Voiceflow Agent Backend with RAG, Qdrant, and Google Calendar Integration - n8n Workflow

Companies wanting to deploy sophisticated chatbots or voice agents.
Developers looking for advanced n8n workflow examples integrating Langchain and vector databases.
E-commerce operations needing automated order status lookups.
Businesses requiring AI-driven scheduling based on natural language input.

This robust n8n workflow acts as the intelligent backbone for a Voiceflow chatbot, voice agent, or phone agent. It solves the complexity of providing immediate, context-aware customer support by integrating a Retrieval-Augmented Generation (RAG) system via Qdrant and a Google Drive knowledge base. Furthermore, this n8n workflow handles two critical transactional tasks: looking up order statuses via an external API and creating time-specific appointments directly in Google Calendar. It utilizes a powerful OpenAI n8n node within the workflow to correctly parse complex date formats from natural language input. This comprehensive setup ensures the AI agent can answer intricate questions and perform actions, moving beyond simple conversational scripts, all managed through this single n8n template.

The entire system is designed to service external calls from a Voiceflow conversational agent via three distinct webhook endpoints, alongside an initial manual setup step for the knowledge base.

1. RAG Knowledge Base Setup (Manual Trigger): The workflow starts with the manual setup path, triggering an HTTP Request to create and refresh the Qdrant collection. It then fetches documents from Google Drive, processes them using the Document Loader and Token Splitter n8n nodes, generates embeddings via an OpenAI n8n node, and finally inserts them into the Qdrant Vector Store n8n node, establishing the RAG knowledge base.


2. Order Tracking Path: The n8norder n8n trigger endpoint receives a request (e.g., order number and email). The workflow calls an external tracking API using an HTTP Request n8n node. The response is formatted by a Set n8n node and sent back immediately via a Webhook Respond n8n node.


3. Appointment Scheduling Path: The n8nappointment n8n trigger activates upon a scheduling request. A Langchain Chain LLM n8n node uses an LLM (gpt-4o-mini) and an Output Parser n8n node to reliably convert natural language dates into structured Google Calendar-compatible timestamps. This structured output is used by the Google Calendar n8n node to create the event, followed by a confirmation response sent back to the agent.


4. RAG Q&A Path: When the Voiceflow agent requires company knowledge, the n8n_rag n8n trigger is invoked. The core of this path is the Langchain Agent n8n node. This agent, configured with a detailed system prompt for customer support, uses its RAG Tool. This tool queries the Qdrant Vector Store to retrieve context, synthesizes the answer using another OpenAI n8n node, and sends the final, context-grounded response via the final webhook.

To deploy this expert n8n workflow, follow these steps:

1. Import: Copy the provided JSON data and import it into your n8n instance as a new workflow.


2. Credentials Setup:

1. Configuration:

Webhook (n8norder, n8nappointment, n8n_rag): Serve as the primary n8n trigger points for external requests originating from the Voiceflow agent. They allow the n8n workflow to run specific backend logic based on the agent's intent.
Langchain Agent n8n node (Retrive Agent): The central decision-making component for the RAG path. It uses a conversationalAgent pattern, backed by an OpenAI n8n node (LLM), and is configured with a detailed prompt ensuring professional, knowledge-base-driven customer support.
Langchain Vector Store Tool (RAG): Defined as a tool for the Agent, allowing the LLM to access external knowledge. It connects the conversational logic to the Qdrant vector database via the retriever n8n node.
Qdrant Vector Store n8n node: Used for both document ingestion (setup phase) and efficient vector search (RAG runtime), leveraging OpenAI embeddings.
Langchain Chain LLM (Concert start date): This specific n8n node is tasked with transforming user-provided, ambiguous date strings into the structured data needed for the Google Calendar n8n node, demonstrating advanced data processing within the n8n workflow.
Google Calendar n8n node: Executes the final action of scheduling an event, utilizing the precisely parsed date/time output generated by the preceding AI logic.

• API URL Tracking (HTTP Request): Performs transactional lookups by calling an external service (e.g., an e-commerce API) to retrieve real-time data like order status, a common requirement for comprehensive n8n templates.