scriptoriumadmin/litellm
1
0
Fork 0
mirror of https://github.com/BerriAI/litellm.git synced 2026-10-06 02:48:13 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 103

fix(bedrock): preserve assistant cache point boundaries

Browse source
This commit is contained in:
amazingniuma-art 2026-09-28 00:38:43 +08:00
parent 9202ac133a
commit 2ca368d27d
3 changed files with 5800 additions and 5759 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

11530
litellm/litellm_core_utils/prompt_templates/factory.py
View file

File diff suppressed because it is too large Load diff

2
litellm/llms/anthropic/chat/transformation.py
View file

@ -1719,7 +1719,7 @@ class AnthropicConfig(AnthropicModelInfo, BaseConfig):
anthropic_system_message_content["cache_control"] = system_message_block["cache_control"]
anthropic_system_message_list.append(anthropic_system_message_content)
elif isinstance(message["content"], list):
content_start: int = len(anthropic_system_message_list)
content_start: int = len(anthropic_system_message_list) # pyright: ignore[reportGeneralTypeIssues] # reset per system message
for _content in message["content"]:
# Skip empty text blocks - Anthropic API raises errors for empty text
text_value = _content.get("text")

27
tests/unit/litellm_core_utils/prompt_templates/test_litellm_core_utils_prompt_templates_factory.py
View file

@ -185,6 +185,33 @@ async def test_bedrock_assistant_list_preserves_message_cache_control(block_cach
assert "toolUse" in blocks[-2]
@pytest.mark.asyncio
async def test_bedrock_assistant_message_cache_point_stays_before_following_assistant_text():
messages: Final = [
{"role": "user", "content": "hi"},
{
"role": "assistant",
"content": [{"type": "text", "text": "cached prefix"}],
"cache_control": {"type": "ephemeral"},
},
{"role": "assistant", "content": [{"type": "text", "text": "uncached suffix"}]},
]
model: Final = "us.anthropic.claude-sonnet-4-5-20250929-v1:0"
expected: Final = [
{"text": "cached prefix"},
{"cachePoint": {"type": "default"}},
{"text": "uncached suffix"},
]
sync_result: Final = _bedrock_converse_messages_pt(messages=messages, model=model, llm_provider="bedrock")
async_result: Final = await BedrockConverseMessagesProcessor._bedrock_converse_messages_pt_async(
messages=messages, model=model, llm_provider="bedrock"
)
assert sync_result[-1]["content"] == expected
assert async_result[-1]["content"] == expected
def _get_gemini_function_response_inline_data_parts(result):
assert isinstance(result, list), "expected Gemini parts list"
assert len(result) == 1, "multimodal function responses should stay in one part"