AI Outbound Engine

AI-driven file parsing engine with configurable rule engine and Next.js Web UI.

Features

• Zero-hardcoding: AI auto-generates parsing rules for any Excel/Word/PDF format
• 5 parsing modes: table, matrix_transpose, card_split, text_parse, multi_sheet
• PDF pre-parsing: Automatically extracts structured data from PDF delivery notes
• Web UI: Upload files, AI rule generation, rule management, data preview

Project Structure

├── ai-engine/          # Core parsing engine (Node.js/TypeScript)
│   ├── src/
│   │   ├── ai/         # DeepSeek AI integration
│   │   ├── engine/     # Rule engine and rule store
│   │   └── readers/    # Excel, Word, PDF readers
│   └── data/rules.json # Saved rules
└── web/                # Next.js Web UI (port 7001)
    └── src/app/        # Upload, rules, data preview pages

Getting Started

AI Engine

cd ai-engine
npm install
npx tsx src/test.ts all

Web UI

cd web
npm install
npm run dev

Open http://localhost:7001

Environment

Variable	Description	Default
DEEPSEEK_API_KEY	DeepSeek API key for AI rule generation	Required

Deployments

Environment	URL	Notes
Production	https://web-nu-nine-14.vercel.app	Vercel
Preview	https://web-aj88x18ie-houpes-projects.vercel.app	Latest build
GitHub	https://github.com/houpe/ai-outbound-engine	Source code

Environment Variables

• DEEPSEEK_API_KEY — configured on Vercel project

Test Templates

Template	Mode	Rows	Description
黎明屯	table	2	单订单发货单（Key-value尾部）
湖南仓	table	167	多订单汇总单（每行含收货信息）
多门店	multi_sheet	21	3个Sheet分门店出库
卡片式	card_split	9	调拨记录卡片堆叠
欢乐牧场	matrix_transpose	15	库存表转出库（门店矩阵）
PDF	table+col_index	39	PDF配送单预解析