From 11e6d40c7a9e53c15a1936aee183fc612c624f43 Mon Sep 17 00:00:00 2001
From: CoreJa <i@coreja.com>
Date: Mon, 26 May 2025 20:10:10 +0800
Subject: [PATCH] bug: fix fn_call error during API response (#8695)

Co-authored-by: Engel Nyst <engel.nyst@gmail.com>
---
 openhands/llm/fn_call_converter.py      |  8 +++++--
 tests/unit/test_llm_fncall_converter.py | 28 +++++++++++++++++++++++++
 2 files changed, 34 insertions(+), 2 deletions(-)

diff --git a/openhands/llm/fn_call_converter.py b/openhands/llm/fn_call_converter.py
index 35ae68b23e..09079d4d76 100644
--- a/openhands/llm/fn_call_converter.py
+++ b/openhands/llm/fn_call_converter.py
@@ -536,8 +536,12 @@ def convert_fncall_messages_to_non_fncall_messages(
             if isinstance(content, str):
                 content = prefix + content
             elif isinstance(content, list):
-                if content and content[-1]['type'] == 'text':
-                    content[-1]['text'] = prefix + content[-1]['text']
+                if content and (
+                    first_text_content := next(
+                        (c for c in content if c['type'] == 'text'), None
+                    )
+                ):
+                    first_text_content['text'] = prefix + first_text_content['text']
                 else:
                     content = [{'type': 'text', 'text': prefix}] + content
             else:
diff --git a/tests/unit/test_llm_fncall_converter.py b/tests/unit/test_llm_fncall_converter.py
index a8a5d816fb..7d3694086b 100644
--- a/tests/unit/test_llm_fncall_converter.py
+++ b/tests/unit/test_llm_fncall_converter.py
@@ -994,3 +994,31 @@ def test_convert_fncall_messages_without_cache_control():
         result[0]['content'][0]['text']
         == 'EXECUTION RESULT of [test_tool]:\ntest content'
     )
+
+
+def test_convert_fncall_messages_with_image_url():
+    """Test that convert_fncall_messages_to_non_fncall_messages handles image URLs correctly."""
+    messages = [
+        {
+            'role': 'tool',
+            'name': 'browser',
+            'content': [
+                {
+                    'type': 'text',
+                    'text': 'some browser tool results',
+                },
+                {
+                    'type': 'image_url',
+                    'image_url': {'url': 'data:image/gif;base64,R0lGODlhAQABAAAAACw='},
+                },
+            ],
+        }
+    ]
+    converted_messages = convert_fncall_messages_to_non_fncall_messages(messages, [])
+    assert len(converted_messages) == 1
+    assert converted_messages[0]['role'] == 'user'
+    assert len(converted_messages[0]['content']) == len(messages[0]['content'])
+    assert (
+        next(c for c in converted_messages[0]['content'] if c['type'] == 'text')['text']
+        == f'EXECUTION RESULT of [{messages[0]["name"]}]:\n{messages[0]["content"][0]["text"]}'
+    )