litellm - 💡(How to fix) Fix bug: Responses API drops cache_control on input_text content blocks (inconsistent with tool cache_control)

from litellm.responses.litellm_completion_transformation.transformation import (
    LiteLLMCompletionResponsesConfig,
)

# Case 1: input_text with cache_control — DROPPED
result = LiteLLMCompletionResponsesConfig.transform_responses_api_request_to_chat_completion_request(
    model="claude-sonnet-4-6",
    input=[{
        "type": "message",
        "role": "user",
        "content": [{
            "type": "input_text",
            "text": "Hello",
            "cache_control": {"type": "ephemeral", "ttl": "1h"},
        }],
    }],
    responses_api_request={},
)
messages = result.get("messages", [])
content = messages[-1].get("content", [])
print(content[0].get("cache_control"))  # None — cache_control was silently dropped

# Case 2: tool with cache_control — PRESERVED (works correctly)
result2 = LiteLLMCompletionResponsesConfig.transform_responses_api_request_to_chat_completion_request(
    model="claude-sonnet-4-6",
    input=[],
    responses_api_request={
        "tools": [{"type": "function", "name": "f", "cache_control": {"type": "ephemeral"}}]
    },
)
print(result2["tools"][0].get("cache_control"))  # {"type": "ephemeral"} — preserved correctly

---

# BUGGY (current):
content_list.append({
    "type": LiteLLMCompletionResponsesConfig._get_chat_completion_request_content_type(
        item.get("type") or "text"
    ),
    "text": text_value,
    # cache_control from item is never read
})

---

if tool.get("cache_control"):
    chat_completion_tool["cache_control"] = tool.get("cache_control")

---

# In _transform_responses_api_content_to_chat_completion_content():
text_block = {
    "type": LiteLLMCompletionResponsesConfig._get_chat_completion_request_content_type(
        item.get("type") or "text"
    ),
    "text": text_value,
}
if item.get("cache_control") is not None:
    text_block["cache_control"] = item["cache_control"]
content_list.append(text_block)