PaperBuddy / 文润

PaperBuddy is a Vercel-ready academic writing assistant for paper polishing, PDF paper understanding, and AI Q&A. It uses a server-side OpenAI-compatible chat completions API, with NVIDIA's endpoint and openai/gpt-oss-120b as the default configuration.

PaperBuddy 是一个可部署到 Vercel 的学术论文辅助网页应用，支持论文文字点评润色、PDF 文献理解和基于论文内容的 AI 问答。项目默认使用 NVIDIA OpenAI-compatible Chat Completions 接口和 openai/gpt-oss-120b 模型。

界面预览 / Screenshots

文润提供两个主要工作区：论文润色和文献解读。论文润色适合逐段检查学术表达，文献解读适合上传 PDF 后边读边问。

Wenrun has two main workspaces: paper polishing and PDF reading. Use paper polishing to review academic wording paragraph by paragraph, and use PDF reading to ask questions while reading a paper.

论文润色 / Paper polishing

复制论文片段后，页面会按段落展示中文翻译、问题分析、润色结果和润色后的中文翻译。

After pasting a paper excerpt, the page shows the Chinese translation, writing issues, polished text, and polished Chinese translation paragraph by paragraph.

文献解读 / PDF reading

上传 PDF 后，页面会同时展示论文预览和对话区，适合追问研究问题、方法、结论和公式含义。

After uploading a PDF, the page shows the paper preview and chat area side by side, so you can ask about research questions, methods, conclusions, and formulas.

---

中文说明

功能

• 论文润色：粘贴论文片段，按段落查看原文、问题分析和润色结果。
• PDF 文献解读：上传 PDF，浏览器本地提取文本，并基于提取内容进行多轮问答。
• 无需用户登录：API Key 只配置在部署平台的服务端环境变量中，不会暴露给前端。
• 本地历史记录：润色历史和 PDF 对话历史保存在浏览器 localStorage 中。

技术栈

• Next.js App Router
• React + TypeScript
• pdfjs-dist PDF 文本提取和页面预览
• Vitest 核心工具测试
• Vercel Serverless API Routes

本地运行

安装依赖：

npm install

复制环境变量文件：

cp .env.example .env.local

在 .env.local 中填入：

LLM_API_BASE_URL=https://integrate.api.nvidia.com/v1
LLM_API_KEY=你的_NVIDIA_API_KEY
LLM_MODEL_ID=openai/gpt-oss-120b
LLM_THINKING_INTENSITY=high
LLM_MAX_TOKENS=4096

启动开发服务：

npm run dev

打开 http://localhost:3000。

部署到 Vercel

推荐流程是先把代码推送到 GitHub，再从 Vercel 导入仓库。

1. 打开 Vercel 并登录。
2. 点击 Add New... → Project。
3. 选择 GitHub 仓库 HuayongHu/PaperBuddy。
4. Framework Preset 保持为 Next.js。
5. 在 Environment Variables 中添加：

LLM_API_BASE_URL=https://integrate.api.nvidia.com/v1
LLM_API_KEY=你的_NVIDIA_API_KEY
LLM_MODEL_ID=openai/gpt-oss-120b
LLM_THINKING_INTENSITY=high
LLM_MAX_TOKENS=4096

6. 点击 Deploy。
7. 部署完成后，打开 Vercel 给出的域名即可使用。

环境变量说明

变量名	必填	默认值	说明
LLM_API_BASE_URL	否	https://integrate.api.nvidia.com/v1	LLM API 根地址。推荐不包含 /chat/completions。
LLM_API_KEY	是	无	NVIDIA API Key，只能配置在服务端环境变量中。
LLM_MODEL_ID	否	openai/gpt-oss-120b	模型 ID。
LLM_THINKING_INTENSITY	否	high	推理强度，可选 low、medium、high。
LLM_MAX_TOKENS	否	4096	单次最大输出 token 数。

使用注意

• PDF 文本提取在浏览器本地完成，服务端不保存 PDF 文件。
• 扫描版或图片型 PDF 可能无法提取文字，问答效果会受影响。
• 润色页前端建议单次输入不超过 2000 字；服务端最大接受 10000 字符。
• PDF 问答最多使用前 60000 字符的提取文本。

常用命令

npm test
npm run build
npm run dev

---

English

What PaperBuddy does

• Paper polishing: paste an academic paragraph or section and get paragraph-by-paragraph comments and revised text.
• PDF understanding: upload a PDF, extract text in the browser, and ask follow-up questions based on the paper content.
• No user accounts: the API key is stored only as a server-side environment variable.
• Local history: polishing history and PDF chat history are stored in browser localStorage.

Tech stack

• Next.js App Router
• React + TypeScript
• pdfjs-dist for PDF text extraction and preview
• Vitest for core utility tests
• Vercel Serverless API Routes

Run locally

Install dependencies:

npm install

Copy the environment file:

cp .env.example .env.local

Set the environment variables in .env.local:

LLM_API_BASE_URL=https://integrate.api.nvidia.com/v1
LLM_API_KEY=your_nvidia_api_key
LLM_MODEL_ID=openai/gpt-oss-120b
LLM_THINKING_INTENSITY=high
LLM_MAX_TOKENS=4096

Start the development server:

npm run dev

Open http://localhost:3000.

Deploy to Vercel

The recommended workflow is to push the code to GitHub first, then import the repository from Vercel.

1. Open Vercel and sign in.
2. Click Add New... → Project.
3. Select the GitHub repository HuayongHu/PaperBuddy.
4. Keep Framework Preset as Next.js.
5. Add these Environment Variables:

LLM_API_BASE_URL=https://integrate.api.nvidia.com/v1
LLM_API_KEY=your_nvidia_api_key
LLM_MODEL_ID=openai/gpt-oss-120b
LLM_THINKING_INTENSITY=high
LLM_MAX_TOKENS=4096

6. Click Deploy.
7. Open the Vercel deployment URL when the build finishes.

Environment variables

Name	Required	Default	Description
LLM_API_BASE_URL	No	https://integrate.api.nvidia.com/v1	LLM API base URL. Prefer omitting /chat/completions.
LLM_API_KEY	Yes	None	NVIDIA API key. Keep it server-side only.
LLM_MODEL_ID	No	openai/gpt-oss-120b	Model ID.
LLM_THINKING_INTENSITY	No	high	Reasoning effort: low, medium, or high.
LLM_MAX_TOKENS	No	4096	Maximum output tokens for one response.

Notes

• PDF text extraction happens in the browser. The server does not store uploaded PDF files.
• Scanned or image-only PDFs may not produce extractable text.
• The polishing page recommends up to 2000 characters per request; the server accepts up to 10000 characters.
• PDF Q&A uses up to the first 60000 extracted characters.

Useful commands

npm test
npm run build
npm run dev

License

MIT