This project is a Retrieval-Augmented Generation (RAG) chatbot that allows users to ask questions and receive accurate answers based only on internal documents.
For this personal project, I used a coffee shop as a sample use case, where documents such as About Us and FAQs stored in Google Docs serve as the chatbot’s knowledge base.
The system uses n8n to handle document ingestion and chat requests, while a Next.js frontend provides the user interface. All responses are generated strictly from the indexed documents, helping reduce hallucinations and ensuring the chatbot only answers using approved information.
- Document-based question answering (RAG)
- Automatic document ingestion from Google Drive
- Semantic search using vector embeddings
- LLM-powered responses with strict grounding rules
- Session-based conversational memory
- API-first design for frontend integration
- Secure access via API key validation
Google Drive
│
▼
[n8n Ingestion Workflow]
├─ Download documents
├─ Split text into chunks
├─ Generate embeddings
└─ Store vectors in a vector database
│
▼
[n8n RAG Chat API Workflow]
├─ API key & input validation
├─ Retrieve relevant document chunks
├─ Generate response using an LLM
└─ Return structured API response
│
▼
[Next.js Frontend]
- Frontend: Next.js (React)
- Workflow Engine: n8n
- LLM & Embeddings: Google Gemini
- Vector Database: Pinecone
- Document Source: Google Drive
- Database: PostgreSQL (for n8n persistence)
- Infrastructure: Docker & Docker Compose
- Clone the repository
- Configure required environment variables (API keys and service configs)
- Start backend services using Docker Compose
- Import and activate the n8n workflows
- Document ingestion workflow
- RAG chat API workflow
- Run the Next.js frontend and connect it to the RAG API
Once running, documents added or updated in the connected Google Drive folder are automatically indexed and immediately available for chat queries.