fix: Fix for the issue encountered when updating the .env file and mo…

…difying model configurations. (#1347)

Co-authored-by: Pavan Kumar <v-kupavan.microsoft.com>
Co-authored-by: Francia Riesco <Fr4nc3@users.noreply.github.com>

.env.sample

@@ -22,9 +22,8 @@ AZURE_SEARCH_DATASOURCE_NAME=
 # Azure OpenAI for generating the answer and computing the embedding of the documents
 AZURE_OPENAI_RESOURCE=
 AZURE_OPENAI_API_KEY=
-AZURE_OPENAI_MODEL=gpt-35-turbo
-AZURE_OPENAI_MODEL_NAME=gpt-35-turbo
-AZURE_OPENAI_EMBEDDING_MODEL=text-embedding-ada-002
+AZURE_OPENAI_MODEL_INFO="{\"model\":\"gpt-35-turbo-16k\",\"modelName\":\"gpt-35-turbo-16k\",\"modelVersion\":\"0613\"}"
+AZURE_OPENAI_EMBEDDING_MODEL_INFO="{\"model\":\"text-embedding-ada-002\",\"modelName\":\"text-embedding-ada-002\",\"modelVersion\":\"2\"}"
 AZURE_OPENAI_TEMPERATURE=0
 AZURE_OPENAI_TOP_P=1.0
 AZURE_OPENAI_MAX_TOKENS=1000

code/backend/batch/utilities/helpers/env_helper.py

@@ -1,3 +1,4 @@
+import json
 import os
 import logging
 import threading
@@ -92,11 +93,19 @@ def __load_config(self, **kwargs) -> None:
 
         self.AZURE_AUTH_TYPE = os.getenv("AZURE_AUTH_TYPE", "keys")
         # Azure OpenAI
+        # Default model info
+        default_azure_openai_model_info = '{"model":"gpt-35-turbo-16k","modelName":"gpt-35-turbo-16k","modelVersion":"0613"}'
+        default_azure_openai_embedding_model_info = '{"model":"text-embedding-ada-002","modelName":"text-embedding-ada-002","modelVersion":"2"}'
+
         self.AZURE_OPENAI_RESOURCE = os.getenv("AZURE_OPENAI_RESOURCE", "")
-        self.AZURE_OPENAI_MODEL = os.getenv("AZURE_OPENAI_MODEL", "")
-        self.AZURE_OPENAI_MODEL_NAME = os.getenv(
-            "AZURE_OPENAI_MODEL_NAME", "gpt-35-turbo"
+
+        # Fetch and assign model info
+        azure_openai_model_info = self.get_info_from_env(
+            "AZURE_OPENAI_MODEL_INFO", default_azure_openai_model_info
         )
+        self.AZURE_OPENAI_MODEL = azure_openai_model_info.get("model")
+        self.AZURE_OPENAI_MODEL_NAME = azure_openai_model_info.get("modelName")
+
         self.AZURE_OPENAI_VISION_MODEL = os.getenv("AZURE_OPENAI_VISION_MODEL", "gpt-4")
         self.AZURE_OPENAI_TEMPERATURE = os.getenv("AZURE_OPENAI_TEMPERATURE", "0")
         self.AZURE_OPENAI_TOP_P = os.getenv("AZURE_OPENAI_TOP_P", "1.0")
@@ -110,9 +119,16 @@ def __load_config(self, **kwargs) -> None:
             "AZURE_OPENAI_API_VERSION", "2024-02-01"
         )
         self.AZURE_OPENAI_STREAM = os.getenv("AZURE_OPENAI_STREAM", "true")
-        self.AZURE_OPENAI_EMBEDDING_MODEL = os.getenv(
-            "AZURE_OPENAI_EMBEDDING_MODEL", ""
+
+        # Fetch and assign embedding model info
+        azure_openai_embedding_model_info = self.get_info_from_env(
+            "AZURE_OPENAI_EMBEDDING_MODEL_INFO",
+            default_azure_openai_embedding_model_info,
         )
+        self.AZURE_OPENAI_EMBEDDING_MODEL = azure_openai_embedding_model_info.get(
+            "model"
+        )
+
         self.SHOULD_STREAM = (
             True if self.AZURE_OPENAI_STREAM.lower() == "true" else False
         )
@@ -267,6 +283,14 @@ def get_env_var_float(self, var_name: str, default: float):
     def is_auth_type_keys(self):
         return self.AZURE_AUTH_TYPE == "keys"
 
