Add possibility to select logging level.

src/promptflow-core/promptflow/_utils/logger_utils.py

@@ -182,7 +182,7 @@ def get_pf_logging_level(default=logging.INFO):
 
 def get_logger(name: str) -> logging.Logger:
     """Get logger used during execution."""
-    logger = logging.Logger(name)
+    logger = logging.getLogger(name)
     logger.setLevel(get_pf_logging_level())
     logger.addHandler(FileHandlerConcurrentWrapper())
     stdout_handler = logging.StreamHandler(sys.stdout)
@@ -207,6 +207,19 @@ def get_logger(name: str) -> logging.Logger:
 service_logger = get_logger("execution.service")
 
 
+def update_logger_levels(log_level: Optional[str] = None) -> None:
+    """
+    Update the logger levels.
+
+    :param log_level: The new logging level. If it is None,
+                      logging level will be taken from
+                      using get_pf_logging_level.
+    :type log_level: Optional[str]
+    """
+    for log in [flow_logger, bulk_logger, logger, service_logger]:
+        log.setLevel(log_level or get_pf_logging_level())
+
+
 logger_contexts = []
 
 

src/promptflow-core/promptflow/core/_serving/flow_invoker.py

@@ -2,6 +2,7 @@
 # Copyright (c) Microsoft Corporation. All rights reserved.
 # ---------------------------------------------------------
 import dataclasses
+import logging
 import os
 from pathlib import Path
 from typing import Callable, Union
@@ -59,7 +60,10 @@ def __init__(
         init_kwargs: dict = None,
         **kwargs,
     ):
-        self.logger = kwargs.get("logger", LoggerFactory.get_logger("flowinvoker"))
+        self.logger = kwargs.get(
+            "logger",
+            LoggerFactory.get_logger("flowinvoker",
+                                     verbosity=kwargs.get('log_level') or logging.INFO))
         self._init_kwargs = init_kwargs or {}
         self.logger.debug(f"Init flow invoker with init kwargs: {self._init_kwargs}")
         # TODO: avoid to use private attribute after we finalize the inheritance
@@ -123,7 +127,8 @@ def _init_connections(self, connection_provider):
                 connection_names=self.flow.get_connection_names(
                     environment_variables_overrides=os.environ,
                 ),
-                provider=ConnectionProvider.init_from_provider_config(connection_provider, credential=self._credential),
+                provider=ConnectionProvider.init_from_provider_config(
+                    connection_provider, credential=self._credential),
                 connections_to_ignore=connections_to_ignore,
                 # fetch connections with name override
                 connections_to_add=list(self.connections_name_overrides.values()),

src/promptflow-devkit/promptflow/_sdk/entities/_flows/_flow_context_resolver.py

@@ -6,7 +6,7 @@
 from functools import lru_cache
 from os import PathLike
 from pathlib import Path
-from typing import Dict, Union
+from typing import Dict, Optional, Union
 
 from promptflow._sdk._configuration import Configuration
 from promptflow._sdk._constants import NODES
@@ -37,19 +37,20 @@ def __init__(self, flow_path: PathLike):
 
     @classmethod
     @lru_cache
-    def resolve(cls, flow: Flow) -> "FlowInvoker":
+    def resolve(cls, flow: Flow, log_level: Optional[int] = None) -> "FlowInvoker":
         """Resolve flow to flow invoker."""
         resolver = cls(flow_path=flow.path)
         resolver._resolve(flow_context=flow.context)
-        return resolver._create_invoker(flow_context=flow.context)
+        return resolver._create_invoker(flow_context=flow.context, log_level=log_level)
 
     @classmethod
     @lru_cache
-    def resolve_async_invoker(cls, flow: Flow) -> "AsyncFlowInvoker":
+    def resolve_async_invoker(cls, flow: Flow, log_level: Optional[int] = None) -> "AsyncFlowInvoker":
         """Resolve flow to flow invoker."""
         resolver = cls(flow_path=flow.path)
         resolver._resolve(flow_context=flow.context)
-        return resolver._create_invoker(flow_context=flow.context, is_async_call=True)
+        return resolver._create_invoker(flow_context=flow.context, is_async_call=True,
+                                        log_level=log_level)
 
     def _resolve(self, flow_context: FlowContext):
         """Resolve flow context."""
@@ -113,7 +114,8 @@ def _resolve_connection_objs(self, flow_context: FlowContext):
         return connections
 
     def _create_invoker(
-        self, flow_context: FlowContext, is_async_call=False
+        self, flow_context: FlowContext, is_async_call=False,
+        log_level: Optional[int] = None
     ) -> Union["FlowInvoker", "AsyncFlowInvoker"]:
         from promptflow.core._serving.flow_invoker import AsyncFlowInvoker, FlowInvoker
 
