archipelago
576c642da4
Demo images / Build & push demo images (push) Failing after 38s
Ollama's embedded manifest failed the typed parse (memory_limit wants a string) so the catalog overlay was skipped for it; AdGuard Home's conventional :3000 collided with Grafana's. The release gate now runs the host-port collision test (repo_app_manifests_have_no_host_port_collisions) so this class can never ship untested again.
64 lines
1.5 KiB
YAML
64 lines
1.5 KiB
YAML
app:
|
|
id: ollama
|
|
name: Ollama
|
|
version: 0.5.4
|
|
upstream:
|
|
kind: github
|
|
repo: ollama/ollama
|
|
description: >-
|
|
Run large language models locally. Download and run AI models like
|
|
Llama, Mistral on your own hardware — served on the node's loopback for
|
|
the AI assistant (Settings → Claude Auth → model backend), never exposed
|
|
to the network.
|
|
|
|
container:
|
|
image: source.archipelago-foundation.org/lfg2025/ollama:latest
|
|
pull_policy: if-not-present
|
|
network: pasta
|
|
|
|
dependencies:
|
|
- storage: 50Gi
|
|
|
|
resources:
|
|
# No memory limit: models are sized by the disk allowance below, and a
|
|
# RAM ceiling would just OOM-kill long inferences.
|
|
disk_limit: 50Gi
|
|
|
|
security:
|
|
capabilities: []
|
|
readonly_root: false
|
|
no_new_privileges: true
|
|
network_policy: isolated
|
|
|
|
ports:
|
|
- host: 11434
|
|
container: 11434
|
|
protocol: tcp
|
|
# local: Ollama's REST API is consumed by the node's own assistant over
|
|
# loopback — never externally reachable, so no gate, no TLS, and no
|
|
# login surface exist at all.
|
|
bind: 127.0.0.1
|
|
auth: local
|
|
|
|
volumes:
|
|
- type: bind
|
|
source: /var/lib/archipelago/ollama
|
|
target: /root/.ollama
|
|
options: [rw]
|
|
|
|
environment: []
|
|
|
|
health_check:
|
|
type: tcp
|
|
endpoint: localhost:11434
|
|
interval: 30s
|
|
timeout: 5s
|
|
retries: 3
|
|
|
|
metadata:
|
|
author: Ollama
|
|
category: community
|
|
icon: /assets/img/app-icons/ollama.png
|
|
repo: https://github.com/ollama/ollama
|
|
tier: optional
|