feat: support new api, json format (#4971)

* feat: support new api, json format

* fix: change openai version

* add: notebook

* fix: embedding call api

* fix: poetry lock

GitOrigin-RevId: 9a594f9dc21775042dd2464eec591d616c5be800

llm_app/model_wrappers/api_clients/clients.py

@@ -23,8 +23,6 @@ def __init__(
     ):
         import openai
 
-        openai.api_requestor.TIMEOUT_SECS = 90
-
         openai.api_key = api_key
         if api_type:
             openai.api_type = api_type
@@ -34,20 +32,21 @@ def __init__(
             openai.api_version = api_version
 
         self.api = openai
+        self.client = openai.OpenAI(api_key=api_key)
 
 
 class OpenAIChatCompletionClient(OpenAIClient):
     @retry(wait=wait_random_exponential(min=1, max=60), stop=stop_after_attempt(6))
     def make_request(self, **kwargs):
         logfun("Calling OpenAI chat completion service %s", str(kwargs)[:100])
-        return self.api.ChatCompletion.create(**kwargs)
+        return self.client.chat.completions.create(**kwargs)
 
 
 class OpenAIEmbeddingClient(OpenAIClient):
     @retry(wait=wait_random_exponential(min=1, max=60), stop=stop_after_attempt(6))
     def make_request(self, **kwargs):
         logfun("Calling OpenAI embedding service %s", str(kwargs)[:100])
-        return self.api.Embedding.create(**kwargs)
+        return self.client.embeddings.create(**kwargs)
 
 
 class HuggingFaceClient(APIClient):

llm_app/model_wrappers/openai_wrapper/api_models.py

@@ -47,6 +47,8 @@ def __call__(self, text: str, locator="gpt-3.5-turbo", **kwargs) -> str:
             kwargs["model"] = locator
 
         messages = MessagePreparer.prepare_chat_messages(text)
+        if "response_format" in kwargs:
+            kwargs["response_format"] = kwargs["response_format"].value
 
         logfun(f"Calling OpenAI API with: {messages}\n")
         response = self.api_client.make_request(messages=messages, **kwargs)
@@ -126,7 +128,7 @@ def __call__(self, text: str, locator="text-embedding-ada-002", **kwargs):
             kwargs["model"] = locator
 
         response = self.api_client.make_request(input=[text], **kwargs)
-        return response["data"][0]["embedding"]
+        return response.data[0].embedding
 
     def apply(
         self,