From 1d63a853bb7d639fc4cab0a658d935b0f88136ca Mon Sep 17 00:00:00 2001
From: Matt Van Horn <455140+mvanhorn@users.noreply.github.com>
Date: Mon, 22 Jun 2026 18:13:56 -0700
Subject: [PATCH] fix(retool): coerce list prompt to str in reward_func

With --apply-chat-template, sample.prompt is a list[dict], so
sample.prompt + sample.response raised TypeError. Coerce to str first,
matching slime's own str(sample.prompt) idiom in sglang_rollout.

Fixes #1829
---
 examples/retool/generate_with_retool.py | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/examples/retool/generate_with_retool.py b/examples/retool/generate_with_retool.py
index ebb5c4f1a..e0bb33d64 100644
--- a/examples/retool/generate_with_retool.py
+++ b/examples/retool/generate_with_retool.py
@@ -410,7 +410,8 @@ async def reward_func(args, sample, **kwargs):
         raise TypeError("Sample must be an instance of Sample class.")
 
     # Build complete solution string
-    solution_str = sample.prompt + sample.response
+    prompt_str = sample.prompt if isinstance(sample.prompt, str) else str(sample.prompt)
+    solution_str = prompt_str + sample.response
 
     # Get ground truth answer - label is a string, not a dict
     ground_truth = sample.label if sample.label is not None else ""