Multi-Workspace Live Chat & Self-Hosted Autonomous AI Engine

Live Customer Chat & Smart Automation
Powered by Sobi AI Core

Every Customer Conversation. One Intelligent Platform

A self-hosted, multi-workspace live-chat platform with an operator inbox, Telegram support bot, modern analytics, and a provider-agnostic AI engine (OpenAI, Anthropic, DeepSeek, Groq, Ollama) featuring Semantic RAG and human handoff.

3-Minute Installation View Live Demo Get Embed Script
Flask & Socket.IO
Next.js 14 & Tailwind
Telegram Support Bot
Provider-Agnostic LLM
7 Locales & Full RTL
Sobi AI Neural Core Platform Overview
100%
Self-Hosted & Full Data Ownership
7 Locales
Full RTL & Multi-Language Support
10+
AI Providers (OpenAI, Claude, DeepSeek...)
<50ms
Real-Time Socket.IO Messaging
Powerful AI Intelligence

Sobi AI — Provider-Agnostic Engine

Configured via a single engine.json file and surfaced through the visual Sobi AI Studio panel.

Smart Answers, Semantic RAG & Automated Human Handoff

Sobi AI is a multi-provider brain with automated fallback, conversation memory, persona controls, and daily budget caps.

  • Provider-Agnostic Engine

    Connects to any OpenAI-compatible API (OpenAI, Groq, Gemini, DeepSeek, Ollama) or Anthropic Claude.

  • Role Routing & Fallback

    Define roles (draft, final) with provider priorities. If one provider fails, Sobi instantly falls back to the next.

  • Semantic RAG (Vector Embeddings)

    Embeds knowledge chunks into vector indices to retrieve precise contextual answers for visitor questions.

  • Human Handoff [[ESCALATE]]

    When unsure, Sobi emits [[ESCALATE]], sends a handoff message, and alerts operators instantly via Telegram.

  • Budget Guard & Cost Analytics

    Daily token and cost caps with full per-call logging (tokens, latency, cost USD) and CSAT feedback.

Playground (Live)
Models & Keys
Persona & RAG
Safety & Handoff
Usage & Cost
Hello! I am Sobi AI. How can I assist your support team today?
OpenAI (GPT-4o) 142ms
System Architecture

Unified & Scalable Infrastructure

Click any component node below to inspect its data flow and default port.

Live Chat Widget

Embeddable Vanilla JS client

Static Script
Flask Backend

Socket.IO & Data Pipeline

Port :5001
Next.js Admin Panel

Operator Inbox & Sobi AI Studio

Port :3001
Telegram Support Bot

Operator Alerts & Outbound Replies

Telegram API
Vanilla JS Chat Widget: Light, zero-dependency script loaded on any customer website over Socket.IO.
Feature Catalog

Everything You Need for Elite Support

A complete self-hosted suite designed for modern teams and growing SaaS products.

Live Chat & Operator Inbox

Real-time messaging over Socket.IO (text, voice, files, images). Message editing, replies, reactions, CSAT ratings, and AI feedback.

Multi-Workspace (Multi-Tenant)

Isolated workspaces per client website with custom branding, logos, helper profiles, operator roles, and billing plans.

Integrated CRM

Manage contacts, pipelines, deals, activity timelines, and custom properties right inside your admin panel.

Analytics & Visitor Tracking

KPI dashboard, conversation analytics, hourly/daily charts, CSAT ratings, and visitor geo-location via MaxMind GeoLite2.

Web Push & Telegram Bot

Browser Web Push notifications (VAPID) and instant Telegram alerts for operators with direct Telegram-to-visitor replies.

7 Locales & Full RTL

First-class support for English, Persian, German, Arabic, Russian, French, and Spanish with native RTL/LTR switching.

Easy Integration

Embed the Widget on Any Website

Enter your workspace ID and backend URL to generate your custom loader script.

index.html
<script
  src="https://your-domain.com/sobinta-chatbox.js"
  data-workspace-id="demo-workspace-123"
  data-api="https://your-domain.com:5001"
  defer></script>
Deployment Guide

Automated 1-Command Setup

The setup.sh script provisions Python virtualenvs, installs Node dependencies, and creates template config files.

Terminal — Quickstart
# 1. Clone project and run automated setup script
git clone https://github.com/sobi-chat/sobi-chatbox.git
cd sobi-chatbox
chmod +x setup.sh && ./setup.sh

# 2. Run Flask Python backend on port 5001
cd backend && python3 chat_server.py

# 3. Run Next.js 14 Admin Panel on port 3000 / 3001
cd frontend && npm run dev