fix(anthropic): silently drop trailing chunks after usage; remove dead _polyfill_result key

- streaming_iterator: in sync __next__, after the usage chunk has been
  merged and emitted, silently consume any trailing provider events
  via 'continue' instead of forwarding them through the queue. Trailing
  chunks would translate to content_block_delta or message_delta and
  violate Anthropic SSE ordering after the final message_delta. The
  async __anext__ already drops these via 'if not self.queued_usage_chunk:'
  gating, so this aligns sync and async behavior.

- handler: drop unused '_polyfill_result' from ANTHROPIC_ONLY_REQUEST_KEYS.
  PolyfillResult is passed as an explicit arg to the adapter methods, never
  through extra_kwargs, so the entry was dead code.

Co-authored-by: Yassin Kortam <yassin@berri.ai>

litellm/llms/anthropic/experimental_pass_through/adapters/handler.py

@@ -37,7 +37,7 @@ if TYPE_CHECKING:
 
 # Anthropic-only keys already mapped by the translator; strip on extra_kwargs re-merge.
 ANTHROPIC_ONLY_REQUEST_KEYS: frozenset[str] = frozenset(
-    {"output_config", "context_management", "_polyfill_result"}
+    {"output_config", "context_management"}
 )
 
 

litellm/llms/anthropic/experimental_pass_through/adapters/streaming_iterator.py

@@ -208,11 +208,13 @@ class AnthropicStreamWrapper(AdapterCompletionStreamWrapper):
                     return self.chunk_queue.popleft()
 
                 if self.queued_usage_chunk:
-                    # Usage has already been merged + emitted. Pass any trailing
-                    # provider events through directly instead of silently
-                    # dropping them in the for-loop body.
-                    self.chunk_queue.append(processed_chunk)
-                    return self.chunk_queue.popleft()
+                    # Usage has already been merged + emitted. Any trailing
+                    # provider events would violate Anthropic SSE ordering
+                    # (no chunks may follow the final ``message_delta``), so
+                    # silently drop them — matches the async ``__anext__``
+                    # behavior where the block-handling logic is gated on
+                    # ``not self.queued_usage_chunk``.
+                    continue
 
                 if should_start_new_block and not self.sent_content_block_finish:
                     # Queue the sequence: content_block_stop -> content_block_start