From 1e3e05b0c326162e59c79ccb6e5123b254e36621 Mon Sep 17 00:00:00 2001
From: Roman Solomatin <36135455+Samoed@users.noreply.github.com>
Date: Wed, 26 Jun 2024 18:11:59 +0500
Subject: [PATCH] openai[patch]: add support for extra_body (#23404)

**Description:** Add support passing extra_body parameter

Some OpenAI compatible API's have additional parameters (for example
[vLLM](https://docs.vllm.ai/en/latest/serving/openai_compatible_server.html#extra-parameters))
that can be passed thought `extra_body`. Same question in
https://github.com/openai/openai-python/issues/767

<!--
If no one reviews your PR within a few days, please @-mention one of
baskaryan, efriis, eyurtsev, ccurme, vbarda, hwchase17.
-->
---
 libs/partners/openai/langchain_openai/chat_models/base.py | 6 ++++++
 libs/partners/openai/langchain_openai/llms/base.py        | 6 ++++++
 2 files changed, 12 insertions(+)

diff --git a/libs/partners/openai/langchain_openai/chat_models/base.py b/libs/partners/openai/langchain_openai/chat_models/base.py
index 872955b771c..1c50e80b902 100644
--- a/libs/partners/openai/langchain_openai/chat_models/base.py
+++ b/libs/partners/openai/langchain_openai/chat_models/base.py
@@ -346,6 +346,9 @@ class BaseChatOpenAI(BaseChatModel):
         http_client as well if you'd like a custom client for sync invocations."""
     stop: Optional[Union[List[str], str]] = Field(default=None, alias="stop_sequences")
     """Default stop sequences."""
+    extra_body: Optional[Mapping[str, Any]] = None
+    """Optional additional JSON properties to include in the request parameters when
+    making requests to OpenAI compatible APIs, such as vLLM."""
 
     class Config:
         """Configuration for this pydantic object."""
@@ -445,6 +448,9 @@ class BaseChatOpenAI(BaseChatModel):
             params["max_tokens"] = self.max_tokens
         if self.stop:
             params["stop"] = self.stop
+        if self.extra_body is not None:
+            params["extra_body"] = self.extra_body
+
         return params
 
     def _combine_llm_outputs(self, llm_outputs: List[Optional[dict]]) -> dict:
diff --git a/libs/partners/openai/langchain_openai/llms/base.py b/libs/partners/openai/langchain_openai/llms/base.py
index 006e6e9422e..ca3ff4eafae 100644
--- a/libs/partners/openai/langchain_openai/llms/base.py
+++ b/libs/partners/openai/langchain_openai/llms/base.py
@@ -137,6 +137,9 @@ class BaseOpenAI(BaseLLM):
     http_async_client: Union[Any, None] = None
     """Optional httpx.AsyncClient. Only used for async invocations. Must specify 
         http_client as well if you'd like a custom client for sync invocations."""
+    extra_body: Optional[Mapping[str, Any]] = None
+    """Optional additional JSON properties to include in the request parameters when
+    making requests to OpenAI compatible APIs, such as vLLM."""
 
     class Config:
         """Configuration for this pydantic object."""
@@ -222,6 +225,9 @@ class BaseOpenAI(BaseLLM):
         if self.max_tokens is not None:
             normal_params["max_tokens"] = self.max_tokens
 
+        if self.extra_body is not None:
+            normal_params["extra_body"] = self.extra_body
+
         # Azure gpt-35-turbo doesn't support best_of
         # don't specify best_of if it is 1
         if self.best_of > 1: