An end-to-end AI-powered portfolio chatbot built using n8n, OpenAI, and Pinecone, implementing a Retrieval-Augmented Generation (RAG) architecture.
Integrated seamlessly into a personal portfolio website to deliver accurate, knowledge-base-driven answers.
This project showcases a production-ready GenAI system where an AI assistant answers user queries only from verified personal data, including:
- Resume
- GitHub repositories
- LinkedIn profile
- Portfolio content
The chatbot avoids hallucinations by retrieving relevant information from a vector database before generating responses.
- 🤖 RAG-based AI Chatbot
- 🔗 n8n AI Agent orchestration
- 🧠 Pinecone Vector Store for semantic search
- 🧬 OpenAI embeddings & chat model
- 💬 Session-based memory (multi-turn conversations)
- 🌐 Website-embedded chatbot
- 🔐 Safe text rendering (no HTML / XSS risks)
- 🚀 Production webhook setup
User (Website Chatbot)
↓
Portfolio Website (JavaScript)
↓
n8n Webhook
↓
AI Agent (System Prompt + Memory)
↓
Pinecone Vector Store (RAG)
↓
OpenAI Chat Model
↓
Respond to Webhook
↓
User
🛠️ Tech Stack
Backend & Automation
n8n – Workflow automation & AI Agent
OpenAI API – Embeddings & chat completion
Pinecone – Vector database (knowledge base)
Webhooks – Frontend ↔ backend communication
Frontend
HTML / CSS / JavaScript
Custom chatbot UI
Session handling using localStorage
📂 Project Structure ├── index.html # Portfolio website ├── main.js # Chatbot logic (frontend) ├── styles.css # UI styling ├── assets/ │ └── screenshots/ │ ├── n8n-workflow.png │ ├── chatbot-ui.png │ ├── pinecone-index.png ├── README.md
The n8n automation consists of the following nodes:
Webhook
Receives user message + sessionId
AI Agent
System instructions
Tool calling
Simple Memory
Maintains session-based context
Pinecone Vector Store
Retrieves relevant knowledge
OpenAI Embeddings
Converts query into vectors
Respond to Webhook
Sends response back to frontend
The chatbot is embedded directly into the portfolio website and supports:
Real-time conversation
Typing indicator
Session persistence
Clean text formatting (markdown-safe)
The chatbot answers questions only from indexed documents:
Resume (PDF)
GitHub repositories
LinkedIn profile
Portfolio data
❌ No hallucination ❌ No guessing ✅ Verified responses only
Each user gets a unique sessionId
Stored in browser localStorage
Sent with every request to n8n
Enables smooth multi-turn conversations
Uses textContent (no innerHTML)
Markdown symbols safely handled
XSS-safe frontend
Stable webhook execution
1️⃣ Clone the repository git clone https://github.com/Akshitvats026/akshit-n8n-rag-agent.git cd your-repo-name
2️⃣ Open the website Open index.html in browser
3️⃣ Configure n8n
Import the workflow
Add OpenAI & Pinecone credentials
Use Production Webhook URL
⚙️ n8n Backend https://akshu-automation.app.n8n.cloud
This project demonstrates:
Real-world RAG implementation
Automation engineering using n8n
System design & AI integration
Full-stack AI development mindset
Production-grade chatbot architecture
Akshit Vats Web Developer | Machine Learning | Generative AI
GitHub: https://github.com/Akshitvats026
LinkedIn: https://www.linkedin.com/in/akshitvats026
If you like this project:
🍴 Fork it
Built with ❤️ using n8n, OpenAI, and Pinecone