+    def get_info_from_env(self, env_var: str, default_info: str) -> dict:
+        # Fetch and parse model info from the environment variable.
+        info_str = os.getenv(env_var, default_info)
+        # Handle escaped characters in the JSON string by wrapping it in double quotes for parsing.
+        if "\\" in info_str:
+            info_str = json.loads(f'"{info_str}"')
+        return {} if not info_str else json.loads(info_str)
+
     @staticmethod
     def check_env():
         for attr, value in EnvHelper().__dict__.items():

code/tests/functional/app_config.py

@@ -1,4 +1,5 @@
 import base64
+import json
 import logging
 import os
 from backend.batch.utilities.helpers.config.conversation_flow import ConversationFlow
@@ -24,11 +25,10 @@ class AppConfig:
         "AZURE_KEY_VAULT_ENDPOINT": "some-key-vault-endpoint",
         "AZURE_OPENAI_API_KEY": "some-azure-openai-api-key",
         "AZURE_OPENAI_API_VERSION": "2024-02-01",
-        "AZURE_OPENAI_EMBEDDING_MODEL": "some-embedding-model",
+        "AZURE_OPENAI_EMBEDDING_MODEL_INFO": '{"model":"some-embedding-model","modelName":"some-embedding-model-name","modelVersion":"some-embedding-model-version"}',
         "AZURE_OPENAI_ENDPOINT": "some-openai-endpoint",
         "AZURE_OPENAI_MAX_TOKENS": "1000",
-        "AZURE_OPENAI_MODEL": "some-openai-model",
-        "AZURE_OPENAI_MODEL_NAME": "some-openai-model-name",
+        "AZURE_OPENAI_MODEL_INFO": '{"model":"some-openai-model","modelName":"some-openai-model-name","modelVersion":"some-openai-model-version"}',
         "AZURE_OPENAI_VISION_MODEL": "some-openai-vision-model",
         "AZURE_OPENAI_RESOURCE": "some-openai-resource",
         "AZURE_OPENAI_STREAM": "True",
@@ -95,6 +95,10 @@ def set(self, key: str, value: str | None) -> None:
     def get(self, key: str) -> str | None:
         return self.config[key]
 
+    def get_from_json(self, config_key: str, field: str) -> str | None:
+        config_json = json.loads(self.config[config_key])
+        return config_json.get(field)
+
     def get_all(self) -> dict[str, str | None]:
         return self.config
 

code/tests/functional/conftest.py

@@ -20,7 +20,7 @@ def setup_default_mocking(httpserver: HTTPServer, app_config: AppConfig):
     ).respond_with_data()
 
     httpserver.expect_request(
-        f"/openai/deployments/{app_config.get('AZURE_OPENAI_EMBEDDING_MODEL')}/embeddings",
+        f"/openai/deployments/{app_config.get_from_json('AZURE_OPENAI_EMBEDDING_MODEL_INFO','model')}/embeddings",
         method="POST",
     ).respond_with_json(
         {
@@ -58,15 +58,15 @@ def setup_default_mocking(httpserver: HTTPServer, app_config: AppConfig):
 
     httpserver.expect_request(
         re.compile(
-            f"/openai/deployments/({app_config.get('AZURE_OPENAI_MODEL')}|{app_config.get('AZURE_OPENAI_VISION_MODEL')})/chat/completions"
+            f"/openai/deployments/({app_config.get_from_json('AZURE_OPENAI_MODEL_INFO','model')}|{app_config.get('AZURE_OPENAI_VISION_MODEL')})/chat/completions"
         ),
         method="POST",
     ).respond_with_json(
         {
             "id": "chatcmpl-6v7mkQj980V1yBec6ETrKPRqFjNw9",
             "object": "chat.completion",
             "created": 1679072642,
-            "model": app_config.get("AZURE_OPENAI_MODEL"),
+            "model": app_config.get_from_json("AZURE_OPENAI_MODEL_INFO", "model"),
             "usage": {
                 "prompt_tokens": 58,
                 "completion_tokens": 68,
@@ -194,7 +194,7 @@ def setup_default_mocking(httpserver: HTTPServer, app_config: AppConfig):
                     "inputs": [{"name": "text", "source": "/document/pages/*"}],
                     "outputs": [{"name": "embedding", "targetName": "content_vector"}],
                     "resourceUri": f"https://localhost:{httpserver.port}/",
-                    "deploymentId": f"{app_config.get('AZURE_OPENAI_EMBEDDING_MODEL')}",
+                    "deploymentId": f"{app_config.get_from_json('AZURE_OPENAI_EMBEDDING_MODEL_INFO','model')}",
                     "apiKey": f"{app_config.get('AZURE_OPENAI_API_KEY')}",
                 },
             ],

