scriptoriumadmin/litellm
1
0
Fork 0
mirror of https://github.com/BerriAI/litellm.git synced 2026-09-10 22:41:41 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 129

fix(ollama): remap tool role to tool_responses for Gemma models

Browse source 
Gemma models served via Ollama expect the role "tool_responses" for
tool result messages instead of the OpenAI-standard "tool" role.
Without this mapping the model ignores the tool result and enters
an infinite tool-calling loop.

Fixes #28530
This commit is contained in:
PRABHU KIRAN VANDRANKI 2026-05-21 21:26:22 -04:00
parent f69b9d6564
commit 81f1dcba2d
1 changed files with 7 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

8
litellm/llms/ollama/chat/transformation.py
View file

@ -287,8 +287,14 @@ class OllamaChatConfig(BaseConfig):
content_str = convert_content_list_to_str(cast(AllMessageValues, m))
images = extract_images_from_message(cast(AllMessageValues, m))
role = cast(str, m.get("role"))
# Gemma models (via Ollama) expect "tool_responses" for tool result
# messages instead of the OpenAI-standard "tool" role. Without this,
# the model ignores the result and loops infinitely.
if role == "tool" and "gemma" in model.lower():
role = "tool_responses"
ollama_message = OllamaChatCompletionMessage(
role=cast(str, m.get("role")),
role=role,
)
if reasoning_content is not None:
ollama_message["thinking"] = reasoning_content