cs-ai-agent
agent
Uyari
Health Uyari
- License — License: Apache-2.0
- Description — Repository has a description
- Active repo — Last push 0 days ago
- Low visibility — Only 5 GitHub stars
Code Gecti
- Code scan — Scanned 12 files during light audit, no dangerous patterns found
Permissions Gecti
- Permissions — No dangerous permissions requested
Purpose
This tool is a comprehensive, AI-driven customer support system that unifies live chat, knowledge base retrieval (RAG), ticketing, and seamless human agent handoff.
Security Assessment
The overall risk is Low. The automated code scan checked 12 files and found no dangerous patterns, no hardcoded secrets, and no excessive permissions. However, since this is a complete customer support platform, it inherently processes and stores sensitive data, including customer chat logs, ticket details, and internal knowledge bases. The system requires network access to communicate with external AI models (via OpenAI-compatible APIs), a vector database (Qdrant), and standard databases (SQLite/MySQL). While the base code appears safe, standard security precautions should be taken during deployment to protect the underlying database and secure the API keys needed for the LLM.
Quality Assessment
The project uses the permissive Apache-2.0 license and is under highly active development, with its most recent push occurring today. It is well-documented and features a solid, modern technology stack (Golang, Next.js, React). However, it currently has very low community visibility with only 5 GitHub stars. This indicates that while the project is actively maintained by its creators, it has not yet been widely tested or adopted by the broader open-source community.
Verdict
Safe to use, but deploy with standard security practices for handling customer data and external API keys, keeping in mind its low current community adoption.
This tool is a comprehensive, AI-driven customer support system that unifies live chat, knowledge base retrieval (RAG), ticketing, and seamless human agent handoff.
Security Assessment
The overall risk is Low. The automated code scan checked 12 files and found no dangerous patterns, no hardcoded secrets, and no excessive permissions. However, since this is a complete customer support platform, it inherently processes and stores sensitive data, including customer chat logs, ticket details, and internal knowledge bases. The system requires network access to communicate with external AI models (via OpenAI-compatible APIs), a vector database (Qdrant), and standard databases (SQLite/MySQL). While the base code appears safe, standard security precautions should be taken during deployment to protect the underlying database and secure the API keys needed for the LLM.
Quality Assessment
The project uses the permissive Apache-2.0 license and is under highly active development, with its most recent push occurring today. It is well-documented and features a solid, modern technology stack (Golang, Next.js, React). However, it currently has very low community visibility with only 5 GitHub stars. This indicates that while the project is actively maintained by its creators, it has not yet been widely tested or adopted by the broader open-source community.
Verdict
Safe to use, but deploy with standard security practices for handling customer data and external API keys, keeping in mind its low current community adoption.
一个以 AI Agent 为核心的智能客服系统,融合在线会话、知识库检索、工单处理与人工协同接管能力。 An AI-agent-first customer support system that unifies live chat, knowledge retrieval, ticketing, and seamless human handoff.
README.md
贝壳AI客服
An AI-agent-first customer support system that unifies live chat, knowledge retrieval, ticketing, and seamless human handoff.
