+    # At runtime we always have some base: real Agent or our dummy.
+    # This is just to avoid blow-ups.
+    Agent = _BaseAgent
+
 
 class DeepEvalPydanticAIAgent(Agent):
 

deepeval/integrations/pydantic_ai/instrumentator.py

@@ -1,40 +1,58 @@
+from __future__ import annotations
+
 import json
 import logging
 import os
 from time import perf_counter
-from typing import Literal, Optional, List
+from typing import Any, List, Optional, TYPE_CHECKING
 
 from deepeval.config.settings import get_settings
 from deepeval.confident.api import get_confident_api_key
 from deepeval.metrics.base_metric import BaseMetric
 from deepeval.prompt import Prompt
 from deepeval.tracing.context import current_trace_context
-from deepeval.tracing.types import Trace
-from deepeval.tracing.otel.utils import to_hex_string
-from deepeval.tracing.tracing import trace_manager
-from deepeval.tracing.otel.utils import normalize_pydantic_ai_messages
 from deepeval.tracing.otel.exporter import ConfidentSpanExporter
-
+from deepeval.tracing.otel.test_exporter import test_exporter
+from deepeval.tracing.otel.utils import (
+    normalize_pydantic_ai_messages,
+    to_hex_string,
+)
+from deepeval.tracing.perf_epoch_bridge import init_clock_bridge
+from deepeval.tracing.tracing import trace_manager
+from deepeval.tracing.types import (
+    AgentSpan,
+    Trace,
+    TraceSpanStatus,
+    ToolCall,
+)
 
 logger = logging.getLogger(__name__)
 
-
 try:
-    from pydantic_ai.models.instrumented import InstrumentationSettings
-    from opentelemetry.sdk.trace import SpanProcessor, TracerProvider
+    # Optional dependencies
+    from opentelemetry.sdk.trace import (
+        ReadableSpan as _ReadableSpan,
+        SpanProcessor as _SpanProcessor,
+        TracerProvider,
+    )
     from opentelemetry.sdk.trace.export import BatchSpanProcessor
     from opentelemetry.exporter.otlp.proto.http.trace_exporter import (
         OTLPSpanExporter,
     )
-    from opentelemetry.sdk.trace import ReadableSpan
+    from pydantic_ai.models.instrumented import (
+        InstrumentationSettings as _BaseInstrumentationSettings,
+    )
 
     dependency_installed = True
 except ImportError as e:
+    dependency_installed = False
+
+    # Preserve previous behavior: only log when verbose mode is enabled.
     if get_settings().DEEPEVAL_VERBOSE_MODE:
         if isinstance(e, ModuleNotFoundError):
             logger.warning(
                 "Optional tracing dependency not installed: %s",
-                e.name,
+                getattr(e, "name", repr(e)),
                 stacklevel=2,
             )
         else:
@@ -43,26 +61,47 @@ except ImportError as e:
                 e,
                 stacklevel=2,
             )
-    dependency_installed = False
+
+    # Dummy fallbacks so imports and class definitions don't crash when
+    # optional deps are missing. Actual use is still guarded by
+    # is_dependency_installed().
+    class _BaseInstrumentationSettings:
+        def __init__(self, *args: Any, **kwargs: Any) -> None:
+            pass
+
+    class _SpanProcessor:
+        def __init__(self, *args: Any, **kwargs: Any) -> None:
+            pass
+
+        def on_start(self, span: Any, parent_context: Any) -> None:
+            pass
+
+        def on_end(self, span: Any) -> None:
+            pass
+
+    class _ReadableSpan:
+        pass
 
 
-def is_dependency_installed():
+def is_dependency_installed() -> bool:
     if not dependency_installed:
         raise ImportError(
-            "Dependencies are not installed. Please install it with `pip install pydantic-ai opentelemetry-sdk opentelemetry-exporter-otlp-proto-http`."
+            "Dependencies are not installed. Please install it with "
+            "`pip install pydantic-ai opentelemetry-sdk "
+            "opentelemetry-exporter-otlp-proto-http`."
         )
     return True
 
