Add support for o1 (#2538)

* add o1 support + bump litellm/openai

* ports

* update exception message for testing

backend/danswer/llm/llm_provider_options.py

@@ -24,6 +24,8 @@ class WellKnownLLMProviderDescriptor(BaseModel):
 
 OPENAI_PROVIDER_NAME = "openai"
 OPEN_AI_MODEL_NAMES = [
+    "o1-mini",
+    "o1-preview",
     "gpt-4",
     "gpt-4o",
     "gpt-4o-mini",

backend/danswer/llm/utils.py

@@ -47,7 +47,9 @@ if TYPE_CHECKING:
 logger = setup_logger()
 
 
-def litellm_exception_to_error_msg(e: Exception, llm: LLM) -> str:
+def litellm_exception_to_error_msg(
+    e: Exception, llm: LLM, fallback_to_error_msg: bool = False
+) -> str:
     error_msg = str(e)
 
     if isinstance(e, BadRequestError):
@@ -94,7 +96,7 @@ def litellm_exception_to_error_msg(e: Exception, llm: LLM) -> str:
         error_msg = "Request timed out: The operation took too long to complete. Please try again."
     elif isinstance(e, APIError):
         error_msg = f"API error: An error occurred while communicating with the API. Details: {str(e)}"
-    else:
+    elif not fallback_to_error_msg:
         error_msg = "An unexpected error occurred while processing your request. Please try again later."
     return error_msg
 

backend/danswer/server/manage/llm/api.py

@@ -18,6 +18,7 @@ from danswer.llm.factory import get_default_llms
 from danswer.llm.factory import get_llm
 from danswer.llm.llm_provider_options import fetch_available_well_known_llms
 from danswer.llm.llm_provider_options import WellKnownLLMProviderDescriptor
+from danswer.llm.utils import litellm_exception_to_error_msg
 from danswer.llm.utils import test_llm
 from danswer.server.manage.llm.models import FullLLMProvider
 from danswer.server.manage.llm.models import LLMProviderDescriptor
@@ -78,7 +79,10 @@ def test_llm_configuration(
     )
 
     if error:
-        raise HTTPException(status_code=400, detail=error)
+        client_error_msg = litellm_exception_to_error_msg(
+            error, llm, fallback_to_error_msg=True
+        )
+        raise HTTPException(status_code=400, detail=client_error_msg)
 
 
 @admin_router.post("/test/default")

backend/requirements/default.txt

@@ -28,14 +28,14 @@ jsonref==1.1.0
 langchain==0.1.17
 langchain-core==0.1.50
 langchain-text-splitters==0.0.1
-litellm==1.43.18
+litellm==1.47.1
 llama-index==0.9.45
 Mako==1.2.4
 msal==1.28.0
 nltk==3.8.1
 Office365-REST-Python-Client==2.5.9
 oauthlib==3.2.2
-openai==1.41.1 
+openai==1.47.0
 openpyxl==3.1.2
 playwright==1.41.2
 psutil==5.9.5

backend/requirements/model_server.txt

@@ -3,7 +3,7 @@ einops==0.8.0
 fastapi==0.109.2
 google-cloud-aiplatform==1.58.0
 numpy==1.26.4
-openai==1.41.1
+openai==1.47.0
 pydantic==2.8.2
 retry==0.9.2
 safetensors==0.4.2

web/src/lib/hooks.ts

@@ -230,6 +230,8 @@ export const useUserGroups = (): {
 
 const MODEL_DISPLAY_NAMES: { [key: string]: string } = {
   // OpenAI models
+  "o1-mini": "O1 Mini",
+  "o1-preview": "O1 Preview",
   "gpt-4": "GPT 4",
   "gpt-4o": "GPT 4o",
   "gpt-4o-2024-08-06": "GPT 4o (Structured Outputs)",
@@ -292,7 +294,7 @@ export function getDisplayNameForModel(modelName: string): string {
 }
 
 export const defaultModelsByProvider: { [name: string]: string[] } = {
-  openai: ["gpt-4", "gpt-4o", "gpt-4o-mini"],
+  openai: ["gpt-4", "gpt-4o", "gpt-4o-mini", "o1-mini", "o1-preview"],
   bedrock: [
     "meta.llama3-1-70b-instruct-v1:0",
     "meta.llama3-1-8b-instruct-v1:0",