RAG Customer Support Agent: GPT-5, Telegram & Pinecone

This n8n workflow creates a powerful Retrieval-Augmented Generation (RAG) agent for customer support, integrating Telegram for real-time user interactions, GPT-5 via OpenAI for advanced reasoning, and Pinecone for vector-based retrieval of relevant FAQs. It listens for incoming Telegram messages, processes queries with an AI agent that maintains a memory buffer of the last 15 interactions per user, retrieves context from a 'Customer FAQ' namespace in Pinecone, and generates accurate, personalized responses sent back via Telegram.

Key benefits include seamless automation of support tickets, reduced response times, and grounded answers that minimize hallucinations by pulling from your knowledge base. The workflow handles multi-turn conversations coherently, supports multilingual queries, and scales effortlessly for e-commerce, SaaS, or community support.

Ideal use cases: Automating FAQ handling for online stores, providing 24/7 policy explanations, or assisting with product troubleshooting. Setup involves configuring Telegram bot credentials, OpenAI API keys, Pinecone index details, and uploading documents for embedding—enabling end-to-end AI-driven support in minutes.