From 0346ac0fa23f3b70c0eb919cda0113c767586c1d Mon Sep 17 00:00:00 2001 From: Mason Daugherty Date: Mon, 18 May 2026 11:51:35 -0700 Subject: [PATCH] fix(ollama): trace invoke model override (#37502) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit If you override the model at invoke time — e.g. `chat.invoke(messages, model="llama3.2")` or `chat.bind(model="llama3.2")` - traces previously logged the model the `ChatOllama` instance was constructed with, not the one that ran. Tracing now reflects the runtime override, so dashboards, filters, and cost/latency breakdowns line up with what really executed. --- libs/partners/ollama/langchain_ollama/chat_models.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/libs/partners/ollama/langchain_ollama/chat_models.py b/libs/partners/ollama/langchain_ollama/chat_models.py index dc8f2ae106d..24c09277e89 100644 --- a/libs/partners/ollama/langchain_ollama/chat_models.py +++ b/libs/partners/ollama/langchain_ollama/chat_models.py @@ -1186,7 +1186,7 @@ class ChatOllama(BaseChatModel): params = self._get_invocation_params(stop=stop, **kwargs) ls_params = LangSmithParams( ls_provider="ollama", - ls_model_name=self.model, + ls_model_name=params.get("model", self.model), ls_model_type="chat", ls_temperature=params.get("temperature", self.temperature), )