fix(anthropic): emit native content blocks when faking a stream

The fake stream iterator only knew text/thinking/redacted_thinking/tool_use, so websearch interception's injected server_tool_use and web_search_tool_result blocks were reduced to a bare content_block_stop on the streaming path

litellm/llms/anthropic/experimental_pass_through/messages/fake_stream_iterator.py

@@ -90,12 +90,12 @@ class FakeAnthropicMessagesStreamIterator:
             }
             chunks.append(f"event: content_block_start\ndata: {json.dumps(content_block_start)}\n\n".encode())
 
-        elif block_type == "tool_use":
+        elif block_type in ("tool_use", "server_tool_use", "mcp_tool_use"):
             content_block_start = {
                 "type": "content_block_start",
                 "index": index,
                 "content_block": {
-                    "type": "tool_use",
+                    "type": block_type,
                     "id": block_dict.get("id"),
                     "name": block_dict.get("name"),
                     "input": {},
@@ -113,6 +113,14 @@ class FakeAnthropicMessagesStreamIterator:
             }
             chunks.append(f"event: content_block_delta\ndata: {json.dumps(content_block_delta)}\n\n".encode())
 
+        else:
+            content_block_start = {
+                "type": "content_block_start",
+                "index": index,
+                "content_block": block_dict,
+            }
+            chunks.append(f"event: content_block_start\ndata: {json.dumps(content_block_start)}\n\n".encode())
+
         content_block_stop = {"type": "content_block_stop", "index": index}
         chunks.append(f"event: content_block_stop\ndata: {json.dumps(content_block_stop)}\n\n".encode())
         return chunks

tests/test_litellm/llms/anthropic/experimental_pass_through/messages/test_fake_stream_iterator.py

@@ -0,0 +1,101 @@
+"""
+Unit tests for FakeAnthropicMessagesStreamIterator.
+
+Regression coverage for websearch interception: the native
+``server_tool_use`` / ``web_search_tool_result`` blocks that
+``_inject_native_blocks`` adds must survive the non-streaming -> streaming
+re-wrap, not collapse into a bare ``content_block_stop``.
+"""
+
+import json
+from typing import Any, Dict, List, cast
+
+from litellm.llms.anthropic.experimental_pass_through.messages.fake_stream_iterator import (
+    FakeAnthropicMessagesStreamIterator,
+)
+from litellm.types.llms.anthropic_messages.anthropic_response import (
+    AnthropicMessagesResponse,
+)
+
+
+def _events(response: Dict[str, Any]) -> List[Dict[str, Any]]:
+    return [
+        json.loads(line.removeprefix("data: "))
+        for chunk in FakeAnthropicMessagesStreamIterator(response=cast(AnthropicMessagesResponse, response))
+        for line in chunk.decode().splitlines()
+        if line.startswith("data: ")
+    ]
+
+
+def _response(content: List[Dict[str, Any]]) -> Dict[str, Any]:
+    return {
+        "id": "msg_123",
+        "type": "message",
+        "role": "assistant",
+        "model": "claude-sonnet-4-5",
+        "content": content,
+        "stop_reason": "end_turn",
+        "usage": {"input_tokens": 10, "output_tokens": 5},
+    }
+
+
+def test_web_search_tool_result_block_is_emitted_with_payload():
+    search_result_block = {
+        "type": "web_search_tool_result",
+        "tool_use_id": "srvtoolu_1",
+        "content": [
+            {
+                "type": "web_search_result",
+                "url": "https://docs.litellm.ai",
+                "title": "LiteLLM docs",
+                "page_age": None,
+            }
+        ],
+    }
+    events = _events(_response([search_result_block, {"type": "text", "text": "hi"}]))
+
+    starts = [e for e in events if e["type"] == "content_block_start"]
+    assert {"index": 0, "type": "content_block_start", "content_block": search_result_block} in starts
+    assert [e["index"] for e in events if e["type"] == "content_block_stop"] == [0, 1]
+
+
+def test_server_tool_use_block_keeps_type_and_input():
+    events = _events(
+        _response(
+            [
+                {
+                    "type": "server_tool_use",
+                    "id": "srvtoolu_1",
+                    "name": "web_search",
+                    "input": {"query": "litellm"},
+                }
+            ]
+        )
+    )
+
+    start = next(e for e in events if e["type"] == "content_block_start")
+    assert start["content_block"] == {
+        "type": "server_tool_use",
+        "id": "srvtoolu_1",
+        "name": "web_search",
+        "input": {},
+    }
+    delta = next(e for e in events if e["type"] == "content_block_delta")
+    assert json.loads(delta["delta"]["partial_json"]) == {"query": "litellm"}
+
+
+def test_unknown_block_type_is_passed_through_verbatim():
+    block = {"type": "code_execution_tool_result", "tool_use_id": "srvtoolu_2", "content": {"stdout": "42"}}
+    events = _events(_response([block]))
+
+    start = next(e for e in events if e["type"] == "content_block_start")
+    assert start["content_block"] == block
+
+
+def test_text_block_still_streams_as_delta():
+    events = _events(_response([{"type": "text", "text": "hello"}]))
+
+    start = next(e for e in events if e["type"] == "content_block_start")
+    assert start["content_block"] == {"type": "text", "text": ""}
+    delta = next(e for e in events if e["type"] == "content_block_delta")
+    assert delta["delta"] == {"type": "text_delta", "text": "hello"}