remove eval (#37678)

* remove eval

* change classifier

Author: xiangyan99

sdk/ai/azure-ai-generative/CHANGELOG.md

@@ -1,9 +1,16 @@
 # Release History
 
+## 1.0.0b10 (Unreleased)
+
+### Other Changes
+
+- Marked as inactive.
+
 ## 1.0.0b9 (2024-09-16)
 
 ### Bugs Fixed
-security bug - code injection
+
+- Bug fixes
 
 ## 1.0.0b8 (2024-03-27)
 

sdk/ai/azure-ai-generative/azure/ai/generative/_version.py

@@ -2,4 +2,4 @@
 # Copyright (c) Microsoft Corporation. All rights reserved.
 # ---------------------------------------------------------
 
-VERSION = "1.0.0b9"
+VERSION = "1.0.0b10"

sdk/ai/azure-ai-generative/azure/ai/generative/evaluate/pf_templates/built_in_metrics/chat/construct_groundedness_request.py

@@ -1,5 +1,6 @@
 from promptflow import tool
 import json
+import ast
 
 
 def normalize_user_text(user_text):
@@ -29,7 +30,7 @@ def construct_groundedness_requests(parsed_chat: dict) -> str:
         question = parsed_chat["questions"][i]
         answer = parsed_chat["answers"][i]
         try:
-            retrieved_documents = eval(
+            retrieved_documents = ast.literal_eval(
                 parsed_chat["retrieved_documents"][i])
         except Exception:
             retrieved_documents = [

sdk/ai/azure-ai-generative/azure/ai/generative/evaluate/pf_templates/built_in_metrics/chat/parse_groundedness_responses.py

@@ -2,6 +2,7 @@
 from typing import List
 import numpy as np
 import re
+import ast
 
 
 def parse_single_response(response: dict) -> list:
@@ -10,7 +11,7 @@ def parse_single_response(response: dict) -> list:
         harm_type = key.replace("generic", "gpt")
         parsed_harm_response = {}
         try:
-            harm_response = eval(response[key])
+            harm_response = ast.literal_eval(response[key])
         except Exception:
             harm_response = response[key]
         if harm_response != "" and isinstance(harm_response, dict):

sdk/ai/azure-ai-generative/azure/ai/generative/evaluate/pf_templates/built_in_metrics/chat/parse_service_response.py

@@ -2,6 +2,7 @@
 from typing import List
 import numpy as np
 import re
+import ast
 
 
 def parse_single_sample(response: dict,
@@ -13,15 +14,15 @@ def parse_single_sample(response: dict,
         if selected_label_keys[harm_type]:
             parsed_harm_response = {}
             try:
-                harm_response = eval(response[key])
+                harm_response = ast.literal_eval(response[key])
             except NameError as e:
                 # fix the eval error if there's "true" in the response
                 m = re.findall("name '(.+)' is not defined", str(e))
                 if m:
                     for word in m:
                         response[key] = response[key].replace(word,
                                                               word.title())
-                    harm_response = eval(response[key])
+                    harm_response = ast.literal_eval(response[key])
                 else:
                     harm_response = ""
             except Exception:

sdk/ai/azure-ai-generative/azure/ai/generative/evaluate/pf_templates/built_in_metrics/qa/parse_groundedness_response.py

@@ -2,6 +2,7 @@
 from typing import List
 import numpy as np
 import re
+import ast
 
 
 def parse_single_sample(response: dict) -> list:
@@ -10,7 +11,7 @@ def parse_single_sample(response: dict) -> list:
         harm_type = key.replace("generic", "gpt")
         parsed_harm_response = {}
         try:
-            harm_response = eval(response[key])
+            harm_response = ast.literal_eval(response[key])
         except Exception:
             harm_response = response[key]
         if harm_response != "" and isinstance(harm_response, dict):

sdk/ai/azure-ai-generative/azure/ai/generative/evaluate/pf_templates/built_in_metrics/qa/parse_service_response.py

@@ -2,6 +2,7 @@
 from typing import List
 import numpy as np
 import re
+import ast
 
 
 def parse_single_sample(response: dict, selected_metrics: dict) -> list:
@@ -12,15 +13,15 @@ def parse_single_sample(response: dict, selected_metrics: dict) -> list:
         if selected_label_keys[harm_type]:
             parsed_harm_response = {}
             try:
-                harm_response = eval(response[key])
+                harm_response = ast.literal_eval(response[key])
             except NameError as e:
                 # fix the eval error if there's "true" in the response
                 m = re.findall("name '(.+)' is not defined", str(e))
                 if m:
                     for word in m:
                         response[key] = response[key].replace(word,
                                                               word.title())
-                    harm_response = eval(response[key])
+                    harm_response = ast.literal_eval(response[key])
                 else:
                     harm_response = ""
             except Exception: