studycn 1.1.0 → 1.2.0

package/AGENT.md

@@ -2,7 +2,7 @@
 
 使用同目录 `study.mjs`，Node.js 22+。读取 `README.md` 后操作；服务与凭证从 `STUDY_URL` / `STUDY_TOKEN` 或 `--token-file` 提供，不把凭证写入题目、日志或对话。
 
-仅操作明确授权题库。先 `whoami` / `banks`，用 JSON 创建 single/multiple/boolean/fill/short/coding 题目，题干和解析必填。填空用 blankAnswers 二维数组，题干按序 {{1}}、{{2}}；简答/编程须 referenceAnswer 与 rubric，options/correct 为空。主观题人工评分，编程不执行代码。作答附件仅由网页试卷创建者逐题开启，题库 Agent 不操作答卷和成绩。素材先 upload，用返回 id 引用。知识点与标签由需求明确或合理编写，不伪称系统自动生成。
+先whoami、banks、capabilities，只操作授权范围。六题型支持文字、图片、视频。题干content/mediaIds、标准答案referenceAnswer/answerMediaIds、解析explanation/explanationMediaIds、选项options[].mediaIds组合使用，素材先upload，用真实返回id引用，不保存临时URL。填空与选择题仍需可机读答案；主观题rubric必填并人工评阅，编程不执行代码。
 
 先本地 validate，再服务端 validate 或 import --dry-run；新建最多100道一批。写入采用固定 --request-id，每个独立操作新标识，同一操作重试复用。
 
@@ -10,4 +10,4 @@
 
 检查退出码与 JSON；0成功，1校验/参数错误，2服务/网络错误。业务命令不依赖自然语言解析。遇到网络错误按输出的 idempotencyKey 重试。写入后 get 核验。
 
-这些接口不支持账号配置、考试发布、答卷改动或成绩发布。不要将题库写入成功表述为考试或成绩已经发布。
+CLI 1.2的call覆盖各业务模块，参见OPERATIONS.md。账号角色、部门范围、对象所有权与凭证模块权限均需满足。只执行用户明确要求的业务动作；自动出题不包含发布考试、发布成绩、权限变更、发送催办。不要把题库写入成功表述为考试或成绩已发布。通用幂等重试仅返回completed/replayed，随后查询核对当前对象。

package/OPERATIONS.md

@@ -0,0 +1,36 @@
+# CLI 1.2 业务操作
+
+`study capabilities`返回模块权限表，`study whoami`返回当前有效授权。
+
+`study call --path PATH --method GET|POST|DELETE --file 参数.json --request-id 固定标识`调用与网页相同的业务操作。GET默认，可省略file和请求标识。路径不带`/api/`或前导斜线。
+
+| 模块 | 路径 | 操作权限 |
+|---|---|---|
+| 项目/部门 | projects / departments；删除 /ID | organization:read/write |
+| 人员 | people；people/move；people/delete | people:read/write |
+| 题库 | banks；删除 banks/ID | questions:read / banks:write |
+| 试卷 | papers；删除 /ID | papers:read/write |
+| 考试 | exams；exams/ID/publish、copy、void、finish、remind | exams:read/write |
+| 名单 | enrollments/考试ID | exams:read / enrollments:write |
+| 答卷 | attempts/答卷ID；/save、submit、terminate、attachments | 本人 self:read/write，管理 grades:read / exams:write |
+| 成绩 | grades；grades/ID/adjust、confirm、publish | grades:read/write |
+| 复核 | reviews | reviews:read/write；本人申请 self:write |
+| 导出 | export/results、analysis、grades、attempts、exams、enrollments | exports:read + grades:read |
+| 题库导出 | export/questions?bankId=ID | exports:read + questions:read |
+| 人员导出 | export/people | exports:read + people:read |
+| 模板 | templates/questions、people、enrollments | exports:read |
+| Excel | imports/preview；imports/commit | questions:write / people:write / enrollments:write，名单导入同时需people:write |
+| 本人 | personal；account；notices/ID | self:read/write |
+| 配置 | settings | settings:read/write |
+| 任务 | jobs；jobs/ai；jobs/ID | jobs:read/write |
+| 凭证 | agent-keys；删除 /ID | credentials:write，不得扩大当前凭证范围 |
+
+示例：`study call --path projects`；`study call --path departments --method POST --file department.json --request-id department-001`；`study call --path "export/results?examId=实际ID" --out results.xlsx`。
+
+固定试卷JSON：`{"name":"试卷","description":"","mode":"fixed","bankIds":["实际题库ID"],"questionIds":["实际题目ID"],"mandatoryIds":[],"quotas":[],"shuffleQuestions":false,"shuffleOptions":false,"allowCorrectSupplement":false,"wrongRatio":0}`。保存已有对象保留id和version；逐题附件通过questionSettings设置。
+
+Excel预览参数 `{"type":"questions","targetId":"实际题库ID"}`，`study call --path imports/preview --method POST --file import.json --upload questions.xlsx --request-id preview-001`。提交参数保留相同type、targetId和预览返回token，再调用imports/commit。
+
+通用请求返回`result`；文件下载用`--out`；媒体附件GET可返回签名URL。写入在请求级事务内执行，与幂等标识一起提交。重试返回completed与replayed，随后查询当前对象，不返还旧数据。逻辑删除保留历史答卷、成绩及日志。
+
+技能ZIP内附首次授权与持久化脚本。首次连接验证后在本人私密目录保存凭证，不调用钥匙串，不保存网页登录密码。豆包宿主需提供Node22+、终端工具、联网与持久化私密目录。

