Hi @kjooleng
Still not working now.
Hi @lannyho
These are the things you need to take note.
Your n8n workflow isn’t working because the AI software (Ollama) and the automation tool (n8n) aren’t communicating correctly. It is essentially a “connection” problem; n8n is trying to send a message, but it’s using the wrong name and hitting a digital wall on your server.
First, there is a simple naming mistake. You told n8n to use a model called “llama3.2,” but your server lists it as “llama3.2:latest.” To a computer, these are two completely different names. You simply need to add the “:latest” part to the model name in your n8n settings so they match perfectly.
Next, the AI software is currently in “private mode.” By default, Ollama only talks to things happening inside its own immediate bubble. You need to change a hidden setting on your Windows Server called an environment variable to tell the AI to stop being private and start listening for requests from other apps.
Even if the AI is willing to talk, your Windows Server has a built-in security guard called a Firewall. This guard blocks almost all incoming traffic to keep the server safe. You need to create a specific “allow” rule that tells the guard to let messages through on Port 11434, which is the specific door the AI uses to communicate.
Additionally, you need to make sure n8n knows exactly where the server is located. If you used the word “localhost” in your settings, n8n is looking for the AI inside itself rather than on the server. You must replace “localhost” with the actual IP address of your Windows Server so n8n knows exactly which machine to call.
Once you fix the name, tell the AI to listen, open the firewall door, and provide the correct server address, the connection should be established. Try these steps one by one, and your AI agent should start responding to your “hello” messages in the chat.
