azure-ai-evaluation 1.4.0__py3-none-any.whl → 1.6.0__py3-none-any.whl

azure/ai/evaluation/simulator/_model_tools/_proxy_completion_model.py

@@ -6,12 +6,15 @@ import copy
 import json
 import time
 import uuid
-from typing import Any, Dict, List, Optional, cast
+from typing import Any, Dict, List, Optional, cast, Union
 
 from azure.ai.evaluation._http_utils import AsyncHttpPipeline, get_async_http_client
 from azure.ai.evaluation._user_agent import USER_AGENT
 from azure.core.exceptions import HttpResponseError
 from azure.core.pipeline.policies import AsyncRetryPolicy, RetryMode
+from azure.ai.evaluation._common.onedp._client import AIProjectClient
+from azure.ai.evaluation._common.onedp.models import SimulationDTO
+from azure.ai.evaluation._common.constants import RAIService
 
 from .._model_tools._template_handler import TemplateParameters
 from .models import OpenAIChatCompletionsModel
@@ -40,14 +43,14 @@ class SimulationRequestDTO:
         headers: Dict[str, str],
         payload: Dict[str, Any],
         params: Dict[str, str],
-        templatekey: str,
+        template_key: str,
         template_parameters: Optional[TemplateParameters],
     ):
         self.url = url
         self.headers = headers
         self.json = json.dumps(payload)
         self.params = params
-        self.templatekey = templatekey
+        self.template_key = template_key
         self.templateParameters = template_parameters
 
     def to_dict(self) -> Dict:
@@ -111,7 +114,7 @@ class ProxyChatCompletionsModel(OpenAIChatCompletionsModel):
     async def get_conversation_completion(
         self,
         messages: List[Dict],
-        session: AsyncHttpPipeline,
+        session: Union[AsyncHttpPipeline, AIProjectClient],
         role: str = "assistant",  # pylint: disable=unused-argument
         **request_params,
     ) -> dict:
@@ -142,7 +145,7 @@ class ProxyChatCompletionsModel(OpenAIChatCompletionsModel):
 
     async def request_api(
         self,
-        session: AsyncHttpPipeline,
+        session: Union[AsyncHttpPipeline, AIProjectClient],
         request_data: dict,
     ) -> dict:
         """
@@ -183,51 +186,72 @@ class ProxyChatCompletionsModel(OpenAIChatCompletionsModel):
             headers=headers,
             payload=request_data,
             params=params,
-            templatekey=self.tkey,
+            template_key=self.tkey,
             template_parameters=self.tparam,
         )
 
         time_start = time.time()
         full_response = None
 
-        response = await session.post(url=self.endpoint_url, headers=proxy_headers, json=sim_request_dto.to_dict())
-
-        if response.status_code != 202:
-            raise HttpResponseError(
-                message=f"Received unexpected HTTP status: {response.status_code} {response.text()}", response=response
+        if(isinstance(session, AIProjectClient)):
+            sim_request_dto = SimulationDTO(
+                headers=headers,
+                params=params,
+                json=json.dumps(request_data),
+                template_key=self.tkey,
+                template_parameters=self.tparam,
             )
-
-        response_data = response.json()
-        self.result_url = cast(str, response_data["location"])
-
-        retry_policy = AsyncRetryPolicy(  # set up retry configuration
-            retry_on_status_codes=[202],  # on which statuses to retry
-            retry_total=7,
-            retry_backoff_factor=10.0,
-            retry_backoff_max=180,
-            retry_mode=RetryMode.Exponential,
-        )
-
-        # initial 15 seconds wait before attempting to fetch result
-        # Need to wait both in this thread and in the async thread for some reason?
-        # Someone not under a crunch and with better async understandings should dig into this more.
-        await asyncio.sleep(15)
-        time.sleep(15)
-
-        async with get_async_http_client().with_policies(retry_policy=retry_policy) as exp_retry_client:
-            token = await self.token_manager.get_token_async()
-            proxy_headers = {
-                "Authorization": f"Bearer {token}",
-                "Content-Type": "application/json",
-                "User-Agent": USER_AGENT,
-            }
-            response = await exp_retry_client.get(  # pylint: disable=too-many-function-args,unexpected-keyword-arg
-                self.result_url, headers=proxy_headers
+            response_data = session.red_teams.submit_simulation(sim_request_dto, headers=headers, params=params)
+            operation_id = response_data["location"].split("/")[-1]
+            
+            request_count = 0
+            flag = True
+            while flag:
+                response = session.evaluations.operation_results(operation_id, headers=headers)
+                if response.status_code == 200:
+                    response_data = cast(List[Dict], response.json())
+                    flag = False
+                else:
+                    request_count += 1
+                    sleep_time = RAIService.SLEEP_TIME**request_count
+                    await asyncio.sleep(sleep_time)
+        else:
+            response = await session.post(url=self.endpoint_url, headers=proxy_headers, json=sim_request_dto.to_dict())
+            # response.raise_for_status()
+            if response.status_code != 202:
+                raise HttpResponseError(
+                    message=f"Received unexpected HTTP status: {response.status_code} {response.text()}", response=response
+                )
+            response_data = response.json()
+            
+            self.result_url = cast(str, response_data["location"])
+            retry_policy = AsyncRetryPolicy(  # set up retry configuration
+                retry_on_status_codes=[202],  # on which statuses to retry
+                retry_total=7,
+                retry_backoff_factor=10.0,
+                retry_backoff_max=180,
+                retry_mode=RetryMode.Exponential,
             )
 