package/README.md

@@ -8,15 +8,15 @@ API：`https://study.aimianshi.online/api/agent/v1`
 
 题库管理 → 管理题目 → 新建/编辑。题干与解析使用可视化编辑器，支持标题、加粗、斜体、列表、引用、表格；可切换 Markdown。右侧实时预览正确答案、选项图片与解析。知识点和标签可手动填，也可由 Agent 在 JSON 中提供，系统不会自动猜测。
 
-选项“图片”按钮可上传或选择已有素材；支持纯图片或图文选项。题目素材可包含图片、音视频、PDF、TXT，单文件最大 40MB。JSON 中图片使用上传接口返回的 `id`，无需人工编写编号。图片不要写进 Markdown 外部链接；应使用选项 `imageId` 或题目 `mediaIds`。Excel 使用已上传完整文件名，继续兼容原导入方式。
+题干、标准答案、解析和每个选项均可组合文字、图片、视频及附件，单文件最大100MB。图片支持PNG/JPG/WebP/GIF，视频支持MP4/WebM/MOV，推荐浏览器通用的H.264/AAC MP4。JSON使用上传返回的编号：题干`mediaIds`、答案`answerMediaIds`、解析`explanationMediaIds`、选项`options[].mediaIds`，兼容图片`imageId`。Excel填写已上传完整文件名，对应“题目素材”“答案素材”“解析素材”“选项素材A–L”，多个用分号分隔。
 
 “Agent JSON”可粘贴结构化题目、校验并应用到编辑器，也可导出 JSON。网页、CLI、Agent API 使用同一 JSON 校验器。草稿、Markdown 模式与页面位置在当前浏览器刷新后保留；关闭弹窗也保留草稿，保存成功清除。退出账号时清除该账号浏览器草稿。
 
 ## 凭证与权限
 
-题库管理 → Agent 接入；超级管理员也可从系统管理 → Agent 接入进入。创建凭证时指定题库、有效期（30/90/365 天）、查询/题目写入/素材上传权限。查询权限包含答案与解析，只应授权题库维护人员。
+题库管理、系统管理或账号设置 → Agent 接入。有效期30/90/365天，按业务模块选择权限；题库操作须指定题库。题库查询包含标准答案与解析，只应授权维护人员。每人可授权本人操作；当前角色允许的操作才可勾选。
 
-完整凭证仅在创建时展示一次，服务器仅存 SHA-256 摘要；网页刷新或关闭后不能再次查看。可随时撤销。凭证继承所属账号当前权限，账号停用、失去题库权限、题库删除、凭证过期或撤销后不能继续操作。凭证不能访问账号管理、邮件/AI配置、考试发布、答卷或成绩发布接口。每次 Agent 写入均记录凭证名称、请求编号、题目版本和操作者。
+完整凭证仅展示一次，服务器仅存SHA-256摘要。可撤销，过期、停用或当前权限变化即时生效。CLI 1.2支持人员、项目部门、试卷考试、名单、答卷、评分复核、配置、后台任务和导出，各模块分别授权。账号密码与邮箱验证码在网页处理。操作仍受账号角色、部门范围、题库范围、对象所有权限制。所有Agent写入记录操作者、凭证名称与请求编号。
 
 HTTP 身份：`Authorization: Bearer <访问凭证>`。网页 Session 不适用于 Agent API。没有自动把 AI 服务配置或其密钥交给 Agent。
 
