Build a Support Chatbot with AcuityScheduling Knowledgebase & n8n

1. Opening Problem Statement

Meet Sarah, the customer support manager at AcuityScheduling.com, a SaaS platform for scheduling and appointments. Every day, Sarah’s team fields dozens of repetitive questions like how to connect iCloud with AcuityScheduling or how to download invoices. These queries take up valuable time, leading to slow response rates, frustrated users, and higher operational costs.

Sarah tried traditional chatbots but found their responses generic or outdated, often leading customers to browse lengthy FAQs or escalate to human agents. Maintaining a dedicated vector store of all documentation seemed overwhelming and costly.

This is where the n8n ‘AcuityScheduling Support Chatbot’ workflow steps in — a smart, automated assistant that taps directly into AcuityScheduling’s existing support portal search API, powered by OpenAI’s GPT-4o model. It offers instant, accurate answers tailored to AcuityScheduling users, saving Sarah dozens of hours weekly and improving user satisfaction dramatically.

2. What This Automation Does

This workflow acts as a conversational AI agent that seamlessly integrates with AcuityScheduling’s knowledgebase. Here’s what happens when it runs:

• Receives Chat Queries: The workflow triggers whenever a chat message is received from users.
• Uses OpenAI GPT-4o Mini: It processes and understands user questions leveraging the GPT-4o mini chat model for natural language comprehension.
• Searches Knowledgebase: Instead of maintaining a separate data store, it directly queries AcuityScheduling’s support portal search API to find relevant articles.
• Formats Retrieved Data: The results returned are cleaned and formatted to suit chatbot responses, saving token costs and improving clarity.
• Returns Aggregated Responses: Multiple relevant articles are combined and presented back to the user in an easy-to-understand manner along with direct links to detailed support documentation.
• Maintains Conversation Context: It uses a simple memory buffer to remember prior user interactions within the same chat session for more coherent conversations.

By using this workflow, Sarah’s team can reduce manual query handling by up to 80%, cut down customer wait time by at least 50%, and avoid the costs and complexity of vector store maintenance.

3. Prerequisites ⚙️

• n8n Account: Make sure you have access to an n8n instance where you can import and deploy workflows.
• OpenAI Account:** 🔑 Obtain API credentials for OpenAI to use the GPT-4o mini model.
• AcuityScheduling Support Portal API: No special credentials needed as this example uses a public search API endpoint from AcuityScheduling’s help center.
• Optional Customization Tools: Access to modify HTTP Request, Set, and Aggregate nodes for data formatting.
• Self-Hosting Option: If you prefer complete control, consider self-hosting your n8n workflow. Learn how at Buldrr powered by Hostinger.

4. Step-by-Step Guide

Step 1: Import the Provided n8n Workflow

First, log in to your n8n editor. Navigate to Workflows → Import. Upload the JSON file of this workflow. You will see the nodes laid out similarly to the screenshot in the workflow editor.

Expected outcome: All nodes including the chat trigger and OpenAI model appear connected.

Common mistake: Forgetting to import the entire JSON payload, resulting in missing nodes.

Step 2: Configure the Chat Message Trigger Node

Click the node named “When chat message received”. This is a Langchain chat trigger node that waits for incoming chat messages from your front-end or chat interface.

No changes required here if you use default webhook URLs. Test invocation can be done through your chat interface connected to n8n.

Visual: The webhook ID is visible in the node’s parameters.

Expected outcome: Workflow triggers every time a chat message comes in.

Common mistake: Using wrong webhook URLs or missing to connect your front-end chat to this trigger.

Step 3: Set Up the OpenAI Chat Model Node

Open the OpenAI Chat Model node. Choose the gpt-4o-mini model from the dropdown. Add your OpenAI API credentials under the credentials tab labeled OpenAi account.

This node interprets and generates natural language responses.

Expected outcome: OpenAI API calls succeed, and the model outputs chat completions.

Common mistake: Incorrect API keys or missing credential configuration causing auth errors.

Step 4: Configure Simple Memory Node

The Simple Memory node is a Langchain memory buffer window that stores conversation context. This helps maintain continuity in dialogues.

No additional parameters are usually necessary here.

Expected: Conversation context is remembered across messages.

Common issue: Memory loss due to misconfiguration or version incompatibilities.

Step 5: Add the Knowledgebase Tool Node

This node uses a custom tool workflow that connects your AI agent to the AcuityScheduling support search API for retrieving knowledgebase articles related to the user’s query.

Verify that the “workflowId” parameter points to your KnowledgeBase Tool Subworkflow node in this workflow.

Expected outcome: Queries are routed to search API dynamically.

Common mistake: Mismatching workflow references breaks retrieval.

Step 6: Inspect the KnowledgeBase Tool Subworkflow

This subworkflow actually performs the HTTP Request to the AcuityScheduling search API:

• POST to https://2al21hjwoz-dsn.algolia.net/1/indexes/*/queries...
• Body contains the query variable from user input
• Headers are set to imitate browser requests for proper API access
• Results limited to top 5 hits and filtered to English language only

Expected: API returns JSON with relevant support articles.

