AI Restaurant Assistant for WhatsApp, Instagram & Messenger

This n8n workflow creates an AI-powered restaurant assistant capable of responding to customer inquiries across WhatsApp, Instagram, and Messenger. It leverages a combination of AI agents, vector stores, and custom logic to provide intelligent and context-aware responses, including menu information, reservation details, and more.

What it does

This workflow is designed to handle incoming messages from various platforms and respond intelligently using AI. While the provided JSON only shows the building blocks, the intended functionality based on the node types and the directory name is as follows:

1. Receives incoming messages: It's triggered by a webhook, likely connected to a messaging platform (WhatsApp, Instagram, Messenger).
2. Processes messages with AI: It uses an AI Agent (Langchain) to understand the user's intent and generate a response.
3. Accesses information from a vector store: It can query a Supabase Vector Store to retrieve relevant information, such as menu items, opening hours, or special offers.
4. Utilizes tools for specific tasks: It can use tools like a Calculator (for price calculations or conversions) or an HTTP Request Tool (to fetch dynamic data, e.g., from a reservation system).
5. Manages conversation history: It stores and retrieves chat history using Postgres Chat Memory to maintain context in ongoing conversations.
6. Responds to the user: It sends back the AI-generated response to the originating messaging platform via the webhook.
7. Manages and processes documents: It can load documents, split them into chunks, and embed them into the vector store for RAG (Retrieval Augmented Generation). This is indicated by the "Google Drive Trigger", "Default Data Loader", "Recursive Character Text Splitter", "Embeddings OpenAI", and "Supabase Vector Store" nodes.
8. Includes conditional logic: Uses If and Switch nodes for branching logic based on message content or AI output.
9. Includes data transformation and manipulation: Uses Function, Edit Fields (Set), Loop Over Items (Split in Batches), Filter, Aggregate, Limit, Split Out, and Summarize nodes for various data handling tasks.

Prerequisites/Requirements

To use this workflow, you will need:

• n8n instance: A running n8n instance.
• OpenAI API Key: For the OpenAI Chat Model and Embeddings OpenAI nodes.
• Supabase Account: For the Supabase Vector Store and Supabase database.
• PostgreSQL Database: For the Postgres Chat Memory.
• Google Drive Account: If you intend to use the Google Drive Trigger for document loading.
• Messaging Platform Integration: Webhooks configured for WhatsApp, Instagram, or Messenger to send incoming messages to n8n.
• n8n Langchain Nodes: Ensure the @n8n/n8n-nodes-langchain package is installed in your n8n instance.

Setup/Usage

1. Import the Workflow:

   • Download the provided JSON file.
   • In your n8n instance, click "Workflows" in the left sidebar.
   • Click "New" and then "Import from JSON".
   • Paste the JSON content or upload the file.

2. Configure Credentials:

   • Locate the "OpenAI Chat Model" and "Embeddings OpenAI" nodes and configure your OpenAI API Key credential.
   • Locate the "Supabase Vector Store" and "Supabase" nodes and configure your Supabase credential (API URL and API Key).
   • Locate the "Postgres Chat Memory" node and configure your PostgreSQL database credential.
   • If using Google Drive, configure your Google Drive credential for the "Google Drive Trigger" and "Google Drive" nodes.

3. Configure Webhooks:

   • The "Webhook" node acts as the entry point for incoming messages. Copy its test or production URL.
   • Configure your messaging platform (WhatsApp, Instagram, Messenger) to send messages to this webhook URL. The exact setup will depend on the platform's API (e.g., Meta's Webhooks for Messenger/Instagram, a WhatsApp Business API provider).

4. Configure AI Agent and Tools:

   • AI Agent: Review the configuration of the "AI Agent" node to ensure it aligns with your desired conversational flow and agent type (e.g., "Conversational React Description").
   • Vector Store: Configure the "Supabase Vector Store" node with your Supabase project details (project URL, API key, table name for vectors, and embedding model).
   • Document Loading (Optional): If you want to ingest documents from Google Drive for your vector store, activate and configure the "Google Drive Trigger", "Default Data Loader", "Recursive Character Text Splitter", and "Embeddings OpenAI" nodes. This typically involves:
     • Setting up the Google Drive Trigger to monitor a specific folder.
     • Using the "Default Data Loader" to extract content from files.
     • Splitting the text with "Recursive Character Text Splitter" for optimal embedding.
     • Generating embeddings with "Embeddings OpenAI".
     • Storing these embeddings in "Supabase Vector Store".
   • Chat Memory: Ensure the "Postgres Chat Memory" is configured correctly to store and retrieve chat history, allowing the AI to maintain context.

5. Activate the Workflow:

   • Once all credentials and configurations are set, activate the workflow by toggling the "Active" switch in the top right corner of the n8n editor.

The workflow is now ready to receive messages and act as an AI restaurant assistant!