Support o1 (#3510)

* support o1

* nit

backend/onyx/llm/chat_llm.py

@@ -453,7 +453,9 @@ class DefaultMultiLLM(LLM):
         if LOG_DANSWER_MODEL_INTERACTIONS:
             self.log_model_configs()
 
-        if DISABLE_LITELLM_STREAMING:
+        if (
+            DISABLE_LITELLM_STREAMING or self.config.model_name == "o1-2024-12-17"
+        ):  # TODO: remove once litellm supports streaming
             yield self.invoke(prompt, tools, tool_choice, structured_response_format)
             return
 

backend/onyx/llm/llm_provider_options.py

@@ -29,6 +29,7 @@ OPENAI_PROVIDER_NAME = "openai"
 OPEN_AI_MODEL_NAMES = [
     "o1-mini",
     "o1-preview",
+    "o1-2024-12-17",
     "gpt-4",
     "gpt-4o",
     "gpt-4o-mini",

backend/requirements/default.txt

@@ -29,7 +29,7 @@ trafilatura==1.12.2
 langchain==0.1.17
 langchain-core==0.1.50
 langchain-text-splitters==0.0.1
-litellm==1.54.1
+litellm==1.55.4
 lxml==5.3.0
 lxml_html_clean==0.2.2
 llama-index==0.9.45

backend/requirements/model_server.txt

@@ -12,5 +12,5 @@ torch==2.2.0
 transformers==4.39.2
 uvicorn==0.21.1
 voyageai==0.2.3
-litellm==1.54.1
+litellm==1.55.4
 sentry-sdk[fastapi,celery,starlette]==2.14.0