 
-from deepeval.tracing.types import AgentSpan
-from deepeval.confident.api import get_confident_api_key
-from deepeval.prompt import Prompt
-from deepeval.tracing.otel.test_exporter import test_exporter
-from deepeval.tracing.context import current_trace_context
-from deepeval.tracing.types import Trace
-from deepeval.tracing.otel.utils import to_hex_string
-from deepeval.tracing.types import TraceSpanStatus, ToolCall
-from deepeval.tracing.perf_epoch_bridge import init_clock_bridge
+if TYPE_CHECKING:
+    # For type checkers, use real types
+    from opentelemetry.sdk.trace import ReadableSpan, SpanProcessor
+    from pydantic_ai.models.instrumented import InstrumentationSettings
+else:
+    # At runtime we always have something to subclass / annotate with
+    InstrumentationSettings = _BaseInstrumentationSettings
+    SpanProcessor = _SpanProcessor
+    ReadableSpan = _ReadableSpan
 
 # OTLP_ENDPOINT = "http://127.0.0.1:4318/v1/traces"
 OTLP_ENDPOINT = "https://otel.confident-ai.com/v1/traces"

deepeval/key_handler.py

@@ -99,7 +99,10 @@ class ModelKeyValues(Enum):
 class EmbeddingKeyValues(Enum):
     # Azure OpenAI
     USE_AZURE_OPENAI_EMBEDDING = "USE_AZURE_OPENAI_EMBEDDING"
+    # Azure OpenAI
+    AZURE_EMBEDDING_MODEL_NAME = "AZURE_EMBEDDING_MODEL_NAME"
     AZURE_EMBEDDING_DEPLOYMENT_NAME = "AZURE_EMBEDDING_DEPLOYMENT_NAME"
+
     # Local
     USE_LOCAL_EMBEDDINGS = "USE_LOCAL_EMBEDDINGS"
     LOCAL_EMBEDDING_MODEL_NAME = "LOCAL_EMBEDDING_MODEL_NAME"

deepeval/metrics/__init__.py

@@ -1,7 +1,6 @@
 from .base_metric import (
     BaseMetric,
     BaseConversationalMetric,
-    BaseMultimodalMetric,
     BaseArenaMetric,
 )
 
@@ -42,6 +41,16 @@ from .mcp_use_metric.mcp_use_metric import MCPUseMetric
 from .turn_relevancy.turn_relevancy import (
     TurnRelevancyMetric,
 )
+from .turn_faithfulness.turn_faithfulness import TurnFaithfulnessMetric
+from .turn_contextual_precision.turn_contextual_precision import (
+    TurnContextualPrecisionMetric,
+)
+from .turn_contextual_recall.turn_contextual_recall import (
+    TurnContextualRecallMetric,
+)
+from .turn_contextual_relevancy.turn_contextual_relevancy import (
+    TurnContextualRelevancyMetric,
+)
 from .conversation_completeness.conversation_completeness import (
     ConversationCompletenessMetric,
 )
@@ -55,13 +64,6 @@ from .multimodal_metrics import (
     ImageCoherenceMetric,
     ImageHelpfulnessMetric,
     ImageReferenceMetric,
-    MultimodalContextualRecallMetric,
-    MultimodalContextualRelevancyMetric,
-    MultimodalContextualPrecisionMetric,
-    MultimodalAnswerRelevancyMetric,
-    MultimodalFaithfulnessMetric,
-    MultimodalToolCorrectnessMetric,
-    MultimodalGEval,
 )
 
 
@@ -69,7 +71,6 @@ __all__ = [
     # Base classes
     "BaseMetric",
     "BaseConversationalMetric",
-    "BaseMultimodalMetric",
     "BaseArenaMetric",
     # Non-LLM metrics
     "ExactMatchMetric",
@@ -119,17 +120,14 @@ __all__ = [
     # Conversational metrics
     "TurnRelevancyMetric",
     "ConversationCompletenessMetric",
+    "TurnFaithfulnessMetric",
+    "TurnContextualPrecisionMetric",
+    "TurnContextualRecallMetric",
+    "TurnContextualRelevancyMetric",
     # Multimodal metrics
     "TextToImageMetric",
     "ImageEditingMetric",
     "ImageCoherenceMetric",
     "ImageHelpfulnessMetric",
     "ImageReferenceMetric",
-    "MultimodalContextualRecallMetric",
-    "MultimodalContextualRelevancyMetric",
-    "MultimodalContextualPrecisionMetric",
-    "MultimodalAnswerRelevancyMetric",
-    "MultimodalFaithfulnessMetric",
-    "MultimodalToolCorrectnessMetric",
-    "MultimodalGEval",
 ]