@@ -52,7 +52,7 @@ study whoami
 
 Mac/Linux 默认安装目录为 `~/.local/share/study`，命令入口为 `~/.local/bin/study`，环境文件为 `~/.local/share/study/env`。安装脚本向当前用户的 `.zshrc` 或 `.bashrc` 追加环境加载行；Bash 同时维护 `.bash_profile`。Windows 默认安装目录为 `%LOCALAPPDATA%\StudyCLI`，命令入口为该目录下的 `bin\study.cmd`，同时更新当前终端和用户 PATH。Windows ARM64/x64、Mac ARM64/x64、Linux ARM64/x64 由安装脚本识别。
 
-Mac/Linux 可用 `STUDY_INSTALL_DIR`、`STUDY_BIN_DIR` 自定义目录；`STUDY_NO_PROFILE=1` 禁止修改终端配置文件，此时自行加载指定安装目录的 `env`。所有安装与运行均不调用系统钥匙串。当前 CLI 版本为 1.1.0；Windows PowerShell 安装脚本已经静态检查，尚未完成 Windows 实机验证。
+Mac/Linux可用`STUDY_INSTALL_DIR`、`STUDY_BIN_DIR`自定义目录；`STUDY_NO_PROFILE=1`禁止修改终端配置，此时自行加载env。当前官网与Skill内置CLI版本1.2.0，不调用系统钥匙串。Windows脚本尚未实机验证。npm版本以实际发布记录为准；官网安装脚本始终下载本站完整包。
 
 已有 Node.js 22+ 也可使用 npm：
 
@@ -108,7 +108,7 @@ study resource --id '素材编号' --out option.png
 | status | `active` / `disabled`，默认 active |
 | id / version | 更新时必须保留读取到的编号与原版本；新建不传 |
 
-填空、简答、编程题的 `options` 和 `correct` 须省略或为空数组。附件不在题库层面开启：试卷创建者在网页试卷编辑中逐题配置 `questionSettings[questionId].allowAttachments`，默认关闭；固定与随机组卷均支持，发布时固定快照。每题最多5个作答附件，每个20MB，交卷后锁定。Agent 凭证仅负责题库，不拥有答卷、作答附件或成绩权限。
+填空、简答、编程题的options和correct为空。选择题正确选项与填空题可接受答案继续采用可机读格式；可追加答案媒体。主观参考答案可用文字或媒体，评分标准仍须文字。答题者附件由试卷questionSettings逐题开启，默认关闭；最多5个，图片视频各100MB，其他附件20MB。交卷后锁定；操作答卷及成绩须另行授权，评分遵循“改分→确认→再次发布”。
 
 答卷采用 `{selected:[], text:"", blanks:[], attachmentIds:[]}`，仍兼容旧选择题数组。简答、编程成绩带逐题待评分状态，全部人工评分后才可确认，再发布。AI 仅提供复核意见。Excel 模板最后三列为“填空答案”“参考答案”“评分标准”，示例覆盖六种题型。
 
@@ -140,14 +140,14 @@ JSON Schema用于形状、范围和必填检查；引用存在、两位小数、
 
 写入题目返回 `{items,count,requestId,replayed}`。POST/PUT 写入（含上传）必须有 `Idempotency-Key`，8–100 位字母、数字、下划线、点、冒号或短横线。相同凭证+标识+内容重试返回原结果、不重复入库，`replayed=true`；相同标识用于不同内容返回409。CLI自动生成标识，建议批次明确传 `--request-id` 以便跨进程重试。每个新操作使用新的标识。
 
-JSON 请求上限2MB；素材请求上限42MB、文件上限40MB。校验成功不代表写入时一定成功：并发修改、重复题、权限变化仍会被最终事务拦截。预览校验不查重；最终写入按题型、题干、选项与正确项检查题库内完全重复题。
+JSON请求2MB，素材文件100MB、请求102MB。校验不等于写入成功，并发版本、重复题及授权变化在最终事务拦截。图片与视频放在私有TOS的Study专用目录，文字在数据库。浏览器调用`/api/resources/ID/access?attemptId=答卷ID`取得单文件5分钟签名URL，随后直接访问TOS，过期重新授权。TOS密钥不发给浏览器。答案和解析媒体在成绩正式发布且到达个人解析开放时间后开放，猜测编号不能提前访问。旧素材兼容本地存储。
 
 401：缺少/失效凭证；403：权限不足；404：题目/素材不存在或不可访问；409：版本冲突、幂等标识冲突；400：格式或重复题；413：超大小；500：服务错误。401/403不能自动扩大权限，409须重读核对，禁止用新版本直接覆盖旧草稿。
 
