给不用桌面客户端、也不会装软件的老师做的:装个浏览器插件,在网页版学习通上 直接读题、判学科、搜答案,选择题替你点中正确选项。
这是桌面版 StudyAnswerHelper 的浏览器插件分支, 界面与功能对齐它的 v1.3.0(四个标签页、学科识别、选择题点选、答案队列、题库、历史)。
桌面版要下载安装 exe,而且读题靠截图识别、点选项靠系统无障碍接口。 这个插件不需要安装任何软件,并且因为题目本身就是网页上的元素, 读题、点选项、填答案都是直接对元素操作 —— 更快、更准,也不存在 "目标程序不暴露界面结构"这类问题。
| 桌面版 | 本插件 | |
|---|---|---|
| 安装 | 下载 exe 并安装 | 浏览器里加载一次文件夹即可 |
| 读题 | 截图 → OCR | 直接读页面(首选);题目是图片时才截图 |
| 点选项 | 系统无障碍接口 | 直接点页面元素 |
| 填答案 | 模拟键盘逐字输入 | 直接写入答题框 |
| 数据存放 | 电脑上的 json 文件 | 浏览器自己的存储空间 |
- 下载
StudyAnswerHelperWeb-v1.0.0.zip,解压到一个你不会乱删的文件夹 (比如D:\学习通答题助手。⚠️ 解压出来的文件夹不能删,插件是从这里运行的) - 打开浏览器(Chrome / Edge / 360 / 任何 Chromium 内核的都行),地址栏输入
chrome://extensions(Edge 是edge://extensions)回车 - 右上角打开 「开发者模式」,然后点 「加载已解压的扩展程序」, 选中第 1 步解压出来的文件夹 —— 完成
装好后浏览器右上角的拼图图标里会出现一个它的图标,建议点图钉把它固定到工具栏。
Windows 可能会弹"已保护你的电脑"之类提示 —— 这是所有未上架商店的扩展都会有的, 因为开发者没有购买代码签名证书。点"仍要运行 / 继续"即可。源码完全公开,可自行核对。
-
打开学习通的答题页面
-
点一下工具栏上的插件图标,面板打开,它会自动读出当前这道题
-
点「搜答案」→ 答案会落在面板的大框里
-
最后一步二选一:
- 选择题:点「填入页面」,它会直接把正确选项点中(界面上会先把正确字母高亮给你看)
- 填空 / 简答:点「填入页面」,答案直接写进答题框
想自己先看看答案对不对,就先不点「填入页面」,改完再点。
几个按钮:
| 按钮 | 作用 |
|---|---|
| 读当前题目 | 重新读页面上这道题(切到下一题后点它) |
| 搜答案 | 三路来源一起找答案:本地题库 / 网络检索 / AI |
| 填入页面 | 选择题点选项,其他题型写进答题框 |
| 把答案存为一条 | 攒多条答案,按顺序一道道填 |
| 存入题库 | 这题的答案靠谱就存起来,下次离线秒出 |
| 截图识题 | 题目是图片(不是文字)时用:拖框圈住它,交给视觉模型读成文字 |
- 自动判学科:数学 / 语文 / 英语 / 物理 / 化学 / 生物 / 历史 / 地理 / 道德与法治 / 信息技术。 判断是纯本地的(不联网、不调模型、瞬间出结果),界面上会显示学科和置信度。 判错了?旁边的下拉框当场改成正确的学科,它会记住。
- 选择题点选:题目有
A. B. C. D.、答案也能落到某个字母时, 界面会把正确字母高亮,点「填入页面」就替你点中它。多选题(AC)逐个点。 - 拿不准就不点:认不出选项、或者答案不是明确字母时,它不会乱点, 而是告诉你为什么(比如"答案里没有明确的选项字母")。
不配也能用:本地题库 + 网络检索都不需要任何密钥。
配了更好(尤其数学题):设置页填一个 OpenAI 兼容接口的地址、模型名和 Key 即可。
DeepSeek / 智谱 / Kimi / 通义 / SiliconFlow / 本机 Ollama 都走同一套协议。
智谱的 glm-4v-flash 有免费额度,搜题和识图都能用。
- 题目、答案、题库、历史全都只存在你自己的浏览器里,不会上传到任何服务器。
- 联网只发生在两种时候:
- 勾了「网络检索」→ 只发出题干文字
- 用了「AI 解答 / 截图识题」→ 只发出题干文字,或你框出来的那一块截图
- 没有账号、没有埋点、不记录你的操作、不碰你的剪贴板。
- API Key 存在浏览器本地存储里,面板上不会显示明文。
- 清理浏览器数据会一并清掉题库,重要的题库请用「题库管理 → 导出全部」存一份文本。
Q:面板提示"这个页面上脚本没生效"?
