Set GPT 4o as default and add O3 mini (#3899)

* quick update to models

* add reqs

* update version

backend/onyx/llm/chat_llm.py

@@ -409,7 +409,11 @@ class DefaultMultiLLM(LLM):
                 # For now, we don't support parallel tool calls
                 # NOTE: we can't pass this in if tools are not specified
                 # or else OpenAI throws an error
-                **({"parallel_tool_calls": False} if tools else {}),
+                **(
+                    {"parallel_tool_calls": False}
+                    if tools and self.config.model_name != "o3-mini"
+                    else {}
+                ),  # TODO: remove once LITELLM has patched
                 **(
                     {"response_format": structured_response_format}
                     if structured_response_format
@@ -469,9 +473,7 @@ class DefaultMultiLLM(LLM):
         if LOG_DANSWER_MODEL_INTERACTIONS:
             self.log_model_configs()
 
-        if (
-            DISABLE_LITELLM_STREAMING or self.config.model_name == "o1-2024-12-17"
-        ):  # TODO: remove once litellm supports streaming
+        if DISABLE_LITELLM_STREAMING:
             yield self.invoke(prompt, tools, tool_choice, structured_response_format)
             return
 

backend/onyx/llm/llm_provider_options.py

@@ -27,6 +27,7 @@ class WellKnownLLMProviderDescriptor(BaseModel):
 
 OPENAI_PROVIDER_NAME = "openai"
 OPEN_AI_MODEL_NAMES = [
+    "o3-mini",
     "o1-mini",
     "o1-preview",
     "o1-2024-12-17",
@@ -91,7 +92,7 @@ def fetch_available_well_known_llms() -> list[WellKnownLLMProviderDescriptor]:
             api_version_required=False,
             custom_config_keys=[],
             llm_names=fetch_models_for_provider(OPENAI_PROVIDER_NAME),
-            default_model="gpt-4",
+            default_model="gpt-4o",
             default_fast_model="gpt-4o-mini",
         ),
         WellKnownLLMProviderDescriptor(

backend/requirements/default.txt

@@ -37,7 +37,7 @@ langchainhub==0.1.21
 langgraph==0.2.59
 langgraph-checkpoint==2.0.5
 langgraph-sdk==0.1.44
-litellm==1.55.4
+litellm==1.60.2
 lxml==5.3.0
 lxml_html_clean==0.2.2
 llama-index==0.9.45
@@ -46,7 +46,7 @@ msal==1.28.0
 nltk==3.8.1
 Office365-REST-Python-Client==2.5.9
 oauthlib==3.2.2
-openai==1.55.3
+openai==1.61.0
 openpyxl==3.1.2
 playwright==1.41.2
 psutil==5.9.5

backend/requirements/model_server.txt

@@ -3,7 +3,7 @@ cohere==5.6.1
 fastapi==0.109.2
 google-cloud-aiplatform==1.58.0
 numpy==1.26.4
-openai==1.55.3
+openai==1.61.0
 pydantic==2.8.2
 retry==0.9.2
 safetensors==0.4.2
@@ -12,5 +12,5 @@ torch==2.2.0
 transformers==4.39.2
 uvicorn==0.21.1
 voyageai==0.2.3
-litellm==1.55.4
+litellm==1.60.2
 sentry-sdk[fastapi,celery,starlette]==2.14.0