AI· 16 November 2026· 3 min read
Workflow Automation with Local AI: Deploying n8n and Qwen2.5:3b for On-Premise Systems
A detailed technical guide on combining n8n with the locally hosted Qwen2.5:3b model to build secure and intelligent automation workflows.

The Power of Local AI in Automation Infrastructure
As a systems integration engineer, I prioritize data security and privacy above all. Sending sensitive client data or ELV system logs to public AI clouds often carries significant risks. This is why Local AI is becoming an essential trend. In this article, I will guide you through deploying n8n — a powerful workflow automation tool — combined with the Qwen2.5:3b language model running entirely on your local server.
Qwen2.5:3b is a Large Language Model (LLM) from Alibaba Cloud, highly optimized for reasoning and coding tasks despite its 3-billion parameter size. Combined with n8n, we can create autonomous agents to process emails, categorize technical incidents, or summarize reports without needing an internet connection.
System Requirements
To ensure smooth operation, your server or PC should meet these minimum specifications:
CPU: Intel Core i5 or higher, or Apple Silicon (M1/M2/M3).
RAM: Minimum 8GB (16GB recommended to run both n8n and the LLM simultaneously).
OS: Ubuntu, macOS, or Windows (with Docker installed).
Step 1: Install Ollama to Run Qwen2.5
Ollama is currently the easiest tool for managing and running LLMs locally. It provides a standard OpenAI-compatible API, making it easy for n8n to connect.
On Linux or macOS, open your Terminal and run:
curl -fsSL https://ollama.com/install.sh | sh
Once installed, pull the Qwen2.5:3b model using this command:
ollama run qwen2.5:3b
This command downloads the model (approximately 1.9GB) and opens a chat window for testing. Once verified, type /exit. Ollama will continue running as a background service on port 11434.
Step 2: Deploy n8n via Docker
For efficient management, I always recommend using Docker. It isolates the environment and simplifies future updates. Use the following command to start n8n:
docker run -d --name n8n -p 5678:5678 -v ~/.n8n:/home/node/.n8n n8nio/n8n
After the container starts, access http://localhost:5678 to set up your n8n admin account.
Step 3: Connect n8n to Local LLM
This is the most critical step to give n8n its "brain". In the n8n interface, create a new Workflow and follow these steps:
Add "Ollama Chat Model" Node: Search for the Ollama node in the n8n library.
Configure Credentials: For the Host, if running n8n in Docker on the same machine as Ollama, enter http://host.docker.internal:11434 (for Windows/Mac) or the machine's LAN IP (for Linux).
Select Model: Enter the exact model name: qwen2.5:3b.
Add "AI Agent" Node: Connect the Ollama node to the input of the AI Agent node to begin processing text requests.
Practical Use Case: Automated System Log Classification
Imagine a network monitoring system sending alerts via Webhooks. Instead of reading hundreds of dry log lines, you can forward that data to n8n, asking Qwen2.5:3b to analyze the priority and only send a Telegram notification for critical failures.
With the Qwen2.5:3b model, response times are nearly instantaneous, allowing automation workflows to run smoothly without incurring per-request API costs.
Security and Optimization Notes
Even though it is running locally, ensure that Ollama's port 11434 is not publicly exposed to the internet without a security layer or VPN. Additionally, if you have an NVIDIA GPU, make sure the NVIDIA Container Toolkit is installed to significantly speed up AI processing compared to using only the CPU.
I hope this guide helps integration engineers add a powerful tool to their kit for optimizing daily operations.