EcoMind - AI E-commerce Expert Dialogue System

English | 日本語 | 简体中文

An intelligent e-commerce knowledge base Q&A system powered by local LLMs. It utilizes RAG (Retrieval-Augmented Generation) technology to answer questions related to cross-border e-commerce.

Key Features

• 🤖 Multi-Model Support: Supports multiple LLM providers including OLLAMA (local), OpenAI, and Anthropic.
• 📚 Knowledge Base Management: Automatically indexes cross-border e-commerce articles (TikTok, AI tools, finance, etc.).
• 🔍 Hybrid Search: Combines semantic search and keyword matching for high accuracy and relevance.
• 💬 Multi-turn Conversation: Maintains context for continuous and natural dialogue.
• 📖 Source Citations: Automatically cites source articles for every answer to ensure traceability.
• 🌊 Streaming Response: Real-time answer generation for a ChatGPT-like experience.
• 🎨 Advanced Markdown: Full support for tables, lists, and code syntax highlighting.
• ⚙️ Web Configuration: Configure model APIs and knowledge base paths directly in the UI without editing code or environment variables.

Quick Start

Installation

macOS (Homebrew)

brew tap ecomind/tap
brew install ecomind

Windows / Linux (NPM)

npm install -g ecomind-ai

Running the App

Once installed, simply run:

ecomind start

This will automatically set up the Python environment and launch the server at http://localhost:8080.

Prerequisites

• Node.js (v16+)
• Python 3.9+
• OLLAMA or oMLX (if using local LLMs)

Manual Installation

If you prefer to set up the environment manually:

1. Clone the repository:

   git clone <repository-url>
   cd EcoMind

2. Install dependencies:

   pip install -r requirements.txt

3. Start the server:

   ./start_ecomind.sh

Setup & Configuration

Once the web interface is open (http://localhost:8080):

1. Click the Gear Icon (⚙️) in the sidebar header.
2. Provider: Select OLLAMA / oMLX for local or OpenAI/Anthropic for cloud models.
3. Model: Enter the model string (e.g., Qwen3-Coder-30B-A3B-Instruct-4bit).
4. Base URL: For local oMLX, use http://127.0.0.1:8000.
5. Wiki Path: Enter the absolute path to your documents folder.
6. Save: Click "Save Configuration". The system will automatically index your documents.

Project Structure

.
├── bin/              # CLI scripts (Node.js)
├── src/
│   ├── api/          # FastAPI application and routes
│   ├── storage/      # Vector DB and persistence layer
│   ├── indexing/     # Document indexing and preprocessing
│   ├── search/       # Search engine logic
│   ├── rag/          # Core RAG processes
│   └── llm/          # LLM client integration
├── frontend/         # Frontend static assets
├── data/             # Database and config files
└── docs/             # Documentation in multiple languages

License

MIT License

Author

Built with EcoMind team