Can't get my AI chatbot launched on API server

[JonTSG]

Describe the problem/error/question

I built an n8n workflow that successfully deployed an AI chatbot on my own laptop. I used Google Gemini to set up the API server and ngrok locally so that I could test it while building. I used Supabase to set up vector DBs for my “brain.” I used OpenAI for the chat model. Everything worked fully on the bot, but then I deployed a new API server on Fly.io and have NOT been able to get this thing working properly. I have encountered just about every error imaginable.

After struggling to get Gemini to debug the issues, I switched to ChatGPT5 because a friend recommended it was great at coding APIs.

I’m encountering similar issues with GPT5 and simply can’t get this thing to respond to me anymore.

I do not know enough about APIs to even give you an idea of what the issue is, other than to show you the last Windows PowerShell test I ran and the output it gave me.

PS C:\Users\sketc\OneDrive\Desktop\output\tsg-chatbot-api> Invoke-RestMethod `

-Uri “https://tsg-chatbot-api.fly.dev/qa-search” -Method Post
-ContentType “application/json” `
-Body ‘{“query”:“how do I stop slicing my driver?”}’

Invoke-RestMethod : {“error”:“different vector dimensions 384 and 1536”}
At line:1 char:1

• Invoke-RestMethod `

•   + CategoryInfo          : InvalidOperation: (System.Net.HttpWebRequest:HttpWebRequest) [Invoke-RestMethod], WebException
    + FullyQualifiedErrorId : WebCmdletWebResponseException,Microsoft.PowerShell.Commands.InvokeRestMethodCommand
  

PS C:\Users\sketc\OneDrive\Desktop\output\tsg-chatbot-api>

Honestly, this might be a huge ask for someone from the community to help, because it involves a lot of API debugging outside of n8n, but I could really use some help!!!

What is the error message (if any)?

Depends on the test, but there’s one above here.

Please share your workflow

Share the output returned by the last node

The tests I’ve been running haven’t even hit the n8n workflow so I don’t have any to show you at this time.

Information on your n8n setup

• n8n version: 1.106.3
• Database (default: SQLite):
• n8n EXECUTIONS_PROCESS setting (default: own, main):
• Running n8n via (Docker, npm, n8n cloud, desktop app): n8n cloud
• Operating system: win11

[solomon]

You should use a webhook to receive the requests in your n8n instance.

Activate your workflow and use the Webhook Production URL:

[JonTSG]

Thanks for the quick reply!

Would I need to adjust any other settings in this webhook node?

Also, how do I test this out and see if it’s working? Can you write me a quick Windows PowerShell command to run to test it?

[solomon]

If you use the GET method, you can open the URL in your browser and you’ll see the execution happening in n8n.

Browsers make GET requests to webpages you access.

You can either use the test URL while listening for a test request, or activate your workflow and use the production URL. If you use the production URL, the execution will only show up in the execution tab.

To choose the right settings for the webhook node, you should refer to the documentation of the service you’re using the make the request.

But n8n’s official webhook docs might also help.

[Bob_Brown]

Bruh!! when everything seems set up right but the chatbot won’t actually launch on the API server, it feels like hunting ghosts. Have you checked server logs or access control rules to see if requests are being blocked silently?

[JonTSG]

Oh man, sorry for the delay on replying!

I should have provided more details obviously.

I abandoned this project because we found Fin AI integrated with our customer service funnel pretty well.

I will probably re-visit some of this, because I discovered I can set up some webhooks with Fin AI to integrate with my Supabase tables I setup and am currently not using because I de-commissioned my manually built chatbot.

Also, to anyone who is a dummy with APIs like me, I recommend ChatGPT5 for setting up a Fly.io server! It helped me immensely with this stuff.