merge main

Author: YoYoJa

.github/CODEOWNERS

@@ -840,6 +840,7 @@
 /eng/common/                                                         @Azure/azure-sdk-eng
 /.github/workflows/                                                  @Azure/azure-sdk-eng
 /.github/CODEOWNERS                                                  @lmazuel @xiangyan99 @kashifkhan @Azure/azure-sdk-eng
+/.github/copilot-instructions.md                                     @l0lawrence @praveenkuttappan @maririos
 /sdk/pullrequest.yml                                                 @scbedd @weshaggard @benbp
 
 /.config/1espt/                                                      @benbp @weshaggard

eng/common/scripts/Update-DocsMsPackages.ps1

@@ -26,7 +26,9 @@ param (
   [string] $DocRepoLocation, # the location of the cloned doc repo
 
   [Parameter(Mandatory = $false)]
-  [string] $PackageSourceOverride
+  [string] $PackageSourceOverride,
+
+  [switch] $SkipLegacyOnboarding
 )
 
 . (Join-Path $PSScriptRoot common.ps1)
@@ -57,7 +59,11 @@ function PackageIsValidForDocsOnboarding($package) {
     -DocRepoLocation $DocRepoLocation
 }
 
-$MONIKERS = @('latest', 'preview', 'legacy')
+$MONIKERS = @('latest', 'preview') 
+if (!$SkipLegacyOnboarding) {
+  $MONIKERS += 'legacy'
+}
+
 foreach ($moniker in $MONIKERS) {
   try {
     Write-Host "Onboarding packages for moniker: $moniker"

eng/tox/install_depend_packages.py

@@ -41,7 +41,7 @@
     "azure-core": "1.11.0",
     "requests": "2.19.0",
     "six": "1.12.0",
-    "cryptography": "38.0.3",
+    "cryptography": "41.0.0",
     "msal": "1.23.0",
     "azure-storage-file-datalake": "12.2.0",
 }

sdk/evaluation/azure-ai-evaluation/azure/ai/evaluation/_evaluate/_evaluate.py

@@ -963,7 +963,7 @@ def _evaluate(  # pylint: disable=too-many-locals,too-many-statements
     if need_oai_run:
         try:
             aoi_name = evaluation_name if evaluation_name else DEFAULT_OAI_EVAL_RUN_NAME
-            eval_run_info_list = _begin_aoai_evaluation(graders, column_mapping, input_data_df, aoi_name)
+            eval_run_info_list = _begin_aoai_evaluation(graders, column_mapping, input_data_df, aoi_name, **kwargs)
             need_get_oai_results = len(eval_run_info_list) > 0
         except EvaluationException as e:
             if need_local_run:

sdk/evaluation/azure-ai-evaluation/azure/ai/evaluation/_evaluate/_evaluate_aoai.py

@@ -79,6 +79,7 @@ def _begin_aoai_evaluation(
     column_mappings: Optional[Dict[str, Dict[str, str]]],
     data: pd.DataFrame,
     run_name: str,
+    **kwargs: Any,
 ) -> List[OAIEvalRunCreationInfo]:
     """
     Use the AOAI SDK to start an evaluation of the inputted dataset against the supplied graders.
@@ -114,15 +115,19 @@ def _begin_aoai_evaluation(
             f"AOAI: Starting evaluation run {idx + 1}/{len(grader_mapping_list)} with {len(selected_graders)} grader(s)..."
         )
         all_eval_run_info.append(
-            _begin_single_aoai_evaluation(selected_graders, data, selected_column_mapping, run_name)
+            _begin_single_aoai_evaluation(selected_graders, data, selected_column_mapping, run_name, **kwargs)
         )
 
     LOGGER.info(f"AOAI: Successfully created {len(all_eval_run_info)} evaluation run(s).")
     return all_eval_run_info
 
