n8n-project

N8n sample project created to the I2A2 Autonomous Agents course

Installation

Requires:

• docker
• docker-compose

Use make init to install docker over colima.

• Set gpu_enabled=true to the .env to have gpu support - only works for linux and windows.
• For macos, you set native_mac=true to the .env to have gpu support.

You can also try make bootstrap to install dependencies, start the containers, and load the default model. Use same variables for gpu support.

Bootstrap

Start the project with make start or docker-compose up -d. Access the http://localhost:5678 and login with your credentials. The first time will require to sign for a free license.

• You can select "I'll not use n8n for work" to simplify the options.
• Then select to receive the license in the next page.
• Go in the settings > Usage and plan and put the active key you received in your email.

Next, you will need to config some connection at n8n credentials:

• Add connection for ollama:
  • Base URL: http://ollama:11434
• Add connection for postgres:
  • Host: pgvector
  • Username: n8n
  • Password: n8n-pgvector
  • Database: db

You can also play around ollama webui at http://localhost:3000.

• If not done yet, load a model using make load-model
• We default to gemma3, but you can load different models using make load-model model=<model_name>

The makefile

Run make help to see all available targets and options.

Targets:
	help                           This help.
	init                           Install dependencies and init container provider
	bootstrap                      install deps and start it with the default model
	start                          start local compose with essential services
	status                         show status of local compose
	stop                           stop local compose
	logs                           show logs of local compose
	load-model                     load model into local compose using model=<model> options, check https://ollama.com/library
	open-ollama                    open ollama in browser

Options:
	model                          Model to use for ollama, default: gemma3 (light); others: mistral, deepseek-r1 (big), llama4 (huge)
	cpu                            Number of CPUs to allocate for colima
	memory                         Memory to allocate for colima
	disk                           Disk space to allocate for colima
	gpu_enabled                    Enable ollama gpu support on linux and windows
	native_mac                     Enable ollama gpu support on macos

Development

Start the project with make start or docker-compose up -d. Access the http://localhost:5678 and login with your credentials.

You can also use gpu for better llm performance:

• make start gpu_enabled=true for windows and linux
• make start native_mac=true for macos

:note: Remember, the makefile commands are not mandatory, but keep them as a live-document reference for the commands you need to use. Feel free to use the direct docker/ollama/etc commands present in the project.

You can also use make commands to open the two available interface:

• make open-n8n Open the n8n UI
• make open-ollama Open the ollama UI

To load models, check for the makefile command:

• make load-model model=<model_name> You can use any model from https://ollama.com/library, or create your own custom models.

For the n8n workflows, please check for workflows/readme.md

Testing

For the n8n workflows, please check for workflows/readme.md

Deployment

TBD - we still need to define a provider, but we can refer to the official server-setups documentation.

Contributions

Feel free to ask for access to the project. You can also fork and open a pull request.