fix: Smooth Error Handling for Subscription Token Exhaustion (#1105)

Co-authored-by: Pavan Kumar <v-kupavan.microsoft.com>

code/create_app.py

@@ -9,7 +9,7 @@
 from os import path
 import sys
 import requests
-from openai import AzureOpenAI, Stream
+from openai import AzureOpenAI, Stream, RateLimitError
 from openai.types.chat import ChatCompletionChunk
 from flask import Flask, Response, request, Request, jsonify
 from dotenv import load_dotenv
@@ -20,6 +20,8 @@
 from azure.mgmt.cognitiveservices import CognitiveServicesManagementClient
 from azure.identity import DefaultAzureCredential
 
+ERROR_429_MESSAGE = "We're currently experiencing a high number of requests for the service you're trying to access. Please wait a moment and try again."
+ERROR_GENERIC_MESSAGE = "An error occurred. Please try again. If the problem persists, please contact the site administrator."
 logger = logging.getLogger(__name__)
 
 
@@ -343,17 +345,14 @@ def conversation_azure_byod():
                 return conversation_with_data(request, env_helper)
             else:
                 return conversation_without_data(request, env_helper)
+        except RateLimitError as e:
+            error_message = str(e)
+            logger.exception("Exception in /api/conversation | %s", error_message)
+            return jsonify({"error": ERROR_429_MESSAGE}), 429
         except Exception as e:
             error_message = str(e)
             logger.exception("Exception in /api/conversation | %s", error_message)
-            return (
-                jsonify(
-                    {
-                        "error": "Exception in /api/conversation. See log for more details."
-                    }
-                ),
-                500,
-            )
+            return jsonify({"error": ERROR_GENERIC_MESSAGE}), 500
 
     async def conversation_custom():
         message_orchestrator = get_message_orchestrator()
@@ -385,17 +384,14 @@ async def conversation_custom():
 
             return jsonify(response_obj), 200
 
+        except RateLimitError as e:
+            error_message = str(e)
+            logger.exception("Exception in /api/conversation | %s", error_message)
+            return jsonify({"error": ERROR_429_MESSAGE}), 429
         except Exception as e:
             error_message = str(e)
             logger.exception("Exception in /api/conversation | %s", error_message)
-            return (
-                jsonify(
-                    {
-                        "error": "Exception in /api/conversation. See log for more details."
-                    }
-                ),
-                500,
-            )
+            return jsonify({"error": ERROR_GENERIC_MESSAGE}), 500
 
     @app.route("/api/conversation", methods=["POST"])
     async def conversation():

code/frontend/src/api/api.ts

@@ -14,5 +14,10 @@ export async function callConversationApi(options: ConversationRequest, abortSig
         signal: abortSignal
     });
 
+    if (!response.ok) {
+        const errorData = await response.json();
+        throw new Error(JSON.stringify(errorData.error));
+    }
+
     return response;
 }

code/frontend/src/pages/chat/Chat.tsx

@@ -114,10 +114,12 @@ const Chat = () => {
       }
     } catch (e) {
       if (!abortController.signal.aborted) {
-        console.error(result);
-        alert(
-          "An error occurred. Please try again. If the problem persists, please contact the site administrator."
-        );
+        if (e instanceof Error) {
+          alert(e.message);
+        }
+        else {
+          alert('An error occurred. Please try again. If the problem persists, please contact the site administrator.');
+        }
       }
       setAnswers([...answers, userMessage]);
     } finally {

code/tests/functional/tests/backend_api/default/test_conversation.py

@@ -666,5 +666,5 @@ def test_post_returns_error_when_downstream_fails(
     assert response.status_code == 500
     assert response.headers["Content-Type"] == "application/json"
     assert json.loads(response.text) == {
-        "error": "Exception in /api/conversation. See log for more details."
+        "error": "An error occurred. Please try again. If the problem persists, please contact the site administrator."
     }

code/tests/functional/tests/backend_api/integrated_vectorization_custom_conversation/test_iv_question_answer_tool.py

