`settings?active=provider&provider=openai` setting page abnormal

[gemnioo]

📱 Client Type

Web (Desktop Browser)

💻 Operating System

macOS

📦 Deployment Platform

Self hosting Docker

🔧 Deployment Mode

server db (lobe-chat-database image)

📌 Version

v2.0.0-next.155

🌐 Browser

Firefox

🐛 Bug Description

When the AI Service Provider settings page is not full-screen, only adding the OpenAI setting causes a flicker! Other providers are fine.

AI Service Provider 设置页面 非全屏下只有添加 OpenAI 设置会出现抖动！其他 Provider 正常

and:

User Output stage output characters are parsed again as Thinking! (Likely a known issue, reproducible in both GPT Responses API and Gemini. After investigation, it seems to be caused by LobeChat's default parsing of the field.

User Output 阶段输出 字符会被再次解析成 Thinking ！(估计是已知问题 在 GPT Responses API 以及 Gemini 下均可复现, 排查后发现可能是 LobeChat 默认解析 字段导致的

📷 Recurrence Steps

No response

🚦 Expected Behavior

No response

📝 Additional Information

No response

🛠️ Willing to Submit a PR?

None

✅ Validations

• Read the docs.
• Check that there isn't already an issue that reports the same bug to avoid creating a duplicate.
• Make sure this is a LobeChat issue and not a third-party library or provider issue.
• Check that this is a concrete bug. For Q&A, please use GitHub Discussions or join our Discord Server.

[github-actions]

Found 1 possible duplicate issue:

1. [Bug] 对于推理模型，能否进一步优化对于输入和输出中<think>标签的处理方式？ #6068

This issue will be automatically closed as a duplicate in 3 days.

• If your issue is a duplicate, please close it and 👍 the existing issue instead
• To prevent auto-closure, add a comment or 👎 this comment

🤖 Generated with Claude Code

[dosubot]

For the OpenAI provider settings flicker in non-full-screen mode: this is likely caused by the extra conditional fields and switches in the OpenAI settings (like proxy URL, API key, responses API toggle, client fetch toggle), which trigger more frequent form reflows and loading skeletons. Combined with sticky headers and Suspense skeletons in the model list, this can cause visible flicker, especially when the window isn't full-screen. Other providers have simpler forms, so they don't trigger the same re-layouts. The recent change to conditionally fetch provider lists based on mobile view may also affect rendering behavior in different screen modes, but the core issue is the more complex OpenAI config UI and its interaction with responsive layout and loading states see code details.

For the repeated parsing of <think></think> as "Thinking": both the backend and frontend treat these tags as signals for reasoning content. The backend strips the tags but uses them to toggle a "thinking" state, emitting reasoning chunks. The frontend normalizes these tags and uses a custom markdown pipeline and renderer to visualize them as "Thinking" overlays. If the tags remain in the output, they're re-interpreted by the markdown renderer, leading to duplicate or persistent "Thinking" states see code details.

No official workarounds are documented, but using full-screen mode or minimizing rapid config changes may reduce flicker. For the <think> issue, filtering or removing these tags from prompts/output before rendering can help avoid repeated "Thinking" overlays.

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}