Switched to AzureOpenAI for api_type=="azure" (microsoft#1232)

* Switched to AzureOpenAI for api_type=="azure"

* Setting AzureOpenAI to empty object if no `openai`

* extra_ and openai_ kwargs

* test_client, support for Azure and "gpt-35-turbo-instruct"

* instruct/azure model in test_client_stream

* generalize aoai support (microsoft#1)

* generalize aoai support

* Null check, fixing tests

* cleanup test

---------

Co-authored-by: Maxim Saplin <[email protected]>

* Returning back model names for instruct

* process model in create

* None check

---------

Co-authored-by: Chi Wang <[email protected]>

autogen/oai/client.py

@@ -11,7 +11,7 @@
 
 from autogen.oai import completion
 
-from autogen.oai.openai_utils import get_key, OAI_PRICE1K
+from autogen.oai.openai_utils import DEFAULT_AZURE_API_VERSION, get_key, OAI_PRICE1K
 from autogen.token_count_utils import count_token
 from autogen._pydantic import model_dump
 
@@ -21,9 +21,10 @@
 except ImportError:
     ERROR: Optional[ImportError] = ImportError("Please install openai>=1 and diskcache to use autogen.OpenAIWrapper.")
     OpenAI = object
+    AzureOpenAI = object
 else:
     # raises exception if openai>=1 is installed and something is wrong with imports
-    from openai import OpenAI, APIError, __version__ as OPENAIVERSION
+    from openai import OpenAI, AzureOpenAI, APIError, __version__ as OPENAIVERSION
     from openai.resources import Completions
     from openai.types.chat import ChatCompletion
     from openai.types.chat.chat_completion import ChatCompletionMessage, Choice  # type: ignore [attr-defined]
@@ -52,8 +53,18 @@ class OpenAIWrapper:
     """A wrapper class for openai client."""
 
     cache_path_root: str = ".cache"
-    extra_kwargs = {"cache_seed", "filter_func", "allow_format_str_template", "context", "api_version", "tags"}
+    extra_kwargs = {
+        "cache_seed",
+        "filter_func",
+        "allow_format_str_template",
+        "context",
+        "api_version",
+        "api_type",
+        "tags",
+    }
     openai_kwargs = set(inspect.getfullargspec(OpenAI.__init__).kwonlyargs)
+    aopenai_kwargs = set(inspect.getfullargspec(AzureOpenAI.__init__).kwonlyargs)
+    openai_kwargs = openai_kwargs | aopenai_kwargs
     total_usage_summary: Optional[Dict[str, Any]] = None
     actual_usage_summary: Optional[Dict[str, Any]] = None
 
@@ -105,46 +116,10 @@ def __init__(self, *, config_list: Optional[List[Dict[str, Any]]] = None, **base
             self._clients = [self._client(extra_kwargs, openai_config)]
             self._config_list = [extra_kwargs]
 
-    def _process_for_azure(
-        self, config: Dict[str, Any], extra_kwargs: Dict[str, Any], segment: str = "default"
-    ) -> None:
-        # deal with api_version
-        query_segment = f"{segment}_query"
-        headers_segment = f"{segment}_headers"
-        api_version = extra_kwargs.get("api_version")
-        if api_version is not None and query_segment not in config:
-            config[query_segment] = {"api-version": api_version}
-            if segment == "default":
-                # remove the api_version from extra_kwargs
-                extra_kwargs.pop("api_version")
-        if segment == "extra":
-            return
-        # deal with api_type
-        api_type = extra_kwargs.get("api_type")
-        if api_type is not None and api_type.startswith("azure") and headers_segment not in config:
-            api_key = config.get("api_key", os.environ.get("AZURE_OPENAI_API_KEY"))
-            config[headers_segment] = {"api-key": api_key}
-            # remove the api_type from extra_kwargs
-            extra_kwargs.pop("api_type")
-            # deal with model
-            model = extra_kwargs.get("model")
-            if model is None:
-                return
-            if "gpt-3.5" in model:
-                # hack for azure gpt-3.5
-                extra_kwargs["model"] = model = model.replace("gpt-3.5", "gpt-35")
-            base_url = config.get("base_url")
-            if base_url is None:
-                raise ValueError("to use azure openai api, base_url must be specified.")
-            suffix = f"/openai/deployments/{model}"
-            if not base_url.endswith(suffix):
-                config["base_url"] += suffix[1:] if base_url.endswith("/") else suffix
-
     def _separate_openai_config(self, config: Dict[str, Any]) -> Tuple[Dict[str, Any], Dict[str, Any]]:
         """Separate the config into openai_config and extra_kwargs."""
         openai_config = {k: v for k, v in config.items() if k in self.openai_kwargs}
         extra_kwargs = {k: v for k, v in config.items() if k not in self.openai_kwargs}
-        self._process_for_azure(openai_config, extra_kwargs)
         return openai_config, extra_kwargs
 
     def _separate_create_config(self, config: Dict[str, Any]) -> Tuple[Dict[str, Any], Dict[str, Any]]:
@@ -156,10 +131,22 @@ def _separate_create_config(self, config: Dict[str, Any]) -> Tuple[Dict[str, Any
     def _client(self, config: Dict[str, Any], openai_config: Dict[str, Any]) -> OpenAI:
         """Create a client with the given config to override openai_config,
         after removing extra kwargs.
+
+        For Azure models/deployment names there's a convenience modification of model removing dots in
+        the it's value (Azure deploment names can't have dots). I.e. if you have Azure deployment name
+        "gpt-35-turbo" and define model "gpt-3.5-turbo" in the config the function will remove the dot
+        from the name and create a client that connects to "gpt-35-turbo" Azure deployment.
         """
         openai_config = {**openai_config, **{k: v for k, v in config.items() if k in self.openai_kwargs}}
-        self._process_for_azure(openai_config, config)
-        client = OpenAI(**openai_config)
+        api_type = config.get("api_type")
+        if api_type is not None and api_type.startswith("azure"):
+            openai_config["azure_deployment"] = openai_config.get("azure_deployment", config.get("model"))
+            if openai_config["azure_deployment"] is not None:
+                openai_config["azure_deployment"] = openai_config["azure_deployment"].replace(".", "")
+            openai_config["azure_endpoint"] = openai_config.get("azure_endpoint", openai_config.pop("base_url", None))
+            client = AzureOpenAI(**openai_config)
+        else:
+            client = OpenAI(**openai_config)
         return client
 
