openai[patch]: Partially Revert Update openai chat model to new base class interface (#19871)

Partially Reverts langchain-ai/langchain#19729

---------

Co-authored-by: Erick Friis <erick@langchain.dev>

libs/partners/openai/langchain_openai/chat_models/base.py

@@ -34,7 +34,11 @@ from langchain_core.callbacks import (
     CallbackManagerForLLMRun,
 )
 from langchain_core.language_models import LanguageModelInput
-from langchain_core.language_models.chat_models import BaseChatModel
+from langchain_core.language_models.chat_models import (
+    BaseChatModel,
+    agenerate_from_stream,
+    generate_from_stream,
+)
 from langchain_core.messages import (
     AIMessage,
     AIMessageChunk,
@@ -474,6 +478,8 @@ class ChatOpenAI(BaseChatModel):
             chunk = ChatGenerationChunk(
                 message=chunk, generation_info=generation_info or None
             )
+            if run_manager:
+                run_manager.on_llm_new_token(chunk.text, chunk=chunk, logprobs=logprobs)
             yield chunk
 
     def _generate(
@@ -483,12 +489,13 @@ class ChatOpenAI(BaseChatModel):
         run_manager: Optional[CallbackManagerForLLMRun] = None,
         **kwargs: Any,
     ) -> ChatResult:
+        if self.streaming:
+            stream_iter = self._stream(
+                messages, stop=stop, run_manager=run_manager, **kwargs
+            )
+            return generate_from_stream(stream_iter)
         message_dicts, params = self._create_message_dicts(messages, stop)
-        params = {
-            **params,
-            **({"stream": self.streaming} if self.streaming else {}),
-            **kwargs,
-        }
+        params = {**params, **kwargs}
         response = self.client.create(messages=message_dicts, **params)
         return self._create_chat_result(response)
 
@@ -569,6 +576,10 @@ class ChatOpenAI(BaseChatModel):
             chunk = ChatGenerationChunk(
                 message=chunk, generation_info=generation_info or None
             )
+            if run_manager:
+                await run_manager.on_llm_new_token(
+                    token=chunk.text, chunk=chunk, logprobs=logprobs
+                )
             yield chunk
 
     async def _agenerate(
@@ -578,12 +589,14 @@ class ChatOpenAI(BaseChatModel):
         run_manager: Optional[AsyncCallbackManagerForLLMRun] = None,
         **kwargs: Any,
     ) -> ChatResult:
+        if self.streaming:
+            stream_iter = self._astream(
+                messages, stop=stop, run_manager=run_manager, **kwargs
+            )
+            return await agenerate_from_stream(stream_iter)
+
         message_dicts, params = self._create_message_dicts(messages, stop)
-        params = {
-            **params,
-            **({"stream": self.streaming} if self.streaming else {}),
-            **kwargs,
-        }
+        params = {**params, **kwargs}
         response = await self.async_client.create(messages=message_dicts, **params)
         return self._create_chat_result(response)