帮助 0 基础开发者从创意到参赛的全流程辅助工具
A full-process assistant tool that helps zero-experience developers go from idea to competition.
帮助 0 基础开发者从创意到参赛的全流程辅助工具。覆盖选题、选型、开发、部署、演示五大阶段，让每一个有创意的人都能创造出自己的项目并成功参赛。
- 集成 MyMemory 多语种翻译 API，自动将中文描述翻译为英文
- 6 渠道并行搜索：GitHub / Devpost / Bing / DuckDuckGo / Wikipedia / ProductHunt
- GitHub：通过 API 搜索代码仓库（多查询回退策略）
- Devpost：通过 r.jina.ai 渲染 JS 后抓取黑客松项目
- Bing：通过 r.jina.ai 渲染 JS 后解析搜索引擎结果（中英双查询）
- DuckDuckGo：通过 Instant Answer API 获取快速答案
- Wikipedia：通过 CORS 友好的 API 搜索百科条目
- ProductHunt：通过 r.jina.ai 渲染 JS 后抓取同类产品
- 自动过滤 Bing 搜索结果中的广告（Sponsored）
- 基于搜索结果数量和命中百分比计算稀缺度评分
- 从原创性、稀缺度、意义感三个维度量化项目价值
- 25+ 常见项目模式查重库，自动检测"换皮"项目
- 匹配时推荐差异化策略，帮助红海项目找到蓝海切入点
- 4 套预设方案：完全免费(¥0) / 经济(¥30-80) / 专业(¥150-400) / 高配(¥400-1000)
- 一键自动推荐技术栈，展示逐项成本分析
- 支持团队规模和开发时长配置
- 自动生成分工建议
- 支持整个项目文件夹递归上传
- 自动过滤
node_modules、.git等无关目录 - 按目录结构分组展示文件列表
- 扫描硬编码密钥（API Key、Token、密码等）
- 检测敏感文件（.env、私钥文件等）
- 检测代码质量问题（debugger、console.log 等）
- 检测
.gitignore配置完整性 - 实时输出安全评分
- 6 步 Git 版本控制教程（从安装到推送 GitHub，每步含可复制命令）
- 自动检测项目类型（React / Vue / Next.js / Python / 静态站点等）
- 推荐最佳部署方案（Vercel / Netlify / Render / Railway）
- 附部署检查清单
- 根据选手输入的项目名称、描述和技术栈，自动解析目标用户和核心功能
- 生成 7 段个性化 Pitch 演讲稿（开场Hook → 问题 → 方案 → Demo → 技术 → 影响力 → 展望）
- 5 个 AI 评审员从 5 个维度自动评分：
- 代码质量评审员（结构、可读性、错误处理、注释、硬编码、依赖、测试、版本控制）
- 用户体验评审员（首屏、导航、反馈、响应式、一致性、无障碍、空状态、性能）
- 创新性评审员（新颖性、差异化、AI整合、问题匹配、可扩展、技术组合、用户洞察、市场潜力）
- 技术深度评审员（复杂度、完整性、架构、API设计、数据处理、安全、性能、部署）
- 演示与表达评审员（价值主张、演示流程、问题陈述、方案清晰度、视觉辅助、技术解释、未来规划、QA准备）
- 40 项标准自动评分并给出针对性反馈
- 雷达图可视化展示
- 生成 P0-P3 优先级改进清单
- 支持用户手动调整 AI 评分
配置 API Key 后，选题评审升级为 AI 语义分析；不配置则自动降级为纯本地逻辑，功能不缺失。
- 输入理解：AI 翻译并提取搜索关键词（替代硬编码词表）
- 语义匹配：AI 判定搜索结果相关度与查重模式命中（替代字符串匹配）
- 建议生成：评分建议、定制差异化策略、蓝海方向推荐
- 功能级对比矩阵：把最相关的相似项目当作竞品，逐项对比功能重合度与市场成熟度，输出「成熟实现（勿重复造轮子）/ 你的差异点（获奖关键）/ 竞品独点（可借鉴）/ 按优先级的优化建议」
配置项（.env，参考 .env.example）：
| 变量 | 必填 | 说明 |
|---|---|---|
HACKCHECK_API_KEY |
是 | OpenAI 兼容 API Key |
AI_BASE_URL |
否 | 默认 https://api.openai-next.com/v1 |
AI_MODEL |
否 | 默认 deepseek-v4-flash（低费率） |
线上部署（Vercel）：在平台环境变量中配置同名变量，api/ai.js 自动生效。
纯静态托管：页面 ⚙ 按钮可填自备 Key（仅存本机浏览器）。
- 前端：HTML5 + CSS3 + 原生 JavaScript（无框架依赖）
- 外部 API：
- MyMemory Translation API - 多语种翻译