-        response.raise_for_status()
-
-        response_data = response.json()
+            # initial 15 seconds wait before attempting to fetch result
+            # Need to wait both in this thread and in the async thread for some reason?
+            # Someone not under a crunch and with better async understandings should dig into this more.
+            await asyncio.sleep(15)
+            time.sleep(15)
+
+            async with get_async_http_client().with_policies(retry_policy=retry_policy) as exp_retry_client:
+                token = await self.token_manager.get_token_async()
+                proxy_headers = {
+                    "Authorization": f"Bearer {token}",
+                    "Content-Type": "application/json",
+                    "User-Agent": USER_AGENT,
+                }
+                response = await exp_retry_client.get(  # pylint: disable=too-many-function-args,unexpected-keyword-arg
+                    self.result_url, headers=proxy_headers
+                )
+            response.raise_for_status()
+            response_data = response.json()
+            
         self.logger.info("Response: %s", response_data)
 
         # Copy the full response and return it to be saved in jsonl.

azure/ai/evaluation/simulator/_model_tools/_template_handler.py

@@ -2,11 +2,12 @@
 # Copyright (c) Microsoft Corporation. All rights reserved.
 # ---------------------------------------------------------
 
-from typing import Dict, List, Optional, TypedDict, cast
-
+from typing import Dict, List, Optional, TypedDict, cast, Union
+from ast import literal_eval
 from typing_extensions import NotRequired
 
 from azure.ai.evaluation._model_configurations import AzureAIProject
+from azure.ai.evaluation._common.onedp._client import AIProjectClient
 
 from ._rai_client import RAIClient
 
@@ -153,7 +154,7 @@ class AdversarialTemplateHandler:
     :type rai_client: ~azure.ai.evaluation.simulator._model_tools.RAIClient
     """
 
-    def __init__(self, azure_ai_project: AzureAIProject, rai_client: RAIClient) -> None:
+    def __init__(self, azure_ai_project: Union[str, AzureAIProject], rai_client: Union[RAIClient, AIProjectClient]) -> None:
         self.azure_ai_project = azure_ai_project
         self.categorized_ch_parameters: Optional[Dict[str, _CategorizedParameter]] = None
         self.rai_client = rai_client
@@ -163,8 +164,11 @@ class AdversarialTemplateHandler:
             categorized_parameters: Dict[str, _CategorizedParameter] = {}
             util = ContentHarmTemplatesUtils
 
-            parameters = await self.rai_client.get_contentharm_parameters()
-
+            if isinstance(self.rai_client, RAIClient):
+                parameters = await self.rai_client.get_contentharm_parameters()
+            elif isinstance(self.rai_client, AIProjectClient):
+                parameters = literal_eval(self.rai_client.red_teams.get_template_parameters())
+            
             for k in parameters.keys():
                 template_key = util.get_template_key(k)
                 categorized_parameters[template_key] = {

azure/ai/evaluation/simulator/_model_tools/models.py

@@ -12,6 +12,8 @@ from abc import ABC, abstractmethod
 from collections import deque
 from typing import Deque, Dict, List, Optional, Union
 from urllib.parse import urlparse
+from azure.ai.evaluation._common.onedp._client import AIProjectClient
+from ._rai_client import RAIClient
 
