AI Customer Service Knowledge Base Builder

Description

Help SMBs quickly build AI-powered customer service systems. Input FAQ documents or website URLs to automatically generate a knowledge base and configure auto-reply capabilities.

帮助中小企业快速搭建AI客服系统。输入FAQ文档或网站URL，自动生成知识库并配置自动回复功能。

Use When

• Setting up customer service automation
• Building FAQ knowledge bases
• Configuring auto-reply systems
• Migrating customer service to AI

Capabilities

• Extract FAQ from documents (PDF, TXT, MD, DOCX)
• Scrape FAQ from website URLs
• Generate structured knowledge base (JSON)
• Test Q&A matching
• Export to common formats (JSON, CSV, Markdown)

Usage

Basic Commands

# Extract from document
node kb-builder.js extract --file ./faq.pdf --output ./kb.json

# Extract from website
node kb-builder.js scrape --url https://example.com/faq --output ./kb.json

# Test knowledge base
node kb-builder.js test --kb ./kb.json --query "退货政策是什么？"

# Export to different formats
node kb-builder.js export --kb ./kb.json --format csv --output ./kb.csv

Configuration

Create config.json:

{
  "language": "zh-CN",
  "minConfidence": 0.7,
  "maxResults": 3,
  "fallbackMessage": "抱歉，我没有找到相关答案。请联系人工客服。"
}

Examples

Example 1: Build from PDF

node kb-builder.js extract --file ./company-faq.pdf --output ./kb.json
node kb-builder.js test --kb ./kb.json --query "如何退货？"

Example 2: Build from Website

node kb-builder.js scrape --url https://shop.example.com/help --output ./kb.json
node kb-builder.js export --kb ./kb.json --format markdown --output ./faq.md

Example 3: Interactive Mode

node kb-builder.js interactive --kb ./kb.json
# Then type questions to test responses

Output Format

Knowledge base JSON structure:

{
  "version": "1.0",
  "language": "zh-CN",
  "entries": [
    {
      "id": "q001",
      "question": "如何退货？",
      "answer": "您可以在收到商品后7天内申请退货...",
      "keywords": ["退货", "退款", "return"],
      "category": "售后服务"
    }
  ]
}

Requirements

• Node.js 18+
• No external API keys needed for basic features
• Optional: OpenAI API key for enhanced matching

Security

• All processing is local
• No data sent to external services (unless using OpenAI enhancement)
• Safe for sensitive business information

Limitations

• PDF extraction requires readable text (not scanned images)
• Website scraping respects robots.txt
• Best results with structured FAQ pages

Author

Created for OpenClaw by Claude

License

MIT

Tags

customer-service, ai, knowledge-base, faq, automation, chatbot, 客服, 知识库