Vane unhandledRejection: TypeError: fetch failed

[jdmoser1]

Describe the bug
Hello. I'm spinning up Vane for the first time and I can't get vane to communicate with a model in Ollama. When I enter a query, I get a "brainstorming" spinning circle with no progress.

I was able to enter the URL for my local Ollama server and Vane is able to get a list of models and select.

I disabled KV cache quantization and it made no difference.

(Edit to add) I tested Ollama directly and had a successful result:

podman exec ollama ollama run llama3.2 "paw patrol haiku"

To Reproduce
Spin up using the compose file below

Expected behavior
A successful response with a web search result

Screenshots

Additional context
vane-compose.yml

services:
  ollama:
    image: docker.io/ollama/ollama:latest
    container_name: ollama
    ports:
      - "11434:11434"
    env_file: 
      - ./ollama.env
    volumes:
      - ./ollama-data:/root/.ollama
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]
    restart: unless-stopped

  vane:
    image: docker.io/itzcrazykns1337/vane:latest
    container_name: vane
    ports:
      - "48081:3000"
    volumes:
      - ./vane-data:/home/vane/data
      - ./vane-uploads:/home/vane/uploads

ollama.env

OLLAMA_HOST=0.0.0.0:11434
#OLLAMA_FLASH_ATTENTION=1
#OLLAMA_KV_CACHE_TYPE=q8_0

[robogeek]

In data/config.json how is your Ollama connection set up?

For this Compose file, the URL to connect with Ollama would be http://ollama:11434. Therefore, in Settings / Models, in the dialog box that appears after clicking "Add Connection", that's the URL to enter for the connection.

However .. I recommend a different setup.

First, run:

docker network create ollama

In ~/docker/ollama have a compose file with the Ollama portion of the system. In this compose file

services:
  ollama:
    # ...
    networks:
      - ollama
    # ...
      
networks:
  ollama:
    external: true

Then, in ~/docker/vane/compose.yaml

services:
  ollama:
    # ...
    networks:
      - ollama
    # ...
      
networks:
  ollama:
    external: true

This way Ollama is a general service - e.g. for use with Open Web UI or with OpenCode - that can be brought up and down separately.

Any service that connects to Ollama is connecting over a private Docker network as http://ollama:11434

Blog post: https://techsparx.com/homelab/services/ollama.html