@@ -132,11 +134,13 @@ def _create_invoker(
                     flow=resolved_flow,
                     connections=connections,
                     streaming=flow_context.streaming,
+                    log_level=log_level,
                 )
             else:
                 return FlowInvoker(
                     flow=resolved_flow,
                     connections=connections,
                     streaming=flow_context.streaming,
                     connection_provider=Configuration.get_instance().get_connection_provider(),
+                    log_level=log_level,
                 )

src/promptflow-devkit/promptflow/_sdk/entities/_flows/base.py

@@ -9,6 +9,7 @@
 from promptflow._constants import DEFAULT_ENCODING, FLOW_FILE_SUFFIX
 from promptflow._sdk.entities._validation import SchemaValidatableMixin
 from promptflow._utils.flow_utils import is_flex_flow, is_prompty_flow, resolve_flow_path
+from promptflow._utils.logger_utils import update_logger_levels
 from promptflow._utils.yaml_utils import load_yaml_string
 from promptflow.core._flow import AbstractFlowBase
 from promptflow.exceptions import UserErrorException
@@ -145,6 +146,7 @@ def __init__(
         **kwargs,
     ):
         self.variant = kwargs.pop("variant", None) or {}
+        self._log_level = kwargs.pop("log_level", None)
         super().__init__(data=dag, code=code, path=path, **kwargs)
 
     @property
@@ -236,14 +238,16 @@ def __call__(self, *args, **kwargs):
         if args:
             raise UserErrorException("Flow can only be called with keyword arguments.")
 
+        if self._log_level:
+            update_logger_levels(self._log_level)
         result = self.invoke(inputs=kwargs)
         return result.output
 
     def invoke(self, inputs: dict) -> "LineResult":
         """Invoke a flow and get a LineResult object."""
         from promptflow._sdk.entities._flows._flow_context_resolver import FlowContextResolver
 
-        invoker = FlowContextResolver.resolve(flow=self)
+        invoker = FlowContextResolver.resolve(flow=self, log_level=self._log_level)
         result = invoker._invoke(
             data=inputs,
         )

src/promptflow-evals/promptflow/evals/evaluate/_evaluate.py

@@ -3,7 +3,7 @@
 # ---------------------------------------------------------
 import inspect
 from types import FunctionType
-from typing import Callable, Dict, Optional
+from typing import Callable, Dict, Optional, Union, cast
 
 import pandas as pd
 
@@ -103,23 +103,27 @@ def evaluate(
     code_client = CodeClient()
 
     evaluator_info = {}
-
-    for evaluator_name, evaluator in evaluators.items():
-        if isinstance(evaluator, FunctionType):
-            evaluator_info.update({evaluator_name: {"client": pf_client, "evaluator": evaluator}})
-        else:
-            evaluator_info.update({evaluator_name: {"client": code_client, "evaluator": evaluator}})
-
-        evaluator_info[evaluator_name]["run"] = evaluator_info[evaluator_name]["client"].run(
-            flow=evaluator,
-            column_mapping=evaluator_config.get(evaluator_name, evaluator_config.get("default", None)),
-            data=data,
-            stream=True,
-        )
+    if evaluator_config is None:
+        evaluator_config = {}
+
+    if evaluators:
+        for evaluator_name, evaluator in evaluators.items():
+            if isinstance(evaluator, FunctionType):
+                evaluator_info.update({evaluator_name: {"client": pf_client, "evaluator": evaluator}})
+            else:
+                evaluator_info.update({evaluator_name: {"client": code_client, "evaluator": evaluator}})
+
+            evaluator_info[evaluator_name]["run"] = evaluator_info[evaluator_name]["client"].run(
+                flow=evaluator,
+                column_mapping=evaluator_config.get(evaluator_name, evaluator_config.get("default", None)),
+                data=data,
+                stream=True,
+            )
 
     evaluators_result_df = None
     for evaluator_name, evaluator_info in evaluator_info.items():
-        evaluator_result_df = evaluator_info["client"].get_details(evaluator_info["run"], all_results=True)
+        evaluator_result_df = cast(
+            Union[PFClient, CodeClient], evaluator_info["client"]).get_details(evaluator_info["run"], all_results=True)
 