 ## Agent 推荐步骤
 
 1. `whoami` 与 `banks` 确认范围。
-2. 根据需求生成六种受支持题型及解析、知识点；有图片先 upload，将返回 id 填入 JSON。
+2. 根据需求生成六种题型、答案、解析与知识点；图片视频先upload，用返回id填写对应媒体字段。
 3. `validate` 本地校验，`validate --remote` 或 `import --dry-run` 校验服务端权限与引用。
 4. 新建使用 create/import，固定 request-id；更新先 get，核对内容后保留 version 执行 update。
 5. 再次 get/questions 核验。考试或成绩发布仍由网页正常人工流程完成。

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "studycn",
-  "version": "1.1.0",
+  "version": "1.2.0",
   "description": "考试中心 Agent CLI",
   "homepage": "https://study.aimianshi.online/",
   "keywords": ["study", "exam", "cli", "agent", "question-bank"],
@@ -8,6 +8,6 @@
   "bin": {"study": "./study.mjs"},
   "engines": {"node": ">=22"},
   "publishConfig": {"registry": "https://registry.npmjs.org/", "access": "public"},
-  "files": ["study.mjs", "question-format.mjs", "question.schema.json", "questions.example.json", "README.md", "AGENT.md"],
+  "files": ["study.mjs", "question-format.mjs", "question.schema.json", "questions.example.json", "README.md", "AGENT.md", "OPERATIONS.md"],
   "license": "UNLICENSED"
 }

