🤖 Custom AI Chatbot Agents — OpenAI Platform

A full-stack, enterprise-grade AI chatbot platform designed to manage and run specialized AI Agents for multiple business departments (Sales, Customer Support, and Lead Prospecting). Built with a highly responsive React + Vite frontend and a robust Node.js/Express backend, integrating the OpenAI API (GPT-4) with real-time response streaming.

---

📈 Business Case & Real-World Impact

A scaling business required AI assistants to interact with clients across 3 distinct commercial touchpoints. Off-the-shelf chatbot software was too expensive, generic, and lacked the deep contextual awareness of the company's private business data and brand tone.

The Solution:

I designed and built a proprietary AI Agent Ecosystem:

• Specialized Agent Prompts: 3 distinct agent profiles (Sales Closer, Technical Support Specialist, and Outbound Lead Prospector) with strict system prompt boundaries.
• Real-time Response Streaming: Implemented server-sent events (SSE) to stream responses chunk-by-chunk, eliminating perceived loading latency.
• Persistent Conversation History: Client-side context management coupled with session storage to maintain conversation continuity.
• Safety Guards: Strict input validation and sanitization filters preventing prompt injection attacks.

Results:

• Simultaneous Handling: Efficiently serves hundreds of concurrent conversations.
• Ultra-low Latency: Average time-to-first-token response reduced to 800ms.
• Increased Conversion: 24/7 instant response increased Sales Team qualification metrics and liberated human agents to handle high-value negotiations.

---

📸 Screenshots

AI Chatbot Lobby	Interactive Streaming Chat

---

🏗️ Architecture & Technical Highlights

custom-ai-chatbot-agents-openai/
├── backend/          # Node.js + Express API, OpenAI integration, System Prompts
└── frontend/         # React (Vite) + Tailwind CSS, Streaming Client

Key Technical Implementations:

1. OpenAI Stream Integration (Backend): Using Express response write headers to stream tokens back to the client as they are generated by GPT-4:

   res.setHeader('Content-Type', 'text/event-stream');
   res.setHeader('Cache-Control', 'no-cache');
   
   const stream = await openai.chat.completions.create({
     model: "gpt-4",
     messages: [{ role: "system", content: systemPrompt }, ...messages],
     stream: true,
   });
   
   for await (const chunk of stream) {
     res.write(`data: ${JSON.stringify(chunk.choices[0]?.delta?.content || '')}\n\n`);
   }

2. Streaming Text Reader (Frontend): Utilizing the browser's fetch API and reader streams to parse SSE text tokens dynamically, updating the React UI state in real-time.

---

⚙️ Setup & Local Installation

Prerequisites

• Node.js (v18 or v20+)
• An OpenAI API Key

---

1. Backend Setup

1. Navigate to the backend folder:

   cd backend

2. Install dependencies:

   npm install

3. Create a .env file from the template:

   cp .env.example .env

4. Populate the .env file with your credentials:

   OPENAI_API_KEY="your-openai-api-key-here"
   PORT=5000

5. Start the backend server:

   npm start

   The API will run on http://localhost:5000.

---

2. Frontend Setup

1. Navigate to the frontend folder:

   cd ../frontend/brio-react

2. Install dependencies:

   npm install

3. Start the Vite development server:

   npm run dev

4. Open your browser to http://localhost:5173.

---

🇧🇷 Resumo em Português

O Ecossistema de IA Customizada é uma plataforma full-stack para gerenciar e executar agentes inteligentes de chat.

• Problema: Soluções prontas de chatbot de IA são genéricas, caras e não integram profundamente com as regras e dados de vendas/suporte da empresa.
• Solução: Backend Express integrado à API da OpenAI (GPT-4) com streaming de respostas em tempo real, controle de histórico e prompts customizados para 3 departamentos: Vendas, Suporte e Prospecção.
• Resultados: Tempo médio de resposta inicial de 800ms, suporte simultâneo a centenas de leads e aumento na eficiência comercial com atendimento 24/7.

---

Developed with ❤️ by João Melo