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

@@ -27,7 +27,7 @@
 @pytest.fixture(autouse=True)
 def completions_mocking(httpserver: HTTPServer, app_config: AppConfig):
     httpserver.expect_oneshot_request(
-        f"/openai/deployments/{app_config.get('AZURE_OPENAI_MODEL')}/chat/completions",
+        f"/openai/deployments/{app_config.get_from_json('AZURE_OPENAI_MODEL_INFO','model')}/chat/completions",
         method="POST",
     ).respond_with_json(
         {
@@ -48,7 +48,7 @@ def completions_mocking(httpserver: HTTPServer, app_config: AppConfig):
             ],
             "created": 1714576877,
             "id": "chatcmpl-9K63hMvVH1DyQJqqM7rFE4oRPFCeR",
-            "model": app_config.get("AZURE_OPENAI_MODEL"),
+            "model": app_config.get_from_json("AZURE_OPENAI_MODEL_INFO", "model"),
             "object": "chat.completion",
             "prompt_filter_results": [
                 {
@@ -72,7 +72,7 @@ def completions_mocking(httpserver: HTTPServer, app_config: AppConfig):
 
     httpserver.expect_oneshot_request(
         re.compile(
-            f"/openai/deployments/({app_config.get('AZURE_OPENAI_MODEL')}|{app_config.get('AZURE_OPENAI_VISION_MODEL')})/chat/completions"
+            f"/openai/deployments/({app_config.get_from_json('AZURE_OPENAI_MODEL_INFO','model')}|{app_config.get('AZURE_OPENAI_VISION_MODEL')})/chat/completions"
         ),
         method="POST",
     ).respond_with_json(
@@ -95,7 +95,7 @@ def completions_mocking(httpserver: HTTPServer, app_config: AppConfig):
             ],
             "created": 1714576891,
             "id": "chatcmpl-9K63vDGs3slJFynnpi2K6RcVPwgrT",
-            "model": app_config.get("AZURE_OPENAI_MODEL"),
+            "model": app_config.get_from_json("AZURE_OPENAI_MODEL_INFO", "model"),
             "object": "chat.completion",
             "prompt_filter_results": [
                 {
@@ -167,7 +167,7 @@ def test_post_responds_successfully(app_url: str, app_config: AppConfig):
         ],
         "created": "response.created",
         "id": "response.id",
-        "model": app_config.get("AZURE_OPENAI_MODEL"),
+        "model": app_config.get_from_json("AZURE_OPENAI_MODEL_INFO", "model"),
         "object": "response.object",
     }
     assert response.headers["Content-Type"] == "application/json"

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

@@ -28,14 +28,14 @@
 @pytest.fixture(autouse=True)
 def completions_mocking(httpserver: HTTPServer, app_config: AppConfig):
     httpserver.expect_oneshot_request(
-        f"/openai/deployments/{app_config.get('AZURE_OPENAI_MODEL')}/chat/completions",
+        f"/openai/deployments/{app_config.get_from_json('AZURE_OPENAI_MODEL_INFO','model')}/chat/completions",
         method="POST",
     ).respond_with_json(
         {
             "id": "chatcmpl-6v7mkQj980V1yBec6ETrKPRqFjNw9",
             "object": "chat.completion",
             "created": 1679072642,
-            "model": app_config.get("AZURE_OPENAI_MODEL"),
+            "model": app_config.get_from_json("AZURE_OPENAI_MODEL_INFO", "model"),
             "usage": {
                 "prompt_tokens": 58,
                 "completion_tokens": 68,
@@ -58,7 +58,7 @@ def completions_mocking(httpserver: HTTPServer, app_config: AppConfig):
     )
 