package/question-format.mjs

@@ -23,7 +23,7 @@ const numeric = (value, path) => {
 };
 export function normalizeQuestion(input) {
   if (!input || typeof input !== "object" || Array.isArray(input)) fail("$", "题目必须是 JSON 对象");
-  const allowed = ["id", "version", "bankId", "kind", "content", "options", "correct", "points", "explanation", "difficulty", "knowledge", "tags", "mediaIds", "status", "ownerId", "createdAt", "updatedAt", "blankAnswers", "referenceAnswer", "rubric"];
+  const allowed = ["id", "version", "bankId", "kind", "content", "options", "correct", "points", "explanation", "difficulty", "knowledge", "tags", "mediaIds", "answerMediaIds", "explanationMediaIds", "status", "ownerId", "createdAt", "updatedAt", "blankAnswers", "referenceAnswer", "rubric"];
   for (const key of Object.keys(input)) if (!allowed.includes(key)) fail(key, `不支持字段 ${key}`);
   const kind = input.kind;
   if (!["single", "multiple", "boolean", "fill", "short", "coding"].includes(kind)) fail("kind", "支持 single 单选、multiple 多选、boolean 判断、fill 填空、short 简答、coding 编程");
@@ -45,7 +45,7 @@ export function normalizeQuestion(input) {
     } else {
       result.referenceAnswer = string(input.referenceAnswer ?? "", "referenceAnswer");
       result.rubric = string(input.rubric ?? "", "rubric");
-      if (!result.referenceAnswer || !result.rubric) fail("referenceAnswer", "主观题须填写参考答案和评分标准，逐题人工评阅");
+      if ((!result.referenceAnswer && !result.answerMediaIds.length) || !result.rubric) fail("referenceAnswer", "主观题须提供文字或媒体参考答案，并填写评分标准，逐题人工评阅");
     }
     return result;
   }
@@ -53,18 +53,19 @@ export function normalizeQuestion(input) {
   if (!Array.isArray(rawOptions) || rawOptions.length < 2 || rawOptions.length > 12) fail("options", "请填写 2–12 个选项");
   const options = rawOptions.map((option, index) => {
     if (!option || typeof option !== "object" || Array.isArray(option)) fail(`options[${index}]`, "选项必须是对象");
-    for (const key of Object.keys(option)) if (!["id", "text", "imageId", "score"].includes(key)) fail(`options[${index}].${key}`, "不支持此选项字段");
+    for (const key of Object.keys(option)) if (!["id", "text", "imageId", "mediaIds", "score"].includes(key)) fail(`options[${index}].${key}`, "不支持此选项字段");
     const id = string(option.id ?? letters[index], `options[${index}].id`, 100);
     const text = string(option.text ?? "", `options[${index}].text`, 4000);
     const imageId = option.imageId ? string(option.imageId, `options[${index}].imageId`, 100) : undefined;
-    if (!id || (!text && !imageId)) fail(`options[${index}]`, "每个选项须有编号，以及文字或已上传图片");
-    return {id, text, ...(imageId ? {imageId} : {}), ...(option.score !== undefined ? {score: numeric(option.score, `options[${index}].score`)} : {})};
+    const mediaIds = list(option.mediaIds, `options[${index}].mediaIds`);
+    if (!id || (!text && !imageId && !mediaIds.length)) fail(`options[${index}]`, "每个选项须有编号，以及文字或已上传媒体");
+    return {id, text, ...(imageId ? {imageId} : {}), ...(mediaIds.length ? {mediaIds} : {}), ...(option.score !== undefined ? {score: numeric(option.score, `options[${index}].score`)} : {})};
   });
   if (new Set(options.map(o => o.id)).size !== options.length) fail("options", "选项编号不能重复");
   const correct = list(input.correct, "correct");
   if (correct.some(id => !options.some(o => o.id === id))) fail("correct", "正确答案必须引用选项 id");
   if (kind === "multiple" ? correct.length < 2 : correct.length !== 1) fail("correct", kind === "multiple" ? "多选至少两个正确项" : "单选和判断须选择一个正确项");
-  if (kind === "boolean" && (options.length !== 2 || options[0].text !== "正确" || options[1].text !== "错误" || options.some(o => o.imageId))) fail("options", "判断题固定两个选项：正确、错误");
+  if (kind === "boolean" && (options.length !== 2 || options[0].text !== "正确" || options[1].text !== "错误")) fail("options", "判断题固定两个选项：正确、错误");
   const explicitScores = options.filter(o => correct.includes(o.id) && o.score !== undefined);
   if (explicitScores.length && explicitScores.length !== correct.length) fail("options.score", "所有正确项须一起填分值，或全部不填以自动均分");
   const scores = equalScores(points, correct.length);
@@ -74,13 +75,14 @@ export function normalizeQuestion(input) {
   return commonQuestion(input, {kind, points, options: scoredOptions, correct});
 }
 function commonQuestion(input, base) {
-  const content = string(input.content, "content"), explanation = string(input.explanation, "explanation");
-  if (!content || !explanation) fail(!content ? "content" : "explanation", "题干和解析必填");
+  const content = string(input.content ?? "", "content"), explanation = string(input.explanation ?? "", "explanation");
+  const mediaIds = list(input.mediaIds, "mediaIds"), answerMediaIds = list(input.answerMediaIds, "answerMediaIds"), explanationMediaIds = list(input.explanationMediaIds, "explanationMediaIds");
+  if ((!content && !mediaIds.length) || (!explanation && !explanationMediaIds.length)) fail(!content && !mediaIds.length ? "content" : "explanation", "题干和解析各需提供文字或媒体");
   const difficulty = input.difficulty ?? "medium", status = input.status ?? "active";
   if (!["easy", "medium", "hard"].includes(difficulty)) fail("difficulty", "难度仅支持 easy / medium / hard");
   if (!["active", "disabled"].includes(status)) fail("status", "状态仅支持 active / disabled");
   const result = {bankId: string(input.bankId, "bankId", 100), ...base, content, explanation, difficulty,
-    knowledge: list(input.knowledge, "knowledge"), tags: list(input.tags, "tags"), mediaIds: list(input.mediaIds, "mediaIds"), status};
+    knowledge: list(input.knowledge, "knowledge"), tags: list(input.tags, "tags"), mediaIds, answerMediaIds, explanationMediaIds, status};
   if (!result.bankId) fail("bankId", "请选择所属题库");
   if (input.id !== undefined) {
     result.id = string(input.id, "id", 100);

