(feat) proxy: add mode in model info

litellm/health_check.py

@@ -68,11 +68,12 @@ async def _perform_health_check(model_list: list):
     tasks = []
     for model in model_list:
         litellm_params = model["litellm_params"]
+        model_info = model.get("model_info", {})
         litellm_params["model"] = litellm.utils.remove_model_id(litellm_params["model"])
         litellm_params["messages"] = _get_random_llm_message()
 
         prepped_params.append(litellm_params)
-        if model.get("mode", None) == "embedding":
+        if model_info.get("mode", None) == "embedding":
             # this is an embedding model
             tasks.append(_check_embedding_model(litellm_params))
         else:

litellm/proxy/proxy_config.yaml

@@ -11,7 +11,8 @@ model_list:
       api_base: os.environ/AZURE_API_BASE
       api_key: os.environ/AZURE_API_KEY
       api_version: "2023-07-01-preview"
-    mode: embedding
+    model_info:
+      mode: embedding
   - model_name: openai-gpt-3.5
     litellm_params:
       model: gpt-3.5-turbo
@@ -20,7 +21,8 @@ model_list:
     litellm_params:
       model: text-embedding-ada-002
       api_key: os.environ/OPENAI_API_KEY
-    mode: embedding
+    model_info:
+      mode: embedding
 
 
 litellm_settings: