☰

Qwant AI Search Summary

Summarize Qwant search results using OpenRouter AI API.

如何安装

您需要先安装一个扩展，例如 篡改猴、Greasemonkey 或 暴力猴，之后才能安装此脚本。

You will need to install an extension such as Tampermonkey to install this script.

您需要先安装一个扩展，例如 篡改猴 或 暴力猴，之后才能安装此脚本。

您需要先安装一个扩展，例如 篡改猴 或 Userscripts ，之后才能安装此脚本。

您需要先安装一款用户脚本管理器扩展，例如 Tampermonkey，才能安装此脚本。

您需要先安装用户脚本管理器扩展后才能安装此脚本。

（我已经安装了用户脚本管理器，让我安装！）

如何安装

您需要先安装一款用户样式管理器扩展，比如 Stylus，才能安装此样式。

您需要先安装一款用户样式管理器扩展，比如 Stylus，才能安装此样式。

您需要先安装一款用户样式管理器扩展，比如 Stylus，才能安装此样式。

您需要先安装一款用户样式管理器扩展后才能安装此样式。

您需要先安装一款用户样式管理器扩展后才能安装此样式。

您需要先安装一款用户样式管理器扩展后才能安装此样式。

（我已经安装了用户样式管理器，让我安装！）

作者
yodaluca23
日安装量
0
总安装量
3
评分
0 0 0
版本
1.0.2
创建于
2025-06-10
更新于
2025-06-12
大小
21.6 KB
许可证
MIT
适用于

Qwant AI Search Summary Userscript

This is a very experimental userscript that uses the OpenRouter.ai API to generate AI-based summaries of Qwant search results, rendered directly in the sidebar. All the built-in/preset models are free in OpenRouters so you just need to add your API key.

Note: This script is barely held together. The code is pretty bad and full of rough edges, but it works — mostly. I'm publishing this earlier than intended since I'm switching to Kagi and won’t be using Qwant anymore.

🚀 Features

  • Summarizes Qwant search results using a chosen LLM from OpenRouter.
  • Renders output with Markdown and LaTeX (via Showdown-KaTeX).
  • Caches previous results for performance and API cost savings.
  • Customizable prompt and model settings through a user menu.
  • Rudimentary cache management and error handling.
  • API key setup via prompt — no need to hardcode.

🧠 Known Limitations

  • Markdown/LaTeX formatting is inconsistent. The model sometimes messes up Markdown or LaTeX despite the detailed system prompt. There's not much I can do here, I've already put a lot of effort into writing a precise, verbose system prompt.
  • It may break at any time. Qwant's site structure might change, OpenRouter might change behavior, or the model output might just stop parsing correctly.
  • The code is messy. Honestly, it's kind of garbage. I'll try to fix it if it breaks, but no promises — I’m moving on to other things.

🔧 Setup

  1. Install this script from this GreasyFork page.
  2. Set your OpenRouter API key via the menu (Userscript > Settings).
  3. Optionally customize the AI model or prompt.

🛠️ Customization

Once installed, you can:

  • Change models (DeepSeek, Gemma, etc.)
  • Write your own prompt for the AI to follow
  • Clear or resize cache
  • Reset settings

💡 Models & Rate Limits

The userscript includes several built-in models (predefined in the UI), and you can use any model from OpenRouter—both free (:free) and paid—just by entering your API key.

To unlock better rate limits (even for free models), it's highly recommended to add at least $10 in credits:

  • < $10 credits: 50 free model requests/day
  • ≥ $10 credits: 1000 free model requests/day

🔁 All users are limited to 20 requests/minute, and extreme usage may be blocked by DDoS protection.

Just click the userscript menu button and follow the prompts.

📦 Dependencies

📜 License

MIT — use and modify freely, but please don’t expect support.

🙏 Thanks

This was built out of frustration and curiosity — if it helps you, that’s cool. But really, consider switching to Kagi if you’re looking for quality search + AI integration.

Disclaimer: This script uses third-party AI services. The summaries may be inaccurate or misleading. Always double-check important information.