@@ -277,5 +277,5 @@ def test_post_returns_error_when_downstream_fails(
     assert response.status_code == 500
     assert response.headers["Content-Type"] == "application/json"
     assert json.loads(response.text) == {
-        "error": "Exception in /api/conversation. See log for more details."
+        "error": "An error occurred. Please try again. If the problem persists, please contact the site administrator."
     }

code/tests/functional/tests/backend_api/sk_orchestrator/test_response_without_tool_call.py

@@ -274,5 +274,5 @@ def test_post_returns_error_when_downstream_fails(
     assert response.status_code == 500
     assert response.headers["Content-Type"] == "application/json"
     assert json.loads(response.text) == {
-        "error": "Exception in /api/conversation. See log for more details."
+        "error": "An error occurred. Please try again. If the problem persists, please contact the site administrator."
     }

code/tests/test_app.py

@@ -2,7 +2,8 @@
 This module tests the entry point for the application.
 """
 
-from unittest.mock import AsyncMock, MagicMock, patch, ANY
+from unittest.mock import AsyncMock, MagicMock, Mock, patch, ANY
+from openai import RateLimitError
 import pytest
 from flask.testing import FlaskClient
 from backend.batch.utilities.helpers.config.conversation_flow import ConversationFlow
@@ -320,7 +321,34 @@ def test_conversaation_custom_returns_error_response_on_exception(
         # then
         assert response.status_code == 500
         assert response.json == {
-            "error": "Exception in /api/conversation. See log for more details."
+            "error": "An error occurred. Please try again. If the problem persists, please contact the site administrator."
+        }
+
+    @patch("create_app.get_orchestrator_config")
+    def test_conversation_custom_returns_error_response_on_rate_limit_error(
+        self, get_orchestrator_config_mock, env_helper_mock, client
+    ):
+        """Test that a 429 response is returned on RateLimitError."""
+        # given
+        response_mock = Mock()
+        response_mock.status_code = 429
+        body_mock = {"error": "Rate limit exceeded"}
+
+        rate_limit_error = RateLimitError("Rate limit exceeded", response=response_mock, body=body_mock)
+        get_orchestrator_config_mock.side_effect = rate_limit_error
+
+        # when
+        response = client.post(
+            "/api/conversation",
+            headers={"content-type": "application/json"},
+            json=self.body,
+        )
+
+        # then
+        assert response.status_code == 429
+        assert response.json == {
+            "error": "We're currently experiencing a high number of requests for the service you're trying to access. "
+                     "Please wait a moment and try again."
         }
 
     @patch("create_app.get_message_orchestrator")
@@ -688,7 +716,35 @@ def test_conversation_azure_byod_returns_500_when_exception_occurs(
         # then
         assert response.status_code == 500
         assert response.json == {
-            "error": "Exception in /api/conversation. See log for more details."
+            "error": "An error occurred. Please try again. If the problem persists, please contact the site administrator."
+        }
+
+    @patch("create_app.conversation_with_data")
+    def test_conversation_azure_byod_returns_429_on_rate_limit_error(
+        self, conversation_with_data_mock, env_helper_mock, client
+    ):
+        """Test that a 429 response is returned on RateLimitError for BYOD conversation."""
+        # given
+        response_mock = Mock()
+        response_mock.status_code = 429
+        body_mock = {"error": "Rate limit exceeded"}
+
+        rate_limit_error = RateLimitError("Rate limit exceeded", response=response_mock, body=body_mock)
+        conversation_with_data_mock.side_effect = rate_limit_error
+        env_helper_mock.CONVERSATION_FLOW = ConversationFlow.BYOD.value
+
+        # when
+        response = client.post(
+            "/api/conversation",
+            headers={"content-type": "application/json"},
+            json=self.body,
+        )
+
+        # then
+        assert response.status_code == 429
+        assert response.json == {
+            "error": "We're currently experiencing a high number of requests for the service you're trying to access. "
+                     "Please wait a moment and try again."
         }
 
     @patch("create_app.AzureOpenAI")