- GitHub Search API - 仓库搜索
- Wikipedia API - 百科搜索
- DuckDuckGo Instant Answer API - 快速答案
- r.jina.ai - JS渲染代理（用于 Bing/Devpost/ProductHunt 搜索）
- corsproxy.io - CORS 代理（备用）
- 数据存储：LocalStorage（本地持久化，无需后端）
- 部署：纯静态文件，可部署到任何静态托管平台
环境要求：仅需 Python 3.8+（用于本地服务器），无需安装任何第三方依赖。
# 克隆仓库
git clone https://github.com/DrenLea/hackcheck.git
cd hackcheck
# 方式一（推荐，启用 AI 增强）：配置密钥后用内置服务器启动
cp .env.example .env # 编辑 .env 填入你的 API Key
python tools/ai_proxy.py # 默认 8080 端口，可用 python tools/ai_proxy.py 9000 换端口
# 访问 http://localhost:8080
# 方式二（无 AI，纯静态）：
python -m http.server 8080
# 方式三（线上部署）：推送到 Vercel，在平台环境变量中配置 HACKCHECK_API_KEY无需安装任何第三方依赖；AI 增强为可选项，不配置密钥也可完整使用全部功能（自动降级为纯本地逻辑）。
# 后端代理测试（16 个用例，需 pytest）
python -m pytest tools/test_ai_proxy.py -q
# 前端 AI 层测试：启动本地服务器后浏览器打开
# http://localhost:8080/tests/ai.test.html （预期 ALL PASS）hackcheck/
├── index.html # 主页面
├── css/
│ └── styles.css # 全局样式
├── js/
│ ├── core.js # 全局状态、导航、总分、导出、初始化
│ ├── topic.js # 阶段1：选题评审
│ ├── tech.js # 阶段2：技术选型
│ ├── dev.js # 阶段3：代码扫描
│ ├── demo.js # 阶段4：Demo辅助
│ ├── pitch.js # 阶段5：Pitch生成与评审
│ ├── ai.js # AI 客户端层（可选增强，自动降级）
│ ├── i18n.js # 中英双语文案
│ └── data.js # 配置数据（技术栈、评审标准、查重库等）
├── api/
│ └── ai.js # Vercel Serverless AI 代理
├── tools/
│ ├── ai_proxy.py # 本地开发服务器（静态文件 + AI 代理）
│ └── test_ai_proxy.py # 代理测试（pytest）
├── tests/
│ └── ai.test.html # 前端 AI 层浏览器测试
├── .env.example # AI 配置示例
├── CHANGELOG.md # 变更日志
└── README.md
- 选题阶段：在首页输入项目描述，点击"搜索相似项目"，系统自动翻译并进行 6 渠道并行搜索，返回稀缺度评分、同类项目列表；配置 AI 后额外输出功能级对比矩阵与优化建议
- 选型阶段：进入"技术选型"模块，选择预设方案，获取技术栈推荐和成本分析
- 开发阶段：进入"代码扫描"模块，上传整个项目文件夹，获取安全评分和改进建议
- 部署阶段：进入"Demo辅助"模块，按 Git 教程完成版本控制，检测项目类型并选择部署方案
- 演示阶段：进入"Pitch生成"模块，输入项目信息生成演讲稿，点击"AI模拟评审"获取5个维度的自动评分和反馈
- 导出报告：点击"导出报告"按钮，下载完整的参赛自查报告
- 全流程陪伴：不是只在最后检查，而是从选题阶段就介入
- 0 基础友好：每个模块都有详细指引，Git 教程包含可复制命令
- 避坑导向：安全扫描检测硬编码密钥、.gitignore 配置等常见陷阱
- 数据驱动：稀缺度评分基于 GitHub 搜索结果，AI 评审基于项目实际信息
- 开箱即用：纯前端实现，无需注册登录，无需后端，打开即用
A full-process assistant tool that helps zero-experience developers go from idea to competition. Covers five stages: topic selection, tech stack selection, development, deployment, and demo — empowering anyone with an idea to build their own project and compete successfully.