     httpserver.expect_oneshot_request(
-        f"/openai/deployments/{app_config.get('AZURE_OPENAI_MODEL')}/chat/completions",
+        f"/openai/deployments/{app_config.get_from_json('AZURE_OPENAI_MODEL_INFO','model')}/chat/completions",
         method="POST",
     ).respond_with_json(
         {
@@ -110,7 +110,7 @@ def test_post_responds_successfully(app_url: str, app_config: AppConfig):
         ],
         "created": "response.created",
         "id": "response.id",
-        "model": app_config.get("AZURE_OPENAI_MODEL"),
+        "model": app_config.get_from_json("AZURE_OPENAI_MODEL_INFO", "model"),
         "object": "response.object",
     }
     assert response.headers["Content-Type"] == "application/json"
@@ -126,7 +126,7 @@ def test_post_makes_correct_calls_to_openai_embeddings_to_get_vector_dimensions(
     verify_request_made(
         mock_httpserver=httpserver,
         request_matcher=RequestMatcher(
-            path=f"/openai/deployments/{app_config.get('AZURE_OPENAI_EMBEDDING_MODEL')}/embeddings",
+            path=f"/openai/deployments/{app_config.get_from_json('AZURE_OPENAI_EMBEDDING_MODEL_INFO','model')}/embeddings",
             method="POST",
             json={
                 "input": [[1199]],
@@ -155,13 +155,15 @@ def test_post_makes_correct_calls_to_openai_embeddings_to_embed_question_to_sear
     verify_request_made(
         mock_httpserver=httpserver,
         request_matcher=RequestMatcher(
-            path=f"/openai/deployments/{app_config.get('AZURE_OPENAI_EMBEDDING_MODEL')}/embeddings",
+            path=f"/openai/deployments/{app_config.get_from_json('AZURE_OPENAI_EMBEDDING_MODEL_INFO','model')}/embeddings",
             method="POST",
             json={
                 "input": [
                     [3923, 374, 279, 7438, 315, 2324, 30]
                 ],  # Embedding of "What is the meaning of life?"
-                "model": app_config.get("AZURE_OPENAI_EMBEDDING_MODEL"),
+                "model": app_config.get_from_json(
+                    "AZURE_OPENAI_EMBEDDING_MODEL_INFO", "model"
+                ),
                 "encoding_format": "base64",
             },
             headers={
@@ -188,7 +190,7 @@ def test_post_makes_correct_calls_to_openai_embeddings_to_embed_question_to_stor
     verify_request_made(
         mock_httpserver=httpserver,
         request_matcher=RequestMatcher(
-            path=f"/openai/deployments/{app_config.get('AZURE_OPENAI_EMBEDDING_MODEL')}/embeddings",
+            path=f"/openai/deployments/{app_config.get_from_json('AZURE_OPENAI_EMBEDDING_MODEL_INFO','model')}/embeddings",
             method="POST",
             json={
                 "input": [
@@ -265,7 +267,7 @@ def test_post_makes_correct_call_to_openai_chat_completions_with_functions(
     verify_request_made(
         mock_httpserver=httpserver,
         request_matcher=RequestMatcher(
-            path=f"/openai/deployments/{app_config.get('AZURE_OPENAI_MODEL')}/chat/completions",
+            path=f"/openai/deployments/{app_config.get_from_json('AZURE_OPENAI_MODEL_INFO','model')}/chat/completions",
             method="POST",
             json={
                 "messages": [
@@ -555,7 +557,7 @@ def test_post_makes_correct_call_to_openai_chat_completions_with_documents(
     verify_request_made(
         mock_httpserver=httpserver,
         request_matcher=RequestMatcher(
-            path=f"/openai/deployments/{app_config.get('AZURE_OPENAI_MODEL')}/chat/completions",
+            path=f"/openai/deployments/{app_config.get_from_json('AZURE_OPENAI_MODEL_INFO','model')}/chat/completions",
             method="POST",
             json={
                 "messages": [
@@ -589,7 +591,7 @@ def test_post_makes_correct_call_to_openai_chat_completions_with_documents(
                         "role": "user",
                     },
                 ],
-                "model": app_config.get("AZURE_OPENAI_MODEL"),
+                "model": app_config.get_from_json("AZURE_OPENAI_MODEL_INFO", "model"),
                 "max_tokens": int(app_config.get("AZURE_OPENAI_MAX_TOKENS")),
                 "temperature": 0,
             },