 
 def _begin_single_aoai_evaluation(
-    graders: Dict[str, AzureOpenAIGrader], data: pd.DataFrame, column_mapping: Optional[Dict[str, str]], run_name: str
+    graders: Dict[str, AzureOpenAIGrader],
+    data: pd.DataFrame,
+    column_mapping: Optional[Dict[str, str]],
+    run_name: str,
+    **kwargs: Any,
 ) -> OAIEvalRunCreationInfo:
     """
     Use the AOAI SDK to start an evaluation of the inputted dataset against the supplied graders.
@@ -146,6 +151,16 @@ def _begin_single_aoai_evaluation(
     LOGGER.info(f"AOAI: Preparing evaluation for {len(graders)} grader(s): {list(graders.keys())}")
     grader_name_list = []
     grader_list = []
+
+    data_source = {}
+    data_source_config = {}
+
+    if kwargs.get("data_source_config") is not None:
+        data_source_config = kwargs.get("data_source_config")
+
+    if kwargs.get("data_source") is not None:
+        data_source = kwargs.get("data_source")
+
     # It's expected that all graders supplied for a single eval run use the same credentials
     # so grab a client from the first grader.
     client = list(graders.values())[0].get_client()
@@ -155,7 +170,8 @@ def _begin_single_aoai_evaluation(
         grader_list.append(grader._grader_config)
     effective_column_mapping: Dict[str, str] = column_mapping or {}
     LOGGER.info(f"AOAI: Generating data source config with {len(effective_column_mapping)} column mapping(s)...")
-    data_source_config = _generate_data_source_config(data, effective_column_mapping)
+    if data_source_config == {}:
+        data_source_config = _generate_data_source_config(data, effective_column_mapping)
     LOGGER.info(f"AOAI: Data source config generated with schema type: {data_source_config.get('type')}")
 
     # Create eval group
@@ -181,7 +197,7 @@ def _begin_single_aoai_evaluation(
 
     # Create eval run
     LOGGER.info(f"AOAI: Creating eval run '{run_name}' with {len(data)} data rows...")
-    eval_run_id = _begin_eval_run(client, eval_group_info.id, run_name, data, effective_column_mapping)
+    eval_run_id = _begin_eval_run(client, eval_group_info.id, run_name, data, effective_column_mapping, data_source)
     LOGGER.info(
         f"AOAI: Eval run created with id {eval_run_id}."
         + " Results will be retrieved after normal evaluation is complete..."
@@ -654,7 +670,10 @@ def _generate_data_source_config(input_data_df: pd.DataFrame, column_mapping: Di
         props = data_source_config["item_schema"]["properties"]
         req = data_source_config["item_schema"]["required"]
         for key in column_mapping.keys():
-            props[key] = {"type": "string"}
+            if key in input_data_df and len(input_data_df[key]) > 0 and isinstance(input_data_df[key].iloc[0], list):
+                props[key] = {"type": "array"}
+            else:
+                props[key] = {"type": "string"}
             req.append(key)
         LOGGER.info(f"AOAI: Flat schema generated with {len(props)} properties: {list(props.keys())}")
         return data_source_config
@@ -821,8 +840,11 @@ def _convert_value_to_string(val: Any) -> str:
             # Safely fetch value
             val = row.get(df_col, None)
 
-            # Convert value to string to match schema's "type": "string" leaves.
-            str_val = _convert_value_to_string(val)
+            if isinstance(val, list):
+                str_val = val
+            else:
+                # Convert value to string to match schema's "type": "string" leaves.
+                str_val = _convert_value_to_string(val)
 
             # Insert into nested dict
             cursor = item_root
@@ -842,7 +864,10 @@ def _convert_value_to_string(val: Any) -> str:
         for col_name in input_data_df.columns:
             if col_name not in processed_cols:
                 val = row.get(col_name, None)
-                str_val = _convert_value_to_string(val)
+                if isinstance(val, list):
+                    str_val = val
+                else:
+                    str_val = _convert_value_to_string(val)
                 item_root[col_name] = str_val
 
         content.append({WRAPPER_KEY: item_root})
@@ -863,6 +888,7 @@ def _begin_eval_run(
     run_name: str,
     input_data_df: pd.DataFrame,
     column_mapping: Dict[str, str],
+    data_source_params: Optional[Dict[str, Any]] = None,
 ) -> str:
     """
     Given an eval group id and a dataset file path, use the AOAI API to
@@ -884,6 +910,10 @@ def _begin_eval_run(
 
     LOGGER.info(f"AOAI: Creating eval run '{run_name}' for eval group {eval_group_id}...")
     data_source = _get_data_source(input_data_df, column_mapping)
+
+    if data_source_params is not None:
+        data_source.update(data_source_params)
+
     eval_run = client.evals.runs.create(
         eval_id=eval_group_id,
         data_source=cast(Any, data_source),  # Cast for type checker: dynamic schema dict accepted by SDK at runtime

sdk/evaluation/azure-ai-evaluation/samples/aoai_score_model_grader_sample.py

@@ -193,7 +193,7 @@ def demonstrate_score_model_grader():
         # 1. Configure Azure OpenAI model using environment variables
         model_config = AzureOpenAIModelConfiguration(
             azure_endpoint=os.environ.get("endpoint"),
-            api_key=os.environ.get("key"),
+            api_key=os.environ.get("api_key"),
             azure_deployment=os.environ.get("deployment_name"),
             api_version="2024-12-01-preview",
         )