scriptoriumadmin/litellm
1
0
Fork 0
mirror of https://github.com/BerriAI/litellm.git synced 2026-11-12 08:42:18 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 207

fix(vertex_ai): read message roles by subscript in the prefix check

Browse source 
m.get on the message union reported Unknown member types against the
repo-wide type gate; role is required on every AllMessageValues shape,
so subscript access is both typed and safe here.
This commit is contained in:
Mrun25 2026-10-17 11:55:01 +05:30
parent cef76f7c75
commit b58633fbd1
1 changed files with 1 additions and 3 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

4
litellm/llms/vertex_ai/context_caching/transformation.py
View file

@ -147,9 +147,7 @@ def separate_cached_messages(
prefix_messages: Final = messages[:first_cached_idx]
suffix_messages: Final = messages[last_cached_idx + 1 :]
if any(m.get("role") != "system" for m in prefix_messages) or any(
m.get("role") == "system" for m in suffix_messages
):
if any(m["role"] != "system" for m in prefix_messages) or any(m["role"] == "system" for m in suffix_messages):
# Not a true prefix: send the conversation uncached so nothing is
# reordered and no system message is silently dropped.
return [], list(messages)