From 249910a25f1cb7f78344f8057682e4a58fcbffdb Mon Sep 17 00:00:00 2001
From: ump45nose <52391318+ump45nose@users.noreply.github.com>
Date: Thu, 27 Aug 2026 20:46:21 +0800
Subject: [PATCH] fix: normalize vLLM Whisper API base URL

---
 .../vllm_model_provider/model/whisper_sst.py  |  6 ++---
 apps/models_provider/tests.py                 | 27 +++++++++++++++++--
 2 files changed, 28 insertions(+), 5 deletions(-)

diff --git a/apps/models_provider/impl/vllm_model_provider/model/whisper_sst.py b/apps/models_provider/impl/vllm_model_provider/model/whisper_sst.py
index 12e01a98400..f776e6444ce 100644
--- a/apps/models_provider/impl/vllm_model_provider/model/whisper_sst.py
+++ b/apps/models_provider/impl/vllm_model_provider/model/whisper_sst.py
@@ -44,8 +44,8 @@ def check_auth(self):
             self.speech_to_text(audio_file)
 
     def speech_to_text(self, audio_file):
-
-        base_url = self.api_url if self.api_url.endswith('v1') else f"{self.api_url}/v1"
+        base_url = self.api_url.rstrip('/')
+        base_url = base_url if base_url.endswith('/v1') else f"{base_url}/v1"
 
         try:
             client = OpenAI(
@@ -67,4 +67,4 @@ def speech_to_text(self, audio_file):
 
         except Exception as err:
             maxkb_logger.error(f":Error: {str(err)}: {traceback.format_exc()}")
-            raise err
\ No newline at end of file
+            raise err
diff --git a/apps/models_provider/tests.py b/apps/models_provider/tests.py
index 7ce503c2dd9..44e7258f159 100644
--- a/apps/models_provider/tests.py
+++ b/apps/models_provider/tests.py
@@ -1,3 +1,26 @@
-from django.test import TestCase
+from io import BytesIO
+from unittest.mock import patch
 
-# Create your tests here.
+from django.test import SimpleTestCase
+
+from models_provider.impl.vllm_model_provider.model.whisper_sst import VllmWhisperSpeechToText
+
+
+class VllmWhisperSpeechToTextTest(SimpleTestCase):
+    @patch('models_provider.impl.vllm_model_provider.model.whisper_sst.OpenAI')
+    def test_normalizes_trailing_slash_in_v1_base_url(self, openai_mock):
+        openai_mock.return_value.audio.transcriptions.create.return_value.text = 'transcript'
+        model = VllmWhisperSpeechToText(
+            api_key='test-key',
+            api_url='https://vllm.example/v1/',
+            model='whisper',
+            params={},
+        )
+
+        result = model.speech_to_text(BytesIO(b'audio'))
+
+        openai_mock.assert_called_once_with(
+            api_key='test-key',
+            base_url='https://vllm.example/v1',
+        )
+        self.assertEqual(result, 'transcript')