 from azure.ai.evaluation._exceptions import ErrorBlame, ErrorCategory, ErrorTarget, EvaluationException
 from azure.ai.evaluation._http_utils import AsyncHttpPipeline
@@ -78,7 +80,7 @@ class LLMBase(ABC):
     async def get_completion(
         self,
         prompt: str,
-        session: AsyncHttpPipeline,
+        session: Union[AsyncHttpPipeline, AIProjectClient],
         **request_params,
     ) -> dict:
         """
@@ -100,7 +102,7 @@ class LLMBase(ABC):
     async def get_all_completions(
         self,
         prompts: List[str],
-        session: AsyncHttpPipeline,
+        session: Union[AsyncHttpPipeline, AIProjectClient],
         api_call_max_parallel_count: int,
         api_call_delay_seconds: float,
         request_error_rate_threshold: float,
@@ -120,7 +122,7 @@ class LLMBase(ABC):
     async def get_conversation_completion(
         self,
         messages: List[dict],
-        session: AsyncHttpPipeline,
+        session: Union[AsyncHttpPipeline, AIProjectClient],
         role: str,
         **request_params,
     ) -> dict:
@@ -274,7 +276,7 @@ class OpenAICompletionsModel(LLMBase):
     async def get_conversation_completion(
         self,
         messages: List[dict],
-        session: AsyncHttpPipeline,
+        session: Union[AsyncHttpPipeline, AIProjectClient],
         role: str = "assistant",
         **request_params,
     ) -> dict:
@@ -304,7 +306,7 @@ class OpenAICompletionsModel(LLMBase):
     async def get_all_completions(  # type: ignore[override]
         self,
         prompts: List[Dict[str, str]],
-        session: AsyncHttpPipeline,
+        session: Union[AsyncHttpPipeline, AIProjectClient],
         api_call_max_parallel_count: int = 1,
         api_call_delay_seconds: float = 0.1,
         request_error_rate_threshold: float = 0.5,
@@ -372,7 +374,7 @@ class OpenAICompletionsModel(LLMBase):
         self,
         request_datas: List[dict],
         output_collector: List,
-        session: AsyncHttpPipeline,
+        session: Union[AsyncHttpPipeline, AIProjectClient],
         api_call_delay_seconds: float = 0.1,
         request_error_rate_threshold: float = 0.5,
     ) -> None:
@@ -433,7 +435,7 @@ class OpenAICompletionsModel(LLMBase):
 
     async def request_api(
         self,
-        session: AsyncHttpPipeline,
+        session: Union[AsyncHttpPipeline, AIProjectClient],
         request_data: dict,
     ) -> dict:
         """
@@ -476,11 +478,12 @@ class OpenAICompletionsModel(LLMBase):
         time_start = time.time()
         full_response = None
 
-        response = await session.post(url=self.endpoint_url, headers=headers, json=request_data, params=params)
-
-        response.raise_for_status()
-
-        response_data = response.json()
+        if(isinstance(session, AIProjectClient)):
+            response_data = session.red_teams.submit_simulation(request_data, headers, params)
+        else:
+            response = await session.post(url=self.endpoint_url, headers=headers, json=request_data, params=params)
+            response.raise_for_status()
+            response_data = response.json()
 
         self.logger.info(f"Response: {response_data}")
 
