From 2b05b33f019181530fd99685a491e2e336ea9178 Mon Sep 17 00:00:00 2001 From: aifanatic Date: Wed, 21 Oct 2026 09:21:02 -0500 Subject: [PATCH] Python: pass judge model in evaluate_traces sample evaluate_traces() requires a keyword-only `model` argument for the evaluator LLM judge. The sample omitted it, so both calls raised TypeError before reaching Foundry. Reuse the FOUNDRY_MODEL deployment the sample already reads for the chat client. Co-Authored-By: Claude Opus 5.5 --- .../evaluation/foundry_evals/evaluate_traces_sample.py | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-) diff --git a/python/samples/05-end-to-end/evaluation/foundry_evals/evaluate_traces_sample.py b/python/samples/05-end-to-end/evaluation/foundry_evals/evaluate_traces_sample.py index bbe4d38ff8e..63ed02f171c 100644 --- a/python/samples/05-end-to-end/evaluation/foundry_evals/evaluate_traces_sample.py +++ b/python/samples/05-end-to-end/evaluation/foundry_evals/evaluate_traces_sample.py @@ -27,10 +27,12 @@ async def main() -> None: - # 1. Set up the chat client + # 1. Set up the chat client. evaluate_traces() also needs the model + # deployment that Foundry should use as the evaluator LLM judge. + model = os.environ.get("FOUNDRY_MODEL", "gpt-4o") chat_client = FoundryChatClient( project_endpoint=os.environ["FOUNDRY_PROJECT_ENDPOINT"], - model=os.environ.get("FOUNDRY_MODEL", "gpt-4o"), + model=model, credential=AzureCliCredential(), ) @@ -54,6 +56,7 @@ async def main() -> None: response_ids=response_ids, evaluators=[FoundryEvals.RELEVANCE, FoundryEvals.GROUNDEDNESS, FoundryEvals.TOOL_CALL_ACCURACY], client=chat_client, + model=model, ) print(f"Status: {results.status}") @@ -78,6 +81,7 @@ async def main() -> None: response_ids=response_ids, evaluators=[FoundryEvals.RELEVANCE, FoundryEvals.COHERENCE], client=chat_client, + model=model, ) print(f"Status: {results.status}") @@ -88,6 +92,7 @@ async def main() -> None: # agent_id="travel-bot", # evaluators=[FoundryEvals.INTENT_RESOLUTION, FoundryEvals.TASK_ADHERENCE], # client=chat_client, + # model=model, # lookback_hours=24, # )