Merge pull request #14965 from Maximgitman/issue-14962-anthropic-streaming-id

Fix Anthropic streaming IDs (#14962)

litellm/llms/anthropic/chat/handler.py

@@ -51,6 +51,7 @@ from litellm.types.utils import (
     ModelResponseStream,
     StreamingChoices,
     Usage,
+    _generate_id,
 )
 
 from ...base import BaseLLM
@@ -490,6 +491,8 @@ class ModelResponseIterator:
         self.content_blocks: List[ContentBlockDelta] = []
         self.tool_index = -1
         self.json_mode = json_mode
+        # Generate response ID once per stream to match OpenAI-compatible behavior
+        self.response_id = _generate_id()
 
         # Track if we're currently streaming a response_format tool
         self.is_response_format_tool: bool = False
@@ -765,6 +768,7 @@ class ModelResponseIterator:
                     )
                 ],
                 usage=usage,
+                id=self.response_id,
             )
 
             return returned_chunk
@@ -936,4 +940,4 @@ class ModelResponseIterator:
             data_json = json.loads(str_line[5:])
             return self.chunk_parser(chunk=data_json)
         else:
-            return ModelResponseStream()
+            return ModelResponseStream(id=self.response_id)

tests/test_litellm/llms/anthropic/chat/test_anthropic_chat_handler.py

@@ -439,3 +439,24 @@ def test_multiple_tools_streaming_has_index_zero():
             assert (
                 parsed.choices[0].index == 0
             ), f"Expected index=0, got {parsed.choices[0].index}"
+
+
+def test_streaming_chunks_have_stable_ids():
+    iterator = ModelResponseIterator(
+        streaming_response=MagicMock(), sync_stream=False, json_mode=False
+    )
+    first_chunk = {
+        "type": "content_block_delta",
+        "index": 0,
+        "delta": {"type": "text_delta", "text": "Hello"},
+    }
+    second_chunk = {
+        "type": "content_block_delta",
+        "index": 0,
+        "delta": {"type": "text_delta", "text": " world"},
+    }
+
+    response_one = iterator.chunk_parser(chunk=first_chunk)
+    response_two = iterator.chunk_parser(chunk=second_chunk)
+
+    assert response_one.id == response_two.id == iterator.response_id