chore: drop added commentary per repo guidance

litellm/google_genai/streaming_iterator.py

@@ -79,13 +79,7 @@ class BaseGoogleGenAIGenerateContentStreamingIterator:
     async def _handle_async_streaming_logging(
         self,
     ):
-        """Handle the logging after all chunks have been collected.
-
-        ``url_route`` below carries no hostname, so the pass-through handler cannot
-        sniff the provider off it and would price every Google stream at Vertex rates.
-        ``custom_llm_provider`` is forwarded so costing uses the provider already
-        resolved for this request.
-        """
+        """Handle the logging after all chunks have been collected."""
         from litellm.proxy.pass_through_endpoints.streaming_handler import (
             PassThroughStreamingHandler,
         )

litellm/proxy/pass_through_endpoints/llm_provider_handlers/vertex_passthrough_logging_handler.py

@@ -404,11 +404,6 @@ class VertexPassthroughLoggingHandler:
         - Builds complete response from chunks
         - Creates standard logging object
         - Logs in litellm callbacks
-
-        ``custom_llm_provider`` is the provider the caller already resolved. Callers that
-        only know the upstream URL leave it unset and the provider is sniffed from the
-        hostname instead. Native google_genai streams pass it explicitly because their
-        url_route carries no hostname to sniff.
         """
         kwargs: dict[str, Any] = {}
         model = model or VertexPassthroughLoggingHandler.extract_model_from_url(url_route)

litellm/proxy/pass_through_endpoints/streaming_handler.py

@@ -138,9 +138,6 @@ class PassThroughStreamingHandler:
         - Anthropic
         - Vertex AI
         - OpenAI
-
-        ``custom_llm_provider`` lets a caller that already knows the provider skip the
-        URL-hostname sniffing the pass-through handlers fall back to.
         """
         try:
             (

tests/test_litellm/proxy/pass_through_endpoints/llm_provider_handlers/test_vertex_passthrough_logging_handler.py

@@ -11,8 +11,6 @@ from litellm.proxy.pass_through_endpoints.llm_provider_handlers.vertex_passthrou
 )
 from litellm.types.utils import ModelResponse, Usage
 
-# gemini/ and vertex_ai/ price this model differently (2x), so it is the only way to
-# observe which rate card a code path actually used.
 MODEL = "gemini-3.1-flash-image"
 PROMPT_TOKENS = 1000
 COMPLETION_TOKENS = 500