GPT-5 RAG Customer Support Agent for Telegram - n8n Workflow

SaaS and E-commerce Companies: Those needing 24/7 automated support leveraging existing FAQ documentation.
Automation Specialists: Users looking for advanced examples of building complex AI agents within n8n.
Developers & Integrators: Individuals utilizing the LangChain n8n node suite for integrating vector databases (Pinecone) with large language models (GPT-5).
n8n Users: Anyone seeking robust n8n templates for generative AI customer service solutions.

Customer support automation demands accuracy and context. Generic chatbots often fail when faced with complex, multi-turn queries or when required to access proprietary policy data. This specialized n8n workflow solves this by implementing a full Retrieval-Augmented Generation (RAG) pipeline directly connected to Telegram.

This setup utilizes the power of the GPT-5 language model, integrated via the n8n LangChain agent node, and leverages a Pinecone vector store (holding your FAQ data) as a knowledge retrieval tool. Every customer query received by the n8n trigger is processed, contextually augmented, and answered with precision. Furthermore, the inclusion of a memory node ensures coherent, multi-turn conversations, making this an ideal, production-ready n8n template for superior customer interaction.

This intelligent n8n workflow operates in a seamless loop, starting with a customer inquiry and ending with a detailed, contextual response:

1. Initiation (The n8n Trigger): The process starts when the Telegram Trigger n8n node receives a new incoming message from a customer. This message triggers the entire pipeline.


2. Agent Orchestration: The received message is passed to the GPT-5 Customer Support Agent n8n node (a LangChain Agent). This agent decides whether the query requires external knowledge retrieval or can be answered conversationally.


3. Contextual Memory: The conversation passes through the Simple Memory n8n node, which maintains a buffer of the last 15 interactions for that specific Telegram chat ID, ensuring the GPT-5 model has necessary conversation context.


4. RAG Retrieval: If external data is needed, the LangChain agent calls the Pinecone Vector Store tool. This n8n node performs a similarity search against the 'awm-n8n' index (Customer FAQ Data & Policies), using the Embeddings OpenAI node to generate the necessary query vectors.


5. Answer Generation: The GPT-5 model receives the original query, the conversation history, and the retrieved relevant context documents. It synthesizes a final, grounded answer.


6. Response: The final, clean output generated by the AI agent is routed to the final Telegram n8n node, which uses the original chat ID to send the automated support response back to the customer.

To deploy this RAG customer support n8n workflow, follow these steps:

1. Import the n8n Workflow: Copy the provided JSON and import it into your n8n instance (Cloud or Self-Hosted) via the 'New' menu > 'Import from JSON'.


2. Credential Setup: You must configure the following credentials for the respective n8n node connections:

1. Index Data: Ensure your customer support data and policies are indexed and searchable within your Pinecone vector database using appropriate OpenAI embeddings.


2. Test and Activate: Save the n8n workflow, ensure all credentials are valid, and activate the Telegram Trigger n8n trigger to begin monitoring incoming messages.

Telegram Trigger (n8n trigger):
Function: Serves as the starting point, listening for any new messages sent to the configured Telegram bot.
Key Configuration: Listens for 'message' updates.
GPT-5 Customer Support Agent (LangChain Agent n8n node):
Function: The central intelligence that orchestrates memory, language model (GPT-5), and tools (Pinecone) to generate a response.
Key Configuration: Takes the input text directly from the Telegram trigger.
OpenAI Chat Model (GPT-5):
Function: Provides the core reasoning and generation capability using the specified 'gpt-5' model.
Key Configuration: Model set explicitly to gpt-5.
Simple Memory (Memory Buffer Window n8n node):
Function: Stores the conversation history (context) to enable multi-turn dialogue. Essential for a robust n8n workflow.
Key Configuration: Session Key uses {{ $json.message.chat.id }} to track individual users; Context Window Length is set to 15 interactions.
Pinecone Vector Store (LangChain Tool n8n node):
Function: Acts as the external knowledge base tool for the AI agent, retrieving grounded facts from the vector database.
Key Configuration: Mode set to 'retrieve-as-tool'. Index name is configured to awm-n8n. Tool description is 'Customer FAQ Data & Policies'.
Embeddings OpenAI:
Function: Generates vector representations of the user query, allowing Pinecone to perform accurate similarity searches for RAG.
Telegram (n8n node):
Function: Sends the synthesized answer back to the customer.
Key Configuration: Text field is mapped to the output of the AI agent ({{ $json.output }}); Chat ID is mapped back to the original triggering chat ID from the Telegram Trigger n8n node.