贝壳AI客服是一个以 AI Agent 为核心 的智能客服系统,面向需要同时处理在线咨询、知识库问答、人工接管和工单流转的团队。
它不是一个简单的聊天机器人,而是一套围绕客服场景设计的完整系统:
- AI Agent 先接待,优先处理常见问题与标准流程
- 知识库检索驱动回答,支持 RAG 场景
- 在低置信度、无答案或命中规则时转人工
- 后台可管理会话、工单、客服组、知识库、AI 配置、Skills 和 MCP
- 提供管理后台与客服工作台两套视角
核心能力
- AI-first 客服流程:AI Agent 优先接待,支持自动回复、兜底与人工协同
- 在线会话系统:支持访客会话、会话分配、转接、关闭、未读状态与实时消息
- 知识库 RAG:支持知识库、文档、切片、检索日志与检索质量分析
- 工单系统:支持会话转工单、工单分类、状态流转与处理闭环
- 客服组织管理:支持客服档案、客服组、排班与分配能力
- AI 扩展能力:支持 Skills、MCP 调试与外部能力接入
- 双工作区体验:管理后台负责配置和运营,客服工作台负责处理会话
AI Agent 工作流程
flowchart TD
A[用户发起咨询<br/>Web Widget / Open API] --> B[创建或匹配会话]
B --> C[进入 AI Agent]
C --> D{是否命中 Skill / 工具流程?}
D -- 是 --> E[执行 Skill / MCP Tool]
D -- 否 --> F[进入知识库检索]
F --> G[召回相关知识片段]
G --> H[LLM 生成回答]
E --> H
H --> I{回答是否可信且可直接返回?}
I -- 是 --> J[返回给用户]
J --> K{问题是否结束?}
K -- 否 --> C
K -- 是 --> L[结束会话或沉淀数据]
I -- 否 --> M[触发转人工策略]
M --> N[进入待接入池 / 指定客服组]
N --> O[客服工作台接管]
O --> P{是否需要工单跟踪?}
P -- 是 --> Q[创建或关联工单]
P -- 否 --> R[人工继续处理]
Q --> R
R --> S[问题解决并关闭]
适用场景
- 官网在线客服
- SaaS 产品支持
- AI + 人工混合接待
- 企业内部服务台或运营支持台
- 需要知识库问答与人工协同的客服团队
技术栈
- Backend: Golang
- Frontend: Next.js 16 + React 19 + shadcn/ui + Tailwind CSS
- Database: SQLite / MySQL
- Vector DB: Qdrant
- AI: OpenAI-compatible LLM / Embedding + RAG + SKILLS + MCP
项目结构
.
├── cmd/ # server / migration / generator
├── internal/
│ ├── controllers/ # API controllers
│ ├── services/ # business services
│ ├── repositories/ # data access
│ ├── models/ # GORM models
│ ├── migration/ # data migrations
│ └── ai/ # LLM / RAG / MCP related logic
├── web/ # admin console (Next.js)
├── widget/ # embeddable customer chat widget
├── config/ # config files
└── docs/ # project docs
快速开始
1. 环境要求
- Go
1.26+ - Node.js
20+ pnpm- Qdrant
2. 准备配置
复制示例配置:
cp config/config.example.yaml config/config.yaml
默认配置使用:
- SQLite:
data/app.db - Backend:
http://127.0.0.1:8083 - Qdrant gRPC:
127.0.0.1:6334
3. 启动 Qdrant
如果你本地还没有 Qdrant,可以用 Docker 快速启动:
docker run -p 6333:6333 -p 6334:6334 qdrant/qdrant
4. 安装前端依赖
cd web
pnpm install
cd ..
5. 启动项目
同时启动后端和前端:
make run
或分别启动:
make run-server
make run-web
常用命令
make run # 同时启动后端和前端
make run-server # 启动后端
make run-web # 启动前端
make build # 构建后端二进制
make test # 运行 Go 测试
make tidy # go mod tidy
make generator # 执行代码生成
make enums # 生成前端枚举
make migration # 执行 migration
系统视角
- 管理后台:负责 AI Agent、知识库、客服组、工单与运营配置
- 客服工作台:负责接管会话、处理消息与人工服务
- Widget:负责承接用户侧咨询入口
这使得贝壳AI客服可以同时覆盖:
- AI 接待
- 人工协同
- 知识驱动回答
- 工单追踪闭环
开源定位
贝壳AI客服适合作为以下方向的开源基础项目:
- AI 客服系统
- AI Helpdesk / AI Support Platform
- RAG + Human Handoff 的落地样板
- 面向企业场景的 AI Agent 应用框架
如果你在寻找一个 以 AI Agent 为中心,而不是仅仅把 LLM 嵌进聊天框 的客服系统,这个项目就是为此设计的。
Yorumlar (0)
Yorum birakmak icin giris yap.
Yorum birakSonuc bulunamadi