English | 简体中文

Open Sales Assistant is a self-hosted AI sales assistant starter kit for small teams and individual sales workflows. It helps a team turn sales materials into a searchable knowledge base, answer customer questions, review AI-generated sales suggestions, and keep basic customer follow-up records.

The included dataset is an English demo for a training course sales scenario. It is kept as an initial example so new users can run the project immediately and understand the workflow.

For Chinese readers, see README.zh-CN.md for the full Chinese project overview, quick start, AI API configuration, and roadmap.

• H5/customer chat entry for sales consultation.
• Admin login and sales knowledge base management.
• Text, document, and image material upload.
• AI-assisted material analysis and pending suggestions.
• FAQ, sales playbook, concern handling, and knowledge summary editing.
• Customer profile, chat history, AI analysis, and follow-up records.
• Admin operation logs for sign-in, data export/backup, knowledge, material, customer, certificate, and settings changes.
• Certificate record import and query demo flow.
• JSON-file storage for local development and small self-hosted trials.
• Planned optional database adapters while keeping JSON storage as the default lightweight mode.
• English-first UI with an English/Chinese switch for customer chat, admin login, and admin console.
• English and Chinese demo fixtures. Local demo mode can switch the demo dataset together with the UI language.

npm install
cp .env.example .env
npm run dev

Docker quick start:

cp .env.example .env
docker compose up -d --build

Open:

• Customer entry: http://localhost:3100/
• Customer login: http://localhost:3100/login.html
• Admin console: http://localhost:3100/admin.html
• Health check: http://localhost:3100/api/health

Default admin credentials are configured in .env:

ADMIN_USERNAME=admin
ADMIN_PASSWORD=请设置一个强密码

Set AI_API_KEY before using AI analysis or chat generation.

Open Sales Assistant uses environment variables for AI settings. It does not store API keys in the admin UI or in data/db.json.

Create your local .env file:

cp .env.example .env

Then edit .env and set your own OpenAI-compatible API key:

AI_PROVIDER=kimi
AI_BASE_URL=https://api.moonshot.cn/v1
AI_API_KEY=your_api_key_here
AI_TEXT_MODEL=kimi-k2.6
AI_VISION_MODEL=kimi-k2.6

Do not commit .env to GitHub. The repository .gitignore already excludes it.

If AI_API_KEY is empty or still contains the placeholder text, the app uses local fallback responses. This is enough to test pages and data flow, but AI analysis and AI chat generation will not call a real model.

Provider examples:

# Kimi / Moonshot
AI_PROVIDER=kimi
AI_BASE_URL=https://api.moonshot.cn/v1
AI_TEXT_MODEL=kimi-k2.6
AI_VISION_MODEL=kimi-k2.6

# OpenAI
AI_PROVIDER=openai
AI_BASE_URL=https://api.openai.com/v1
AI_TEXT_MODEL=gpt-4o-mini
AI_VISION_MODEL=gpt-4o-mini

# DeepSeek
AI_PROVIDER=deepseek
AI_BASE_URL=https://api.deepseek.com
AI_TEXT_MODEL=deepseek-chat
AI_VISION_MODEL=deepseek-chat

# DashScope / Tongyi Qianwen
AI_PROVIDER=dashscope
AI_BASE_URL=https://dashscope.aliyuncs.com/compatible-mode/v1
AI_TEXT_MODEL=qwen-plus
AI_VISION_MODEL=qwen-vl-plus

The /api/health endpoint only reports whether AI is configured and which model names are active. It never returns the API key.

Start here:

• English overview: README.md
• Chinese overview: README.zh-CN.md

npm test

The test command runs JavaScript syntax checks and a local smoke test for the health endpoint plus the main static pages. It does not require an AI API key.

data/db.json contains sample English training-course sales data. Treat it as demo content only:

• It demonstrates how sales materials become a knowledge base.
• It is not a production dataset.
• Replace it with your own product, service, pricing, FAQ, and compliance rules before real use.

To restore the committed demo dataset after local experiments:

npm run reset-demo

Language-specific demo reset commands:

npm run reset-demo:en
npm run reset-demo:zh

The default UI language is English. Selecting Chinese from the UI switches the interface to Chinese and, in local demo mode, switches the demo dataset to the Chinese fixture. In production, public demo switching is disabled unless DEMO_LANGUAGE_SWITCH=true is explicitly configured.

This repository is not a SaaS platform. The current goal is to provide a clear, self-hosted team sales assistant that contributors can run locally, adapt, and improve.

Future work may add stronger team collaboration, role permissions, database storage, integrations, and evaluation workflows, but the open-source baseline remains focused on helping a team own its sales assistant without depending on a hosted commercial platform.

JSON file storage will remain supported as the default lightweight mode. Future PostgreSQL/MySQL support should be added through optional storage adapters, not by removing the current JSON workflow.

• Improve sales follow-up workspace: status, owner, next action, and reminders.
• Improve operation logs and safer data backup/restore workflows.
• Add editable AI suggestions and knowledge version history.
• Add optional PostgreSQL/MySQL storage adapters while preserving JSON storage.
• Improve Docker deployment and production self-hosting examples.
• Add customer segmentation and sales funnel dashboards.
• Add connector examples for CRM, forms, chat tools, and spreadsheets.
• Add tests and prompt evaluation cases for stable community contributions.

This project is being prepared as an open-source sales assistant starter kit. Codex would be used for maintainer workflows such as issue triage, PR review, test generation, release notes, documentation updates, and prompt/evaluation improvements.

MIT