     @classmethod
@@ -242,8 +229,9 @@ def yes_or_no_filter(context, response):
             full_config = {**config, **self._config_list[i]}
             # separate the config into create_config and extra_kwargs
             create_config, extra_kwargs = self._separate_create_config(full_config)
-            # process for azure
-            self._process_for_azure(create_config, extra_kwargs, "extra")
+            api_type = extra_kwargs.get("api_type")
+            if api_type and api_type.startswith("azure") and "model" in create_config:
+                create_config["model"] = create_config["model"].replace(".", "")
             # construct the create params
             params = self._construct_create_params(create_config, extra_kwargs)
             # get the cache_seed, filter_func and context

test/oai/test_client.py

@@ -31,10 +31,15 @@ def test_aoai_chat_completion():
         filter_dict={"api_type": ["azure"], "model": ["gpt-3.5-turbo", "gpt-35-turbo"]},
     )
     client = OpenAIWrapper(config_list=config_list)
-    # for config in config_list:
-    #     print(config)
-    #     client = OpenAIWrapper(**config)
-    #     response = client.create(messages=[{"role": "user", "content": "2+2="}], cache_seed=None)
+    response = client.create(messages=[{"role": "user", "content": "2+2="}], cache_seed=None)
+    print(response)
+    print(client.extract_text_or_completion_object(response))
+
+    # test dialect
+    config = config_list[0]
+    config["azure_deployment"] = config["model"]
+    config["azure_endpoint"] = config.pop("base_url")
+    client = OpenAIWrapper(**config)
     response = client.create(messages=[{"role": "user", "content": "2+2="}], cache_seed=None)
     print(response)
     print(client.extract_text_or_completion_object(response))
@@ -93,21 +98,23 @@ def test_chat_completion():
 def test_completion():
     config_list = config_list_openai_aoai(KEY_LOC)
     client = OpenAIWrapper(config_list=config_list)
-    response = client.create(prompt="1+1=", model="gpt-3.5-turbo-instruct")
+    model = "gpt-3.5-turbo-instruct"
+    response = client.create(prompt="1+1=", model=model)
     print(response)
     print(client.extract_text_or_completion_object(response))
 
 
 @pytest.mark.skipif(skip, reason="openai>=1 not installed")
 @pytest.mark.parametrize(
-    "cache_seed, model",
+    "cache_seed",
     [
-        (None, "gpt-3.5-turbo-instruct"),
-        (42, "gpt-3.5-turbo-instruct"),
+        None,
+        42,
     ],
 )
-def test_cost(cache_seed, model):
+def test_cost(cache_seed):
     config_list = config_list_openai_aoai(KEY_LOC)
+    model = "gpt-3.5-turbo-instruct"
     client = OpenAIWrapper(config_list=config_list, cache_seed=cache_seed)
     response = client.create(prompt="1+3=", model=model)
     print(response.cost)
@@ -117,7 +124,8 @@ def test_cost(cache_seed, model):
 def test_usage_summary():
     config_list = config_list_openai_aoai(KEY_LOC)
     client = OpenAIWrapper(config_list=config_list)
-    response = client.create(prompt="1+3=", model="gpt-3.5-turbo-instruct", cache_seed=None)
+    model = "gpt-3.5-turbo-instruct"
+    response = client.create(prompt="1+3=", model=model, cache_seed=None)
 
     # usage should be recorded
     assert client.actual_usage_summary["total_cost"] > 0, "total_cost should be greater than 0"
@@ -138,15 +146,15 @@ def test_usage_summary():
     assert client.total_usage_summary is None, "total_usage_summary should be None"
 
     # actual usage and all usage should be different
-    response = client.create(prompt="1+3=", model="gpt-3.5-turbo-instruct", cache_seed=42)
+    response = client.create(prompt="1+3=", model=model, cache_seed=42)
     assert client.total_usage_summary["total_cost"] > 0, "total_cost should be greater than 0"
     assert client.actual_usage_summary is None, "No actual cost should be recorded"
 
 
 if __name__ == "__main__":
-    test_aoai_chat_completion()
-    test_oai_tool_calling_extraction()
-    test_chat_completion()
+    # test_aoai_chat_completion()
+    # test_oai_tool_calling_extraction()
+    # test_chat_completion()
     test_completion()
-    # test_cost()
-    test_usage_summary()
+    # # test_cost()
+    # test_usage_summary()

test/oai/test_client_stream.py

@@ -286,7 +286,9 @@ def test_chat_tools_stream() -> None:
 def test_completion_stream() -> None:
     config_list = config_list_openai_aoai(KEY_LOC)
     client = OpenAIWrapper(config_list=config_list)
-    response = client.create(prompt="1+1=", model="gpt-3.5-turbo-instruct", stream=True)
+    # Azure can't have dot in model/deployment name
+    model = "gpt-35-turbo-instruct" if config_list[0].get("api_type") == "azure" else "gpt-3.5-turbo-instruct"
+    response = client.create(prompt="1+1=", model=model, stream=True)
     print(response)
     print(client.extract_text_or_completion_object(response))