一体化的 AI Agent 服务平台 — 知识库 RAG、Telegram Bot、WhatsApp 通道、多语言客服All-in-one AI Agent service platform — Knowledge Base RAG, Telegram Bot, WhatsApp Channel, Multi-language Customer Service
AI Servicer 是一个完整的 AI Agent 服务平台，提供从知识库构建到多渠道分发的全链路能力。用户上传文档，系统自动完成解析（PDF/TXT/Markdown）、中文切片、ONNX 向量化 embedding 和存储。用户提问时，系统通过语义相似度检索最相关片段，结合大模型生成精准回答。AI Servicer is a complete AI Agent service platform, providing end-to-end capabilities from knowledge base construction to multi-channel distribution. Users upload documents, and the system automatically handles parsing (PDF/TXT/Markdown), text chunking, ONNX vector embedding, and storage. When users ask questions, the system retrieves the most relevant fragments through semantic similarity search and generates accurate answers with large language models.
平台支持 Telegram Bot 和 WhatsApp 集成，可通过面板直接向多个群组广播消息和文件。内置中英文双语界面，支持独立部署，可对接多种大模型。The platform supports Telegram Bot and WhatsApp integration, allowing broadcasting messages and files to multiple groups directly from the dashboard. Built-in bilingual interface, standalone deployment, supporting multiple LLMs.
覆盖 AI 客服全环节，从知识库管理到多渠道分发Covering all aspects of AI customer service, from knowledge base management to multi-channel distribution
上传文档自动解析、切片、向量化。查询时语义检索最相关片段，结合 LLM 生成答案。支持同名文档自动替换。Auto-parse, chunk, and vectorize uploaded documents. Semantic retrieval of relevant fragments for LLM-powered answers. Auto-replace on re-upload.
支持 DeepSeek 等主流大语言模型，通过 API 直接调用。RAG 上下文嵌入 user message，确保模型准确引用知识库内容。Supports DeepSeek and other mainstream LLMs via direct API calls. RAG context embedded in user message for accurate knowledge base references.
原生 fetch 长轮询 Telegram API。群聊 @bot 触发 AI 回答。面板支持多群组选择、文字/文件/文档批量发送。Native fetch long-polling Telegram API. @bot trigger in group chats. Batch send text/files/docs to multiple groups.
通过 WhatsApp Business API 接入，支持客户消息自动回复、知识库问答、文件发送。面板统一管理所有通道。Integrate via WhatsApp Business API for auto-reply, knowledge base Q&A, and file delivery. Unified management across all channels.
纯 HTML iframe 聊天控件，粘贴即用。支持自定义样式，可嵌入任何网站。公开 API 端点无需鉴权。Pure HTML iframe widget, copy-paste ready. Customizable style, embeddable on any site. Public API endpoint requires no auth.
JWT 认证 + 注册/登录。自动创建商户、项目和 Bot。Token 用量实时统计，支持多商户管理。JWT auth + register/login. Auto-create merchants, projects, and bots. Real-time token usage tracking, multi-merchant support.
多渠道统一接入，一个后台管理所有客户对话Unified multi-channel integration, manage all customer conversations from one dashboard
内嵌聊天 Widget，粘贴 iframe 代码即可在任何网站启用 AI 客服。支持自定义样式，无需登录。Embedded chat widget, paste iframe code to enable AI customer service on any website. Customizable style, no login required.
Telegram Bot 自动回复，群聊 @bot 触发。面板支持多群组广播、文字/文件/文档发送。原生 fetch 长轮询。Telegram Bot auto-reply, @bot trigger in groups. Batch broadcast, send text/files/docs. Native fetch long-polling.
WhatsApp Business API 集成，自动回复客户消息，知识库问答，文件发送。面板统一配置管理。WhatsApp Business API integration, auto-reply to customer messages, knowledge base Q&A, file delivery. Unified dashboard management.
前端 React + 后端 Express，本地 ONNX 向量化，支持多种 LLM 接入React frontend + Express backend, local ONNX vectorization, multi-LLM support
从文档上传到 AI 回答的完整链路Complete pipeline from document upload to AI answer
轻量化技术选型，单机可运行，无需 GPULightweight tech stack, runs on a single machine, no GPU required
React 18 + TypeScript + Vite + Tailwind CSS。Lucide 图标、React Router SPA。中英文 i18n 通过 React Context 实现。Lucide icons, React Router SPA. Bilingual i18n via React Context.
Express 4 + CommonJS。JWT 认证Auth、bcrypt 加密Encryption、multer 文件上传Uploads。
@xenova/transformers + ONNX（sentence-transformers/all-MiniLM-L6-v2）。纯 CPU 推理，无需 GPU。CPU-only inference, no GPU needed.
支持 DeepSeek 等主流 API 模型。直接 HTTP 调用，可灵活切换，延迟更低。Supports DeepSeek and other mainstream LLM APIs. Direct HTTP calls for lower latency, flexible model switching.
JSON 文件存储（免数据库）。用户/商户/项目/Bot/文档/向量全部以 JSON 持久化。File-based JSON storage (no database needed). Users, merchants, projects, bots, docs, vectors all persisted as JSON.
Ubuntu 22.04 + Nginx。50M 文件上传限制upload limit。完全独立运行Fully standalone。
轻量化设计，单机部署即可应对日常使用Lightweight design, single-machine deployment handles daily usage
© 2026 0xAi | AI Servicer