Missing tokens or bytes in the middle of Google AI Studio responses (?)

[hosiken]

The response from Google AI Studio will be incomplete, with garbled characters and incomplete code. This results in missing parts of the code and garbled multibyte characters.

• Model: Gemini-1.5-flash-latest or Gemini-1.5-pro-latest
• Prompt that reproduces this issue (Not limited to this, but easy to confirm)
  • シベリアンハスキーのかわいさを標準出力に向かって熱弁するPythonコード
  • Python code that speaks enthusiastically about the cuteness of Siberian huskies to standard output.
• VSCode Version: 1.91.1 (user setup) f1e16e1e6214d7c44d078b1f0607b2388f29d729
• CodeGPT Version: v3.4.47 and v3.4.55 (pre-release), Language setting is Japanese
• Windows 10 21H1、Using under the firewall
• OpenAI (gpt-4o-mini) does not have the problem

def husky_love():
  """シベリアンハスキーのか���������������弁する関数"""

  print("みなさん、こんにちは！今日は��ベリアンハスキー����いて語らせてください！")
  print("あの美しいブルーの瞳、凛々しい顔立ち、そしてふわふわの白い毛�����！\")\n  print(\�����い������！\")\n  print(\��と������えたがり、遊び相手には元気いっぱいに走り回ります！")
  print("その姿を見ていると、こっちまで元気をもらえます！")
  print("そして、忘れてはいけないのが、あの美しい鳴き���！")
  print("まるでオオカミの遠吠えのような、力強くも美しい歌声は、心を揺さぶるものがあります！")
  print("シベリアンハスキーは、まさに完璧な犬種！")
  print("もし、あなたが犬を飼うことを���え��いる��ら、ぜひシベリアンハスキーを検討してみてください！")
  print("きっと、あなたの人生を豊かなものにしてくれるでしょう！")

husky_love()

import random

# List of enthusiastic phrases
phrases = [
    "OMG, Siberian Huskies are the CUTEST things EVER!",
    "Their fluffy fur and piercing blue eyes are just too much to handle!",
    "I can't even with their adorable little noses and playful personalities!",
    \Their howl is music to my ears, even if it's at 3 AM!",
    "I would do anything for a cuddle with a Siberian Husky!",
    "They're the perfect blend of majestic and adorable!",
    \ = random.choice(phrases)

# Print the phrase to standard output
print(enthusiastic_phrase)

[hosiken]

The blank user prompt in the chat history which can be seen in the screenshot is also maybe a bug.

[hosiken]

This bug no longer occurs in CodeGPT v3.5.39 and appears to be fixed, so I'm closing the issue.