Common trouble: API key or headers outdated leading to 403 errors.

Step 7: Process API Results with ‘Has Results?’ Node

The Has Results? node checks if any articles were found. This is a conditional node that routes the workflow either to processing results or returning an empty response.

Expected: Workflow branches correctly based on results presence.

Common mistake: Improper condition setup preventing correct routing.

Step 8: Split and Extract Relevant Article Fields

The Results to Items node splits the returned hits array into individual items for easier processing.

Then the Extract Relevant Fields node maps each article’s title, body, and constructs a direct URL link.

Example assignment for URL: https://help.acuityscheduling.com/hc/en-us/articles/{{id}}

Expected: Clean, formatted data ready for the chatbot response.

Common issue: Incorrect JSON path causing empty fields.

Step 9: Aggregate and Format the Response

The Aggregate Response node combines all the processed items back into a single response payload that the AI agent can present.

This reduces token usage and optimizes chat delivery.

Expected: Neatly formatted summary sent to user.

Step 10: Connect Workflow and Test

Ensure all nodes are properly connected as per the original JSON. Activate the workflow from the top right corner.

Test by sending sample queries like “How do I connect my iCloud to AcuityScheduling?” and verify the response links to relevant support articles.

Common mistake: Not activating the workflow or skipping test queries.

5. Customizations ✏️

• Use Alternative LLM Models: In the OpenAI Chat Model node, switch from gpt-4o-mini to other models like GPT-4 or GPT-3.5 Turbo. This changes answer quality and costs.
• Add More Knowledgebase Sources: Expand the Knowledgebase Tool Subworkflow to include other company support portals by editing the HTTP Request node URL and payload.
• Change Language Filters: Modify the facetFilters parameter in HTTP Request to support other locales besides “en-us” for multilingual support.
• Improve Memory Retention: Adjust the Simple Memory node’s window size or use a different Langchain memory node type for more complex conversation state management.
• Customize Response Formatting: Edit the Extract Relevant Fields and Aggregate Response nodes to tailor the answer snippets and links based on branding or markup preferences.

6. Troubleshooting 🔧

Problem: “OpenAI API authentication failed”

Cause: Invalid or missing API key in OpenAI credential setup.

Solution: Go to Credentials → OpenAi account in n8n, re-enter a valid API key from your OpenAI account dashboard. Test connection by running the OpenAI Chat Model node manually.

Problem: “HTTP 403 Forbidden from Acuity Support Search API”

Cause: Headers missing or changed in the HTTP Request node, or API endpoint is deprecated.

Solution: Verify the headerParameters of the HTTP node match the live browser request headers for the support site. Update API keys if necessary.

Problem: “No search results returned”

Cause: The query string sent to the API is empty or incorrectly formatted.

Solution: Inspect the HTTP node’s POST body JSON. Confirm the query parameter is passed correctly from your chatbot input.

7. Pre-Production Checklist ✅

• Verify OpenAI API credentials are correct by testing the OpenAI Chat Model node.
• Confirm the HTTP Request node returns valid results by temporarily logging output of the Acuity Support Search API node.
• Check all workflow connections are intact and node parameter references are accurate.
• Run test queries through the chat interface and confirm relevant, formatted responses appear.
• Backup your workflow JSON before deploying to production for rollback options.

8. Deployment Guide

To deploy, activate the workflow in your n8n editor by clicking the Activate toggle button at the top right.

Integrate the chat trigger webhook URL into your chat interface or front-end to start sending user messages to this bot.

Monitor usage and errors via the n8n execution logs. Adjust memory size or API call frequencies if needed to optimize performance.

9. FAQs

Can I use Google Bard or other LLMs instead of OpenAI’s GPT?

Yes, if the alternative LLM has function calling support similar to the Langchain nodes, you can replace the OpenAI Chat Model node. Just update credentials and test thoroughly.

Does this workflow use a vector store or costly database?

No, this workflow cleverly avoids vector stores by querying the Acuity Scheduling support search API directly, saving time and expenses.

Is my data safe?

All queries and results flow through your n8n instance and OpenAI’s API over HTTPS, ensuring secure communication. Always secure your API keys and webhook URLs.

Can this handle high volumes of queries?

Yes, depending on your n8n hosting and OpenAI usage plan, it can scale, but monitor API limits.

10. Conclusion

By following this guide, you have successfully created a powerful, focused support chatbot tailored specifically for AcuityScheduling.com users. This workflow leverages existing support portal search rather than costly vector databases, saving you time, money, and maintenance headaches.

Sarah and her team can now provide faster, accurate help to customers consistently, reducing manual workload by up to 80% and improving customer happiness significantly.

Next, consider extending the bot to other knowledgebases, integrating with customer ticketing systems, or adding multi-language support to broaden your users’ reach.

Keep experimenting and improving — your support automation journey just got a huge boost!