Using Ollama vision models

Gulianrdgd · September 2, 2025, 12:56pm

It would help if there was a node for:

Ollama vision models

My use case:

Currently, image analysis is only available through the AI vendor nodes (Gemini, Anthropic, OpenAI). However, Ollama supports vision models that can analyze images. I’d like to use local Ollama vision models for image analysis.

Any resources to support this?

github.com/ollama/ollama

docs/api.md

main

# API

## Endpoints

- [Generate a completion](#generate-a-completion)
- [Generate a chat completion](#generate-a-chat-completion)
- [Create a Model](#create-a-model)
- [List Local Models](#list-local-models)
- [Show Model Information](#show-model-information)
- [Copy a Model](#copy-a-model)
- [Delete a Model](#delete-a-model)
- [Pull a Model](#pull-a-model)
- [Push a Model](#push-a-model)
- [Generate Embeddings](#generate-embeddings)
- [List Running Models](#list-running-models)
- [Version](#version)

## Conventions

### Model names

This file has been truncated. show original

Are you willing to work on this?

Yes I have actually worked on this but I need some help before continuing

My question:

I’ve implemented this by adding Ollama as a new AI vendor (similar to Gemini/Anthropic/OpenAI), which works for vision tasks.

However, I’m questioning this approach because Ollama is not really like the other vendors. The vendor interface also feels empty because Ollama does not have that many capabilities. The other vendors have features such as image generation, which Ollama does not (yet)

Should I continue with Ollama as a separate AI vendor? Or should I change the AI agent node to allow image upload when using an Ollama chat model?

Here is my current implementation:

I have included screenshots.

Stuart_Halliwell · September 12, 2025, 9:50am

I’d love to have this as I’m trying to setup an image classifier for all my photography work.

Ideally I’d have n8n running a local file search on my server, picking up on the images from multiple folders, doing some local AI classification and image keyword generation, slapping the whole lot into a vector DB (again all local) and potentially sorting into folders by key classification and highlighting potential duplicates.

Looking to use the llava:7b model, Qdrant, n8n and openwebui/ollama all on the homelab to do this.

Got the tools and the back end, just not the nodes it seems…

Keeping an eye on this request :).

S.

jan · October 13, 2025, 1:55pm

New version [email protected] got released which includes the GitHub PR 19371.