@@ -533,7 +536,7 @@ class OpenAIChatCompletionsModel(OpenAICompletionsModel):
     async def get_conversation_completion(
         self,
         messages: List[dict],
-        session: AsyncHttpPipeline,
+        session: Union[AsyncHttpPipeline, AIProjectClient],
         role: str = "assistant",
         **request_params,
     ) -> dict:
@@ -544,7 +547,7 @@ class OpenAIChatCompletionsModel(OpenAICompletionsModel):
         ----------
         messages: List of messages to query the model with.
         Expected format: [{"role": "user", "content": "Hello!"}, ...]
-        session: AsyncHttpPipeline object to query the model with.
+        session: Union[AsyncHttpPipeline, AIProjectClient] object to query the model with.
         role: Not used for this model, since it is a chat model.
         request_params: Additional parameters to pass to the model.
         """
@@ -560,7 +563,7 @@ class OpenAIChatCompletionsModel(OpenAICompletionsModel):
     async def get_completion(
         self,
         prompt: str,
-        session: AsyncHttpPipeline,
+        session: Union[AsyncHttpPipeline, AIProjectClient],
         **request_params,
     ) -> dict:
         """
@@ -569,7 +572,7 @@ class OpenAIChatCompletionsModel(OpenAICompletionsModel):
         Parameters
         ----------
         prompt: Prompt str to query model with.
-        session: AsyncHttpPipeline object to use for the request.
+        session: Union[AsyncHttpPipeline, AIProjectClient] object to use for the request.
         **request_params: Additional parameters to pass to the request.
         """
         messages = [{"role": "system", "content": prompt}]
@@ -583,7 +586,7 @@ class OpenAIChatCompletionsModel(OpenAICompletionsModel):
     async def get_all_completions(
         self,
         prompts: List[str],  # type: ignore[override]
-        session: AsyncHttpPipeline,
+        session: Union[AsyncHttpPipeline, AIProjectClient],
         api_call_max_parallel_count: int = 1,
         api_call_delay_seconds: float = 0.1,
         request_error_rate_threshold: float = 0.5,

azure/ai/evaluation/simulator/_simulator.py

@@ -11,7 +11,7 @@ import re
 import warnings
 from typing import Any, Callable, Dict, List, Optional, Union, Tuple
 
-from promptflow.core import AsyncPrompty
+from azure.ai.evaluation._legacy._adapters._flows import AsyncPrompty
 from tqdm import tqdm
 
 from azure.ai.evaluation._common._experimental import experimental

{azure_ai_evaluation-1.4.0.dist-info → azure_ai_evaluation-1.6.0.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: azure-ai-evaluation
-Version: 1.4.0
+Version: 1.6.0
 Summary: Microsoft Azure Evaluation Library for Python
 Home-page: https://github.com/Azure/azure-sdk-for-python
 Author: Microsoft Corporation
@@ -28,8 +28,14 @@ Requires-Dist: azure-identity>=1.16.0
 Requires-Dist: azure-core>=1.30.2
 Requires-Dist: nltk>=3.9.1
 Requires-Dist: azure-storage-blob>=12.10.0
+Requires-Dist: httpx>=0.25.1
+Requires-Dist: pandas<3.0.0,>=2.1.2
+Requires-Dist: openai>=1.73.0
+Requires-Dist: ruamel.yaml<1.0.0,>=0.17.10
+Requires-Dist: msrest>=0.6.21
+Requires-Dist: Jinja2>=3.1.6
 Provides-Extra: redteam
-Requires-Dist: pyrit>=0.8.0; extra == "redteam"
+Requires-Dist: pyrit==0.8.1; extra == "redteam"
 
 # Azure AI Evaluation client library for Python
 
@@ -376,6 +382,34 @@ This project has adopted the [Microsoft Open Source Code of Conduct][code_of_con
 
 # Release History
 
+## 1.6.0 (2025-05-07)
+
+### Features Added
+- New `<evaluator>.binary_aggregate` field added to evaluation result metrics. This field contains the aggregated binary evaluation results for each evaluator, providing a summary of the evaluation outcomes.
+- Added support for Azure Open AI evaluation via 4 new 'grader' classes, which serve as wrappers around Azure Open AI grader configurations. These new grader objects can be supplied to the main `evaluate` method as if they were normal callable evaluators. The new classes are:
+    - AzureOpenAIGrader (general class for experienced users)
+    - AzureOpenAILabelGrader
+    - AzureOpenAIStringCheckGrader
+    - AzureOpenAITextSimilarityGrader
+
+### Breaking Changes
+- In the experimental RedTeam's scan method, the `data_only` param has been replaced with `skip_evals` and if you do not want data to be uploaded, use the `skip_upload` flag.
+
+### Bugs Fixed
+- Fixed error in `evaluate` where data fields could not contain numeric characters. Previously, a data file with schema:
+    ```
+    "query1": "some query", "response": "some response"
+    ```
+    throws error when passed into `evaluator_config` as `{"evaluator_name": {"column_mapping": {"query": "${data.query1}", "response": "${data.response}"}},}`.
+    Now, users may import data containing fields with numeric characters. 
+
+
+## 1.5.0 (2025-04-04)
+
+### Features Added
+
+- New `RedTeam` agent functionality to assess the safety and resilience of AI systems against adversarial prompt attacks
+
 ## 1.4.0 (2025-03-27)
 
 ### Features Added