feat(advisor): add AdvisorInterceptionConfig type definition

Introduces a typed config class for advisor_interception_params used
under litellm_settings in proxy config.

Made-with: Cursor

litellm/types/integrations/advisor_interception.py

@@ -0,0 +1,28 @@
+"""
+Type definitions for Advisor Interception integration.
+"""
+
+from typing import List, Optional, TypedDict
+
+
+class AdvisorInterceptionConfig(TypedDict, total=False):
+    """
+    Configuration parameters for AdvisorInterceptionLogger.
+
+    Used in proxy_config.yaml under litellm_settings:
+        litellm_settings:
+          advisor_interception_params:
+            default_advisor_model: "advisor-model"
+            enabled_providers: ["openai", "vertex_ai"]
+    """
+
+    default_advisor_model: Optional[str]
+    """
+    The model_name (from model_list) or litellm model string to use as the
+    advisor model. When running behind the proxy, this should match a
+    model_name in model_list so the router resolves the correct deployment
+    and credentials.
+    """
+
+    enabled_providers: List[str]
+    """List of LLM provider names to enable interception for (e.g., ['openai', 'vertex_ai'])"""