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

@@ -61,14 +61,14 @@ def setup_config_mocking(httpserver: HTTPServer):
 @pytest.fixture(autouse=True)
 def completions_mocking(httpserver: HTTPServer, app_config: AppConfig):
     httpserver.expect_oneshot_request(
-        f"/openai/deployments/{app_config.get('AZURE_OPENAI_MODEL')}/chat/completions",
+        f"/openai/deployments/{app_config.get_from_json('AZURE_OPENAI_MODEL_INFO','model')}/chat/completions",
         method="POST",
     ).respond_with_json(
         {
             "id": "chatcmpl-6v7mkQj980V1yBec6ETrKPRqFjNw9",
             "object": "chat.completion",
             "created": 1679072642,
-            "model": app_config.get("AZURE_OPENAI_MODEL"),
+            "model": app_config.get_from_json("AZURE_OPENAI_MODEL_INFO", "model"),
             "usage": {
                 "prompt_tokens": 58,
                 "completion_tokens": 68,
@@ -92,7 +92,7 @@ def completions_mocking(httpserver: HTTPServer, app_config: AppConfig):
 
     httpserver.expect_oneshot_request(
         re.compile(
-            f"/openai/deployments/({app_config.get('AZURE_OPENAI_MODEL')}|{app_config.get('AZURE_OPENAI_VISION_MODEL')})/chat/completions"
+            f"/openai/deployments/({app_config.get_from_json('AZURE_OPENAI_MODEL_INFO','model')}|{app_config.get('AZURE_OPENAI_VISION_MODEL')})/chat/completions"
         ),
         method="POST",
     ).respond_with_json(
@@ -125,7 +125,7 @@ def test_post_responds_successfully_when_not_filtered(
 ):
     # given
     httpserver.expect_oneshot_request(
-        f"/openai/deployments/{app_config.get('AZURE_OPENAI_MODEL')}/chat/completions",
+        f"/openai/deployments/{app_config.get_from_json('AZURE_OPENAI_MODEL_INFO','model')}/chat/completions",
         method="POST",
     ).respond_with_json(
         {
@@ -175,7 +175,7 @@ def test_post_responds_successfully_when_not_filtered(
         ],
         "created": "response.created",
         "id": "response.id",
-        "model": app_config.get("AZURE_OPENAI_MODEL"),
+        "model": app_config.get_from_json("AZURE_OPENAI_MODEL_INFO", "model"),
         "object": "response.object",
     }
     assert response.headers["Content-Type"] == "application/json"
@@ -186,7 +186,7 @@ def test_post_responds_successfully_when_filtered(
 ):
     # given
     httpserver.expect_oneshot_request(
-        f"/openai/deployments/{app_config.get('AZURE_OPENAI_MODEL')}/chat/completions",
+        f"/openai/deployments/{app_config.get_from_json('AZURE_OPENAI_MODEL_INFO','model')}/chat/completions",
         method="POST",
     ).respond_with_json(
         {
@@ -236,7 +236,7 @@ def test_post_responds_successfully_when_filtered(
         ],
         "created": "response.created",
         "id": "response.id",
-        "model": app_config.get("AZURE_OPENAI_MODEL"),
+        "model": app_config.get_from_json("AZURE_OPENAI_MODEL_INFO", "model"),
         "object": "response.object",
     }
     assert response.headers["Content-Type"] == "application/json"
@@ -247,7 +247,7 @@ def test_post_makes_correct_call_to_openai_from_post_prompt_tool(
 ):
     # given
     httpserver.expect_oneshot_request(
-        f"/openai/deployments/{app_config.get('AZURE_OPENAI_MODEL')}/chat/completions",
+        f"/openai/deployments/{app_config.get_from_json('AZURE_OPENAI_MODEL_INFO','model')}/chat/completions",
         method="POST",
     ).respond_with_json(
         {
@@ -280,7 +280,7 @@ def test_post_makes_correct_call_to_openai_from_post_prompt_tool(
     verify_request_made(
         mock_httpserver=httpserver,
         request_matcher=RequestMatcher(
-            path=f"/openai/deployments/{app_config.get('AZURE_OPENAI_MODEL')}/chat/completions",
+            path=f"/openai/deployments/{app_config.get_from_json('AZURE_OPENAI_MODEL_INFO','model')}/chat/completions",
             method="POST",
             json={
                 "messages": [
@@ -289,7 +289,7 @@ def test_post_makes_correct_call_to_openai_from_post_prompt_tool(
                         "role": "user",
                     }
                 ],
-                "model": app_config.get("AZURE_OPENAI_MODEL"),
+                "model": app_config.get_from_json("AZURE_OPENAI_MODEL_INFO", "model"),
                 "max_tokens": int(app_config.get("AZURE_OPENAI_MAX_TOKENS")),
             },
             headers={