Echo Assistant

中文 README

AI-native Enterprise Workflow Assistant

Echo Assistant is an AI-native workflow assistant designed for enterprise operations, customer support automation, knowledge-driven task execution, and natural language driven business workflows.

Echo Assistant helps teams automate customer support and operational workflows using AI.

Core Features

AI Customer Support

Answers customer and internal support questions from an enterprise knowledge base, with source-aware responses and workflow-specific suggestions.

Natural Language Workflow Rules

Turns plain-language operating policies into executable workflow logic, so business teams can configure rules without writing code.

Workflow Automation

Connects knowledge retrieval, AI reasoning, validation, and tool calls into repeatable workflows for operations teams.

Demo Scenario

The current demo uses e-commerce operations as one example scenario:

• Customer support QA for refund and after-sales policies
• Natural language recommendation and exposure rules
• Product intake, enrichment, and recommendation result preview

The product positioning is broader than this demo. Echo Assistant is intended to support enterprise workflows across support, operations, internal knowledge, approval flows, and other process-heavy teams.

Architecture

User
  ↓
React Frontend
  ↓
Node.js API
  ↓
AI Orchestration / RAG
  ↓
Workflow Engine
  ↓
Business Tools & Data

Screenshots

Customer Support QA

Natural Language Workflow Rule

Recommendation Result Preview

Tech Stack

• Frontend: React + Vite + Ant Design + TypeScript
• Backend: Express + TypeScript + LangChain + LangGraph + Zod + Dotenv
• AI orchestration with knowledge retrieval, streaming responses, and tool calls
• Local workflow data for demo products, recommendation rules, knowledge, and orders

Example Workflow

1. Upload enterprise policies, SOPs, or internal documents
2. Ask a support or operations question in natural language
3. Echo Assistant retrieves relevant knowledge and checks workflow context
4. Echo Assistant generates an actionable answer or converts an operating rule into executable workflow logic

Example:

User: Customer requests refund after 7 days. What should we do?

Echo Assistant: According to the refund SOP, orders exceeding 7 days require manual approval from operations. Suggested response template has been generated.

Run Locally

npm install
cp server/.env.example server/.env
# edit server/.env and set OPENAI_API_KEY / OPENAI_BASE_URL / OPENAI_MODEL
npm run dev

Frontend: http://localhost:5173

Backend: http://localhost:3001

OpenAI Config

server/.env supports custom OpenAI-compatible API hosts:

OPENAI_API_KEY=your_api_key
OPENAI_BASE_URL=https://api.openai.com
OPENAI_MODEL=gpt-4o-mini

OPENAI_BASE_URL can be either a host or a /v1 base URL. The server normalizes https://your-host to https://your-host/v1.

Project Structure

client/
  src/
    App.tsx                 # Product shell and workflow scenario pages
server/
  data/
    seed/ecommerce-sop.md   # Built-in demo SOP knowledge
    orders.csv              # Mock demo order data for query_order tool
  src/
    app.ts                  # Express app entry and middleware injection
    apis/                   # API route definitions
    controllers/            # Request validation and business orchestration
    middlewares/            # App middleware and error handling
    services/               # Agent, knowledge, product, and workflow services
samples/
  product-intake-demo.xlsx
  smart-thermos-after-sales-sop.md

API Demo

curl -X POST http://localhost:3001/api/chat \
  -H "Content-Type: application/json" \
  -d '{"message":"订单 E1001 用户想退款，应该怎么处理？","sessionId":"demo"}'

Useful APIs:

• GET /api/readiness: knowledge, model, and tool readiness
• GET /api/knowledge: list knowledge documents
• POST /api/knowledge: add a knowledge document with { "title": "...", "content": "..." }
• GET /api/knowledge/:id: preview a knowledge document
• DELETE /api/knowledge/:id: delete uploaded knowledge
• POST /api/chat/stream: stream agent status, answer tokens, and final trace data by SSE
• POST /api/products/enrich: enrich product intake rows for the demo workflow
• POST /api/recommendations/rules: convert a natural language demo rule into executable recommendation logic
• GET /api/recommendations/feed: preview recommendation results from saved demo rules