Files
docker-compose-cookbooks/invokeai
codeandClaude Fable 5 8ab413804a normalize all stacks to conventions
- VOL_PATH always defaults to /data; vscode/opencode VOL_PROJECTS nests it
- TZ defaults to America/Vancouver (was UTC in etherpad, ntfy, redis)
- compose section order fixed (volumes before environment, command after networks)
- all default secrets are the literal ChangeThisPassword (grep-friendly,
  users must change them); removed a real key from ollama sample.env
- sample.env synced with compose: missing vars added with defaults, dead vars removed
- invoiceninja_dev/openarchiver_dev: app services had no environment block,
  wired all sample.env vars through so the stacks actually work
- opencode: sample.env OLLAMA_HOST renamed to OLLAMA_HOST_URL to match compose
- TZ values unquoted consistently

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-16 06:24:54 +00:00
..
2026-05-04 19:15:58 +00:00

Invokeai

Overview

Makes images, needs gpu.

Project Details

  • Project Repository: Github
  • Container Image: ghcr.io/invoke-ai/invokeai
  • Documentation: Docs
  • Reverse Proxy Port: You Can set this in env, I have it set to 9090

Additional Notes / Gotchas

To get started, go to the models tab and install: RealVisXL V3 or Juggernaut XL Also in the model manager, for upscale install: RealESRGAN_x4plus And last install Tile SDXL version

Dockhand Stack, Deploy from Git

Cookbooks Repository stackname: invokeai Compose file path: invokeai/compose.yaml Additional env file (optional): invokeai/sample.env

Then "Load" SERVICENAME/sample.env into the Environmental variables in dockhand

Create the Stack

Docker Run

docker run -d \
  --name invokeai \
  --restart unless-stopped \
  -v "/data/invokeai:/invokeai" \
  -e INVOKEAI_ROOT=/invokeai \
  -e INVOKEAI_PORT=9090 \
  -p 9090:9090 \
  --gpus all \
  ghcr.io/invoke-ai/invokeai:latest