package/question.schema.json

@@ -4,13 +4,16 @@
   "title": "考试中心题目 v1",
   "type": "object",
   "additionalProperties": false,
-  "required": ["bankId", "kind", "content", "explanation"],
+  "required": ["bankId", "kind"],
+  "$defs": {
+    "media": {"type": "array", "maxItems": 100, "uniqueItems": true, "items": {"type": "string", "minLength": 1, "maxLength": 200}}
+  },
   "properties": {
     "id": {"type": "string", "minLength": 1, "maxLength": 100},
     "version": {"type": "integer", "minimum": 1},
     "bankId": {"type": "string", "minLength": 1, "maxLength": 100},
     "kind": {"enum": ["single", "multiple", "boolean", "fill", "short", "coding"]},
-    "content": {"type": "string", "minLength": 1, "maxLength": 20000, "description": "题干，支持 Markdown、GFM 表格"},
+    "content": {"type": "string", "maxLength": 20000, "description": "题干，支持 Markdown、GFM 表格；文字或 mediaIds 至少一项"},
     "options": {
       "type": "array", "maxItems": 12,
       "items": {
@@ -19,20 +22,23 @@
           "id": {"type": "string", "minLength": 1, "maxLength": 100, "description": "省略时按顺序生成 A–L"},
           "text": {"type": "string", "maxLength": 4000},
           "imageId": {"type": "string", "maxLength": 100, "description": "上传接口返回的 id；仅图片"},
+          "mediaIds": {"$ref": "#/$defs/media", "description": "选项图片、视频及附件，使用上传后的编号"},
           "score": {"type": "number", "minimum": 0, "multipleOf": 0.01, "description": "正确项一起填写或一起省略自动均分；错误项须为 0"}
         }
       }
     },
     "correct": {"type": "array", "maxItems": 12, "uniqueItems": true, "items": {"type": "string", "minLength": 1, "maxLength": 200}, "description": "选择题引用 options.id；填空和主观题为空数组或省略"},
     "blankAnswers": {"type": "array", "minItems": 1, "maxItems": 20, "items": {"type": "array", "minItems": 1, "maxItems": 20, "uniqueItems": true, "items": {"type": "string", "minLength": 1, "maxLength": 2000}}, "description": "填空逐空可接受答案，与题干 {{1}}、{{2}} 顺序一致"},
-    "referenceAnswer": {"type": "string", "minLength": 1, "maxLength": 20000},
+    "referenceAnswer": {"type": "string", "maxLength": 20000},
     "rubric": {"type": "string", "minLength": 1, "maxLength": 20000, "description": "简答和编程题人工评分标准"},
     "points": {"type": "number", "exclusiveMinimum": 0, "maximum": 10000, "multipleOf": 0.01, "default": 10},
-    "explanation": {"type": "string", "minLength": 1, "maxLength": 20000},
+    "explanation": {"type": "string", "maxLength": 20000},
     "difficulty": {"enum": ["easy", "medium", "hard"], "default": "medium"},
     "knowledge": {"type": "array", "maxItems": 100, "uniqueItems": true, "items": {"type": "string", "minLength": 1, "maxLength": 200}, "default": []},
     "tags": {"type": "array", "maxItems": 100, "uniqueItems": true, "items": {"type": "string", "minLength": 1, "maxLength": 200}, "default": []},
     "mediaIds": {"type": "array", "maxItems": 100, "uniqueItems": true, "items": {"type": "string", "minLength": 1, "maxLength": 200}, "default": []},
+    "answerMediaIds": {"$ref": "#/$defs/media", "description": "标准答案媒体，仅答案开放后向本人展示"},
+    "explanationMediaIds": {"$ref": "#/$defs/media", "description": "解析媒体，仅答案开放后向本人展示"},
     "status": {"enum": ["active", "disabled"], "default": "active"},
     "ownerId": {"type": "string", "description": "get 返回的只读字段，输入时忽略"},
     "createdAt": {"type": "string", "description": "只读"},
@@ -40,13 +46,15 @@
   },
   "dependentRequired": {"id": ["version"], "version": ["id"]},
   "allOf": [
+    {"anyOf": [{"required": ["content"], "properties": {"content": {"minLength": 1}}}, {"required": ["mediaIds"], "properties": {"mediaIds": {"minItems": 1}}}]},
+    {"anyOf": [{"required": ["explanation"], "properties": {"explanation": {"minLength": 1}}}, {"required": ["explanationMediaIds"], "properties": {"explanationMediaIds": {"minItems": 1}}}]},
     {"if": {"properties": {"kind": {"enum": ["single", "multiple"]}}}, "then": {"required": ["options"], "properties": {"options": {"minItems": 2}}}},
     {"if": {"properties": {"kind": {"enum": ["single", "multiple", "boolean"]}}}, "then": {"required": ["correct"], "properties": {"correct": {"minItems": 1}}}},
     {"if": {"properties": {"kind": {"const": "multiple"}}}, "then": {"properties": {"correct": {"minItems": 2}}}},
     {"if": {"properties": {"kind": {"enum": ["single", "boolean"]}}}, "then": {"properties": {"correct": {"maxItems": 1}}}},
     {"if": {"properties": {"kind": {"const": "boolean"}}}, "then": {"properties": {"options": {"minItems": 2, "maxItems": 2}}}},
     {"if": {"properties": {"kind": {"const": "fill"}}}, "then": {"required": ["blankAnswers"]}},
-    {"if": {"properties": {"kind": {"enum": ["short", "coding"]}}}, "then": {"required": ["referenceAnswer", "rubric"]}},
+    {"if": {"properties": {"kind": {"enum": ["short", "coding"]}}}, "then": {"required": ["rubric"], "anyOf": [{"required": ["referenceAnswer"], "properties": {"referenceAnswer": {"minLength": 1}}}, {"required": ["answerMediaIds"], "properties": {"answerMediaIds": {"minItems": 1}}}]}},
     {"if": {"properties": {"kind": {"enum": ["fill", "short", "coding"]}}}, "then": {"properties": {"options": {"maxItems": 0}, "correct": {"maxItems": 0}}}}
   ]
 }

