Ai Voice Chatbot With Elevenlabs & Openai For Customer Service And Restaurants 2846 – n8n Workflows – Free Template

What This Workflow Does

This n8n workflow creates an advanced Voice RAG (Retrieval-Augmented Generation) Chatbot that combines speech recognition with AI-powered responses. Users can ask questions to a voice agent powered by ElevenLabs, which are then processed through OpenAI’s language models using context retrieved from a Qdrant vector database. The system generates intelligent, contextual responses based on your uploaded documents and knowledge base.

How It Works

The workflow operates through a series of integrated steps:

• A voice query is submitted to the ElevenLabs voice agent through a webhook trigger
• The question is extracted from the webhook body and passed to the n8n AI Agent
• The AI Agent queries the Qdrant vector database using OpenAI embeddings to retrieve relevant document excerpts
• Retrieved context is combined with the original question and sent to the OpenAI language model
• The model generates a contextual response based on your knowledge base
• The response is returned through the webhook and processed by ElevenLabs for voice synthesis
• A buffer window memory maintains conversation context for multi-turn interactions

Use Cases

• Customer Support Chatbots: Deploy voice-activated customer service that answers questions using your company’s documentation and knowledge base
• Educational Tutoring Systems: Create interactive learning assistants that explain concepts using course materials and educational documents
• Internal Knowledge Management: Build employee-facing voice assistants that retrieve information from company policies, procedures, and training materials
• Healthcare Information Systems: Develop accessible voice interfaces that answer patient questions using medical literature and treatment guidelines
• Personal Research Assistants: Create voice-activated research tools that query academic papers, articles, and reference materials on demand

Nodes Used

• Manual Trigger: Initiates workflow testing
• Webhook: Receives voice queries from ElevenLabs with question data
• HTTP Request: Makes API calls to external services
• Google Drive: Retrieves documents for knowledge base creation
• Document Default Data Loader: Processes and prepares documents for vectorization
• Text Splitter Token Splitter: Breaks documents into manageable chunks for embedding
• Embeddings OpenAI: Converts text chunks into vector embeddings
• Vector Store Qdrant: Stores and manages vector embeddings in a searchable database
• Tool Vector Store: Enables the AI agent to query the vector database
• LM Chat OpenAI: Generates responses using OpenAI’s language model
• Agent: Orchestrates the entire RAG process and manages agent behavior
• Memory Buffer Window: Maintains conversation history for context
• Respond to Webhook: Returns the generated response to ElevenLabs
• Sticky Note: Provides workflow documentation and instructions

Prerequisites

• Active n8n instance or n8n cloud account
• OpenAI API key with access to GPT models and embeddings
• ElevenLabs account with voice agent configuration
• Qdrant vector database instance (cloud or self-hosted)
• Google Drive account with documents to use as knowledge base
• Documents formatted as PDF, DOCX, TXT, or other standard formats
• Basic understanding of RAG systems and webhook configuration
• Network access to configure webhook endpoints between n8n and ElevenLabs

Difficulty Level

Advanced. This workflow requires knowledge of AI/ML concepts including RAG systems, vector embeddings, and prompt engineering. Users should be comfortable configuring multiple API integrations, setting up vector databases, and debugging complex multi-step automations. Previous experience with n8n and API workflows is strongly recommended.

This workflow template is shared under the n8n fair-code license. Free to use and modify.