- Integrated MyMemory multilingual translation API — auto-translates Chinese descriptions to English
- 6-channel parallel search: GitHub / Devpost / Bing / DuckDuckGo / Wikipedia / ProductHunt
- GitHub: API-based repository search (multi-query fallback strategy)
- Devpost: Scrapes hackathon projects via r.jina.ai (JS rendering)
- Bing: Parses search engine results via r.jina.ai (JS rendering, dual zh/en queries)
- DuckDuckGo: Instant Answer API for quick answers
- Wikipedia: CORS-friendly API for encyclopedia articles
- ProductHunt: Scrapes similar products via r.jina.ai (JS rendering)
- Auto-filters sponsored (ad) results from Bing
- Scarcity score calculated from search result count and hit percentage
- Quantifies project value across three dimensions: originality, scarcity, and social impact
- 25+ common project pattern library — auto-detects "reskin" projects
- Recommends differentiation strategies to help red-ocean projects find blue-ocean angles
- 4 preset plans: Free (¥0) / Budget (¥30-80) / Pro (¥150-400) / Premium (¥400-1000)
- One-click tech stack recommendation with itemized cost breakdown
- Configurable team size and development duration
- Auto-generated task assignment suggestions
- Recursive folder upload support
- Auto-filters
node_modules,.git, and other irrelevant directories - File list grouped by directory structure
- Scans for hardcoded secrets (API keys, tokens, passwords, etc.)
- Detects sensitive files (.env, private key files, etc.)
- Detects code quality issues (debugger, console.log, etc.)
- Checks
.gitignoreconfiguration completeness - Real-time security score output
- 6-step Git version control tutorial (from installation to pushing to GitHub, with copyable commands)
- Auto-detects project type (React / Vue / Next.js / Python / static site, etc.)
- Recommends best deployment platform (Vercel / Netlify / Render / Railway)
- Includes deployment checklist
- Auto-parses target users and core features from project name, description, and tech stack
- Generates 7-section personalized pitch script (Hook → Problem → Solution → Demo → Tech → Impact → Future)
- 5 AI reviewers auto-score across 5 dimensions:
- Code Quality Reviewer (structure, readability, error handling, comments, hardcoding, dependencies, testing, version control)
- UX Reviewer (first impression, navigation, feedback, responsiveness, consistency, accessibility, empty states, performance)
- Innovation Reviewer (novelty, differentiation, AI integration, problem fitting, scalability, tech combination, user insight, market potential)
- Tech Depth Reviewer (complexity, completeness, architecture, API design, data handling, security, performance, deployment)
- Presentation Reviewer (value proposition, demo flow, problem statement, solution clarity, visual aids, tech explanation, future plans, Q&A readiness)
- 40 criteria auto-scored with targeted feedback
- Radar chart visualization
- P0-P3 priority improvement list
- Users can manually adjust AI scores
With an API key configured, topic review upgrades to AI semantic analysis; without one, it automatically falls back to pure local logic with no loss of functionality.
- Input Understanding: AI translates and extracts search keywords (replaces hardcoded vocabulary)
- Semantic Matching: AI judges search result relevance and duplicate-pattern hits (replaces string matching)
- Suggestion Generation: score advice, tailored differentiation strategies, blue-ocean direction recommendations
- Feature Comparison Matrix: treats the most relevant similar projects as competitors, compares feature overlap and market maturity item by item, and outputs "mature implementations (don't reinvent) / your differentiators (award key) / competitor strengths (worth borrowing) / prioritized recommendations"
Configuration (.env, see .env.example):
| Variable | Required | Description |
|---|---|---|
HACKCHECK_API_KEY |
Yes | OpenAI-compatible API key |
AI_BASE_URL |
No | Defaults to https://api.openai-next.com/v1 |
AI_MODEL |
No | Defaults to deepseek-v4-flash (low cost) |
Online deployment (Vercel): set the same environment variables on the platform — api/ai.js picks them up automatically.