package/study.mjs

@@ -29,6 +29,13 @@ const help = `考试中心 CLI（Node.js 22+）
   resources                         列出可访问素材
   upload --file FILE [--request-id KEY]
   resource --id ID --out FILE        下载素材
+  access --id ID                    获取素材短时访问链接
+  delete --id ID --request-id KEY    删除题目（保留历史答卷）
+  capabilities                      查看业务操作与权限说明
+  call --path PATH [--method GET|POST|DELETE] [--file JSON]
+       [--upload FILE] [--out FILE] [--request-id KEY]
+                                    操作项目、部门、人员、试卷、考试、名单、
+                                    答卷、成绩、复核、配置、凭证、任务与导出
   schema [--out FILE]                获取本地 JSON Schema
 
 通用参数：--url URL、--token-file FILE、--help、--json。
@@ -38,7 +45,7 @@ const help = `考试中心 CLI（Node.js 22+）
 `;
 const output = value => process.stdout.write(JSON.stringify(value, null, 2) + "\n");
 const error = (message, code = "CLI_ERROR", exit = 1) => {const e = new Error(message); e.code = code; e.exit = exit; throw e;};
-const valueOptions = new Set(["url", "token-file", "request-id", "file", "id", "out", "bank", "kind", "status", "knowledge", "q", "limit", "offset"]);
+const valueOptions = new Set(["url", "token-file", "request-id", "file", "id", "out", "bank", "kind", "status", "knowledge", "q", "limit", "offset", "path", "method", "upload"]);
 const flags = new Set(["dry-run", "remote", "help", "json"]);
 function parse(args) {
   let command; const options = {};
@@ -66,13 +73,13 @@ async function readJson(filename) {
 const requireId = value => {if (!value || typeof value !== "string") error("请使用 --id 指定编号");return encodeURIComponent(value);};
 async function run() {
   const {command, options} = parse(process.argv.slice(2));
-  if (command === "version") {output({name: "考试中心 CLI", version: "1.1.0", node: process.versions.node});return;}
-  if (command === "help" || options.help) {if(options.json) output({name: "考试中心 CLI", version: "1.1.0", help}); else process.stdout.write(help);return;}
+  if (command === "version") {output({name: "考试中心 CLI", version: "1.2.0", node: process.versions.node});return;}
+  if (command === "help" || options.help) {if(options.json) output({name: "考试中心 CLI", version: "1.2.0", help}); else process.stdout.write(help);return;}
   if (command === "schema") {
     const value = JSON.parse(await fs.readFile(new URL("./question.schema.json", import.meta.url), "utf8"));
     if (options.out) {await fs.writeFile(options.out, JSON.stringify(value, null, 2) + "\n");output({ok: true, file: path.resolve(options.out)});}else output(value);return;
   }
-  if (!["whoami", "banks", "questions", "get", "validate", "create", "import", "update", "resources", "upload", "resource"].includes(command)) error("未知命令：" + command);
+  if (!["whoami", "banks", "questions", "get", "validate", "create", "import", "update", "resources", "upload", "resource", "access", "delete", "call", "capabilities"].includes(command)) error("未知命令：" + command);
   if (options["dry-run"] && !["create", "import", "update"].includes(command)) error("--dry-run 仅支持 create / import / update；本命令未执行");
   if (options.remote && command !== "validate") error("--remote 仅支持 validate；本命令未执行");
   let items;
@@ -99,13 +106,32 @@ async function run() {
     const headers = {"Authorization": "Bearer " + token};
     if (body !== undefined) {headers["Idempotency-Key"] = requestId;if (!(body instanceof FormData)) headers["Content-Type"] = "application/json";}
     let response;
-    try {response = await fetch(new URL("/api/agent/v1/" + endpoint, base), {method, headers, body: body === undefined ? undefined : body instanceof FormData ? body : JSON.stringify(body), redirect: "error", signal: AbortSignal.timeout(60000)});}
+    try {response = await fetch(new URL("/api/agent/v1/" + endpoint, base), {method, headers, body: body === undefined ? undefined : body instanceof FormData ? body : JSON.stringify(body), redirect: "error", signal: AbortSignal.timeout(180000)});}
     catch {error("网络请求未完成；写入重试请复用 --request-id " + requestId, "NETWORK_ERROR", 2);}