         # drop input columns
         evaluator_result_df = evaluator_result_df.drop(

src/promptflow-evals/promptflow/evals/evaluators/chat/__init__.py

@@ -6,7 +6,7 @@
 
 from promptflow.connections import AzureOpenAIConnection
 from promptflow.evals.evaluators import GroundednessEvaluator, RelevanceEvaluator, CoherenceEvaluator, FluencyEvaluator
-from typing import List, Dict
+from typing import Any, Dict, List
 from concurrent.futures import ThreadPoolExecutor, as_completed
 import json
 import logging
@@ -170,8 +170,8 @@ def _evaluate_turn(self, turn_num, questions, answers, contexts, evaluator):
             return {}
 
     def _aggregate_results(self, per_turn_results: List[Dict]):
-        scores = {}
-        reasons = {}
+        scores: Dict[str, Any] = {}
+        reasons: Dict[str, Any] = {}
 
         for turn in per_turn_results:
             for metric, value in turn.items():
@@ -184,7 +184,7 @@ def _aggregate_results(self, per_turn_results: List[Dict]):
                         scores[metric] = []
                     scores[metric].append(value)
 
-        aggregated = {}
+        aggregated: Dict[str, Any] = {}
         evaluation_per_turn = {}
 
         for metric, values in scores.items():

src/promptflow-evals/promptflow/evals/evaluators/coherence/flow/parse_score.py

@@ -2,9 +2,11 @@
 import numpy as np
 import re
 
+from typing import Optional
+
 
 @tool
-def parse_score(llm_output: str = None):
+def parse_score(llm_output: Optional[str] = None):
     score = np.nan
     if llm_output:
         match = re.search(r'\d', llm_output)

src/promptflow-evals/promptflow/evals/evaluators/content_safety/flow/evaluate_with_rai_service.py

@@ -1,7 +1,7 @@
 from promptflow.core import tool
 import time
 from constants import RAIService, Tasks
-from typing import List
+from typing import Any, Dict, List
 import numpy as np
 import re
 from utils import get_harm_severity_level
@@ -68,7 +68,7 @@ def fetch_result(operation_id: str, rai_svc_url: str, credential: TokenCredentia
         time.sleep(sleep_time)
 
 
-def parse_response(batch_response: List[dict], metric_name: str) -> List[List[dict]]:
+def parse_response(batch_response: List[dict], metric_name: str) -> Dict[str, Any]:
     # Fix the metric name if it's "hate_fairness"
     # Eventually we will remove this fix once the RAI service is updated
     key = metric_name

src/promptflow-evals/promptflow/evals/evaluators/content_safety/flow/utils.py

@@ -1,8 +1,10 @@
 import constants
 import numpy as np
 
+from typing import Union
 
-def get_harm_severity_level(harm_score: int) -> str:
+
+def get_harm_severity_level(harm_score: int) -> Union[str, float]:
     HARM_SEVERITY_LEVEL_MAPPING = {constants.HarmSeverityLevel.Safe: [0, 1],
                                    constants.HarmSeverityLevel.Low: [2, 3],
                                    constants.HarmSeverityLevel.Medium: [4, 5],

src/promptflow-evals/promptflow/evals/evaluators/content_safety/hate_unfairness.py

@@ -6,7 +6,8 @@
 
 
 class HateUnfairnessEvaluator:
-    def __init__(self, project_scope: dict, credential: Optional[TokenCredential] = None):
+    def __init__(self, project_scope: dict, credential: Optional[TokenCredential] = None,
+                 log_level: Optional[int] = None):
         """
         Initialize an evaluator for hate unfairness score.
 
