Hi!
I’m trying to set up a local Ollama-based AI agent in n8n (version 1.114.0) without using any external API keys.
Environment:
n8n version: 1.114.0
Ollama running locally via docker-compose
No external AI services (OpenAI, Anthropic etc.)
HTTP Request with Bearer token works fine
Data is loaded into NocoDB successfully
The issue appears only when using Ollama Chat Model or LLM Chain
Problem:
When I remove the “API Key” field from the LLM nodes (because Ollama is local and doesn’t require one), n8n throws a connection error:
“The resource you are requesting could not be found” (404)
Or sometimes Ollama returns “Ollama running” on a GET request, and POST is rejected
As a result, the Ollama Chat Model node cannot connect to any upstream chain, and LLM Chain node breaks completely
It looks like:
The LLM node in 1.114.0 forces authentication logic, even for local Ollama
Or it uses an incorrect endpoint (GET instead of POST)
Or something in the internal LLM proxy is not compatible with Ollama running inside Docker container
Question:
What is the correct way to configure the Ollama Chat Model (or LLM Chain) in n8n 1.114.0 so it works with a local Ollama instance without requiring an API key?
Do I need:
a specific base URL format?
a custom header?
a separate LLM agent configuration?
a workaround for the 404 / GET–POST mismatch?
Any official guidance or fixes for Ollama ↔️︎ n8n integration would be greatly appreciated.