     if (binary && response.ok) return new Uint8Array(await response.arrayBuffer());
     let value;try {value = await response.json();}catch {error("服务未返回有效 JSON（HTTP " + response.status + "）", "INVALID_RESPONSE", 2);}
     if (!response.ok) {const e = new Error(value.error?.message || value.error || "请求失败");e.code = value.error?.code || "HTTP_ERROR";e.status = response.status;e.requestId = value.error?.requestId;e.idempotencyKey = requestId;e.exit = 2;throw e;}
     return value;
   }
+  if (command === "capabilities") return output(await request("capabilities"));
+  if (command === "call" || command === "delete") {
+    const method = command === "delete" ? "DELETE" : (options.method || "GET").toUpperCase();
+    const endpoint = command === "delete" ? "questions/" + requireId(options.id) : options.path;
+    if (!endpoint || !["GET", "POST", "DELETE"].includes(method)) error("call 须指定 --path 和有效 --method");
+    let body = options.file ? await readJson(options.file) : undefined;
+    let payload = {method, path: endpoint, ...(body ? {body} : {})};
+    if (options.upload) {
+      const stat = await fs.stat(options.upload);if(!stat.size || stat.size > 100 * 1024 * 1024) error("上传文件须为 1 字节至 100MB");
+      const form = new FormData();form.set("operation", JSON.stringify(payload));form.set("file", new File([await fs.readFile(options.upload)], path.basename(options.upload)));payload = form;
+    }
+    const response = await request("operations", "POST", payload);
+    if (options.out && response.result?.binary) {
+      const bytes = Buffer.from(response.result.binary, "base64");await fs.writeFile(options.out, bytes);return output({ok: true, file: path.resolve(options.out), size: bytes.length});
+    }
+    if (response.result?.binary && !options.out) error("下载或导出请指定 --out；操作已返回结果", "OUTPUT_REQUIRED");
+    if (options.out) {await fs.writeFile(options.out, JSON.stringify(response.result ?? response, null, 2) + "\n");return output({ok: true, file: path.resolve(options.out)});}
+    return output(response);
+  }
   if (["whoami", "banks", "resources"].includes(command)) return output(await request(command === "whoami" ? "me" : command));
   if (command === "questions") {
     const params = new URLSearchParams();
@@ -126,13 +152,23 @@ async function run() {
   if (command === "update") return output(await request("questions/" + requireId(items[0].id), "PUT", items[0]));
   if (command === "upload") {
     if(!options.file || options.file === "-") error("上传请指定真实文件");
-    const stat = await fs.stat(options.file);if(!stat.size || stat.size > 40 * 1024 * 1024) error("上传文件须为 1 字节至 40MB");
+    const stat = await fs.stat(options.file);if(!stat.size || stat.size > 100 * 1024 * 1024) error("上传文件须为 1 字节至 100MB");
     const form = new FormData();form.set("file", new File([await fs.readFile(options.file)], path.basename(options.file)));
     return output({...await request("resources", "POST", form), requestId});
   }
+  if (command === "access") return output(await request("resources/" + requireId(options.id) + "/access"));
   if (command === "resource") {
     if(!options.out) error("下载请使用 --out 指定保存路径");
-    const bytes = await request("resources/" + requireId(options.id), "GET", undefined, true);
+    const access = await request("resources/" + requireId(options.id) + "/access");
+    let bytes;
+    if (access.url.startsWith("/")) bytes = await request("resources/" + requireId(options.id), "GET", undefined, true);
+    else {
+      const url = new URL(access.url);
+      if (url.protocol !== "https:" || url.username || url.password || !url.hostname.endsWith(".volces.com")) error("媒体访问地址不符合预期", "INVALID_MEDIA_URL", 2);
+      const response = await fetch(url, {redirect: "error", signal: AbortSignal.timeout(180000)});
+      if (!response.ok) error("媒体下载失败，请重新获取授权链接", "MEDIA_ERROR", 2);
+      bytes = new Uint8Array(await response.arrayBuffer());
+    }
     await fs.writeFile(options.out, bytes);output({ok: true, file: path.resolve(options.out), size: bytes.length});return;
   }
 }