@@ -15,6 +16,8 @@ def __init__(self, project_scope: dict, credential: Optional[TokenCredential] =
         :type project_scope: dict
         :param credential: The credential for connecting to Azure AI project.
         :type credential: TokenCredential
+        :param log_level: The logging level.
+        :type log_level: Optional[int]
 
         **Usage**
 
@@ -30,7 +33,7 @@ def __init__(self, project_scope: dict, credential: Optional[TokenCredential] =
         # Load the flow as function
         current_dir = Path(__file__).resolve().parent
         flow_dir = current_dir / "flow"
-        self._flow = load_flow(source=flow_dir)
+        self._flow = load_flow(source=flow_dir, log_level=log_level)
 
     def __call__(self, *, question: str, answer: str, **kwargs):
         """Evaluates hate unfairness content.

src/promptflow-evals/promptflow/evals/evaluators/content_safety/self_harm.py

@@ -6,7 +6,8 @@
 
 
 class SelfHarmEvaluator:
-    def __init__(self, project_scope: dict, credential: Optional[TokenCredential] = None):
+    def __init__(self, project_scope: dict, credential: Optional[TokenCredential] = None,
+                 log_level: Optional[int] = None):
         """
         Initialize an evaluator for self harm score.
 
@@ -15,6 +16,8 @@ def __init__(self, project_scope: dict, credential: Optional[TokenCredential] =
         :type project_scope: dict
         :param credential: The credential for connecting to Azure AI project.
         :type credential: TokenCredential
+        :param log_level: The logging level.
+        :type log_level: Optional[int]
 
         **Usage**
 
@@ -30,7 +33,7 @@ def __init__(self, project_scope: dict, credential: Optional[TokenCredential] =
         # Load the flow as function
         current_dir = Path(__file__).resolve().parent
         flow_dir = current_dir / "flow"
-        self._flow = load_flow(source=flow_dir)
+        self._flow = load_flow(source=flow_dir, log_level=log_level)
 
     def __call__(self, *, question: str, answer: str, **kwargs):
         """Evaluates self harm content.

src/promptflow-evals/promptflow/evals/evaluators/content_safety/sexual.py

@@ -6,7 +6,8 @@
 
 
 class SexualEvaluator:
-    def __init__(self, project_scope: dict, credential: Optional[TokenCredential] = None):
+    def __init__(self, project_scope: dict, credential: Optional[TokenCredential] = None,
+                 log_level: Optional[int] = None):
         """
         Initialize an evaluator for sexual score.
 
@@ -15,6 +16,8 @@ def __init__(self, project_scope: dict, credential: Optional[TokenCredential] =
         :type project_scope: dict
         :param credential: The credential for connecting to Azure AI project.
         :type credential: TokenCredential
+        :param log_level: The logging level.
+        :type log_level: Optional[int]
 
         **Usage**
 
@@ -30,7 +33,7 @@ def __init__(self, project_scope: dict, credential: Optional[TokenCredential] =
         # Load the flow as function
         current_dir = Path(__file__).resolve().parent
         flow_dir = current_dir / "flow"
-        self._flow = load_flow(source=flow_dir)
+        self._flow = load_flow(source=flow_dir, log_level=log_level)
 
     def __call__(self, *, question: str, answer: str, **kwargs):
         """Evaluates sexual content.

src/promptflow-evals/promptflow/evals/evaluators/content_safety/violence.py

@@ -6,7 +6,8 @@
 
 
 class ViolenceEvaluator:
-    def __init__(self, project_scope: dict, credential: Optional[TokenCredential] = None):
+    def __init__(self, project_scope: dict, credential: Optional[TokenCredential] = None,
+                 log_level: Optional[int] = None):
         """
         Initialize an evaluator for violence score.
 
@@ -15,6 +16,8 @@ def __init__(self, project_scope: dict, credential: Optional[TokenCredential] =
         :type project_scope: dict
         :param credential: The credential for connecting to Azure AI project.
         :type credential: TokenCredential
+        :param log_level: The logging level.
+        :type log_level: Optional[int]
 
         **Usage**
 
@@ -30,7 +33,7 @@ def __init__(self, project_scope: dict, credential: Optional[TokenCredential] =
         # Load the flow as function
         current_dir = Path(__file__).resolve().parent
         flow_dir = current_dir / "flow"
-        self._flow = load_flow(source=flow_dir)
+        self._flow = load_flow(source=flow_dir, log_level=log_level)
 
     def __call__(self, *, question: str, answer: str, **kwargs):
         """Evaluates violence content.