A:确认打开的是学习通 / 超星的答题页面(chaoxing.com 或 xuexitong.com 域名),
然后刷新一次页面再点图标。插件只对这两个域名生效,别的网站不会打扰你。
Q:读到的题目不对 / 读不出来? A:页面结构太多变了。先点搜答案页的「手动改题」自己贴一道题用着; 如果某类页面固定读不出来,把页面地址告诉我,我去补对应的选择器。
Q:点了「填入页面」没反应? A:多半是答题框还没渲染完,或光标没在页面上。先点一下页面空白处,再点「填入页面」。 如果还不行,改用手填 —— 答案就显示在大框里,可以直接复制。
Q:AI 报 401 / 连接失败?
A:检查密钥有没有填、接口地址和模型名是不是一对(比如智谱的地址要配 glm-4v-flash 这类模型名)。
设置页有「测试连接」可以一键验证。
Q:会不会被网站检测到? A:它只是像你一样在页面上点击和填写,没有绕过任何验证,也没有自动提交的能力 —— 最终提交永远是你自己点的那一下。
chrome://extensions → 找到它 → 点「移除」。之后解压出来的文件夹可以直接删掉。
A Chrome/Edge extension branch of the desktop app StudyAnswerHelper. It is for teachers who have no desktop client and cannot install software — they use the web version of Xuexitong (Chaoxing) in a browser. The UI and feature set mirror the desktop v1.3.0: four tabs, subject detection, automatic option clicking, answer queue, local question bank, and history.
Unlike the desktop app, this extension works directly on the page's DOM: it reads the question from the page, clicks the correct option as a real element, and writes the answer straight into the answer field. There is no screen capture, no OCR, and no accessibility interface involved in the normal flow.
- Download
StudyAnswerHelperWeb-v1.0.0.zipand unzip it somewhere permanent (e.g.D:\StudyAnswerHelperWeb). Do not delete that folder — the extension runs from it. - Open
chrome://extensions(oredge://extensions), enable Developer mode. - Click Load unpacked and pick the folder. Done.
Windows may warn that the app is unrecognised — expected for any extension not published through a store, because a code-signing certificate costs money. The source is public.
Open a Xuexitong assignment page, click the extension icon — it reads the current question automatically. Press 搜答案 (Search) to look it up, then 填入页面 (Fill in): for multiple-choice it clicks the correct option, otherwise it writes the answer into the answer box.
Subject detection and option clicking behave exactly as in the desktop app: the subject chip shows the detected subject plus a confidence value, and the correct letter is highlighted before anything is clicked. When it cannot resolve a letter it refuses to click and says why.
No — the local bank and web search need nothing. Adding an OpenAI-compatible endpoint (such as DeepSeek, Zhipu, Moonshot, Qwen, SiliconFlow, or a local Ollama) improves answers, especially for maths, and is required for reading questions that are images.
Everything (questions, answers, bank, history, key) stays in your own browser storage. The only data that ever leaves is the question text, or the cropped screenshot you selected. No account, no telemetry, no keystroke logging, no clipboard access. Submitting the answer is always your own click.
- Offline self-test: 15 checks pass (
node tools/selftest-web.js) — manifest validity, syntax of every file, subject detection for all ten subjects, choice/answer-letter parsing, bank fuzzy matching, AI response parsing, endpoint assembly. - The zip is verified with Python's
zipfile: CRC clean and byte-identical to the sources.
Not yet verified: real end-to-end behaviour on a live Xuexitong page (no account to log in with on this machine). After installing, please run through section 三 once and report any page where reading the question fails — that is the part most likely to need a selector fix.
源码开放,可自行审计与修改。用于学习用途;答题结果请自行核对后再提交。 The source is open for auditing and modification. Answer candidates from web search are unverified — always check them before submitting.