Pure static hosting: use the ⚙ button on the page to enter your own key (stored only in your local browser).
- Frontend: HTML5 + CSS3 + Vanilla JavaScript (no framework dependencies)
- External APIs:
- MyMemory Translation API — Multilingual translation
- GitHub Search API — Repository search
- Wikipedia API — Encyclopedia search
- DuckDuckGo Instant Answer API — Quick answers
- r.jina.ai — JS rendering proxy (for Bing/Devpost/ProductHunt search)
- corsproxy.io — CORS proxy (fallback)
- Data Storage: LocalStorage (client-side persistence, no backend needed)
- Deployment: Pure static files, deployable to any static hosting platform
Requirements: Python 3.8+ only (for the local server) — no third-party dependencies.
# Clone the repository
git clone https://github.com/DrenLea/hackcheck.git
cd hackcheck
# Option 1 (recommended, enables AI enhancement): configure key, then use the built-in server
cp .env.example .env # Edit .env and fill in your API key
python tools/ai_proxy.py # Port 8080 by default; use `python tools/ai_proxy.py 9000` to change
# Visit http://localhost:8080
# Option 2 (no AI, pure static):
python -m http.server 8080
# Option 3 (online): deploy to Vercel and set HACKCHECK_API_KEY in platform env varsNo third-party dependencies to install; AI enhancement is optional — all features work fully without a key (auto-fallback to pure local logic).
# Backend proxy tests (16 cases, requires pytest)
python -m pytest tools/test_ai_proxy.py -q
# Frontend AI-layer tests: start the local server, then open in a browser
# http://localhost:8080/tests/ai.test.html (expect ALL PASS)hackcheck/
├── index.html # Main page
├── css/
│ └── styles.css # Global styles
├── js/
│ ├── core.js # Global state, navigation, overall score, export, init
│ ├── topic.js # Stage 1: Topic review
│ ├── tech.js # Stage 2: Tech selection
│ ├── dev.js # Stage 3: Code scan
│ ├── demo.js # Stage 4: Demo assistance
│ ├── pitch.js # Stage 5: Pitch generation & review
│ ├── ai.js # AI client layer (optional enhancement, auto-fallback)
│ ├── i18n.js # Bilingual (zh/en) strings
│ └── data.js # Configuration data (tech stack, review criteria, pattern library, etc.)
├── api/
│ └── ai.js # Vercel Serverless AI proxy
├── tools/
│ ├── ai_proxy.py # Local dev server (static files + AI proxy)
│ └── test_ai_proxy.py # Proxy tests (pytest)
├── tests/
│ └── ai.test.html # Browser tests for the frontend AI layer
├── .env.example # AI configuration example
├── CHANGELOG.md # Changelog
└── README.md
- Topic Selection: Enter your project description on the home page, click "Search Similar Projects" — the system auto-translates and runs a 6-channel parallel search, returning a scarcity score and similar project list; with AI configured, it additionally outputs a feature comparison matrix and optimization advice
- Tech Selection: Go to the "Tech Selection" module, choose a preset plan, and get tech stack recommendations with cost analysis
- Development: Go to the "Code Scan" module, upload your entire project folder, and get a security score with improvement suggestions
- Deployment: Go to the "Demo Assistant" module, follow the Git tutorial for version control, detect project type, and choose a deployment plan
- Presentation: Go to the "Pitch Generation" module, enter project info to generate a speech script, click "AI Review" to get auto-scoring and feedback across 5 dimensions
- Export Report: Click "Export Report" to download a complete competition self-check report
- Full-Process Companion: Not just a final check — intervenes from the topic selection stage
- Beginner-Friendly: Every module has detailed guidance; Git tutorial includes copyable commands
- Pitfall-Oriented: Security scanner detects hardcoded secrets, .gitignore config, and other common traps
- Data-Driven: Scarcity score based on GitHub search results; AI review based on actual project info
- Out-of-the-Box: Pure frontend implementation — no registration, no backend, just open and use
MIT