github/langchain
1
0
Fork 0
mirror of https://github.com/hwchase17/langchain.git synced 2025-09-12 00:11:17 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix(anthropic): remove beta header warning for TTL (#32832)

Browse Source 
No longer beta as of Aug 13
This commit is contained in:
Mason Daugherty
2025-09-05 14:28:58 -04:00
committed by GitHub
parent 0b3af47335
commit 4e39c164bb
1 changed files with 12 additions and 26 deletions
Download Patch File Download Diff File Expand all files Collapse all files

38
libs/partners/anthropic/langchain_anthropic/chat_models.py
View File

@@ -999,24 +999,27 @@ class ChatAnthropic(BaseChatModel):
.. dropdown:: Extended caching .. dropdown:: Extended caching
.. versionadded:: 0.3.15
The cache lifetime is 5 minutes by default. If this is too short, you can The cache lifetime is 5 minutes by default. If this is too short, you can
apply one hour caching by enabling the ``'extended-cache-ttl-2025-04-11'`` apply one hour caching by setting ``ttl`` to ``'1h'``.
beta header:
.. code-block:: python .. code-block:: python
llm = ChatAnthropic( llm = ChatAnthropic(
model="claude-3-7-sonnet-20250219", model="claude-3-7-sonnet-20250219",
betas=["extended-cache-ttl-2025-04-11"],
) )
and specifying ``"cache_control": {"type": "ephemeral", "ttl": "1h"}``. messages = [{
"role": "user",
"content": [
{
"type": "text",
"text": f"{long_text}",
"cache_control": {"type": "ephemeral", "ttl": "1h"},
},
],
}]
.. important:: response = llm.invoke(messages)
Specifying a `ttl` key under `cache_control` will not work unless the
beta header is set!
Details of cached token counts will be included on the ``InputTokenDetails`` Details of cached token counts will be included on the ``InputTokenDetails``
of response's ``usage_metadata``: of response's ``usage_metadata``:
@@ -1432,23 +1435,6 @@ class ChatAnthropic(BaseChatModel):
# If cache_control is provided in kwargs, add it to last message # If cache_control is provided in kwargs, add it to last message
# and content block. # and content block.
if "cache_control" in kwargs and formatted_messages: if "cache_control" in kwargs and formatted_messages:
cache_control = kwargs["cache_control"]
# Validate TTL usage requires extended cache TTL beta header
if (
isinstance(cache_control, dict)
and "ttl" in cache_control
and (
not self.betas or "extended-cache-ttl-2025-04-11" not in self.betas
)
):
msg = (
"Specifying a 'ttl' under 'cache_control' requires enabling "
"the 'extended-cache-ttl-2025-04-11' beta header. "
"Set betas=['extended-cache-ttl-2025-04-11'] when initializing "
"ChatAnthropic."
)
warnings.warn(msg, stacklevel=2)
if isinstance(formatted_messages[-1]["content"], list): if isinstance(formatted_messages[-1]["content"], list):
formatted_messages[-1]["content"][-1]["cache_control"] = kwargs.pop( formatted_messages[-1]["content"][-1]["cache_control"] = kwargs.pop(
"cache_control" "cache_control"