sync-docker-image-cli

Command line client for the Docker image sync web service. Trigger a sync to your own registry, then watch it finish — without leaving the terminal.

Zero runtime dependencies. Node 18+.

Install

npm install -g sync-docker-image-cli

Log in

Generate a token in the web UI at https://<your-app>/tokens, then:

sdi login https://your-app.vercel.app

Paste the token at the prompt — it is not echoed, so it stays out of your shell history. For non-interactive logins, pipe it instead of passing --token:

cat token.txt | sdi login https://your-app.vercel.app --token-stdin

The token is stored in ~/.config/sync-docker-image/config.json (mode 0600). For CI, skip login and export environment variables instead:

export SDI_URL=https://your-app.vercel.app
export SDI_TOKEN=sdi_xxxxxxxx

Use

# Sync one tag. The destination defaults to the registry/namespace
# configured on the server.
sdi copy nginx:1.27

# Block until it finishes, showing which GitHub Actions step is running
sdi copy ghcr.io/owner/app:v1 --wait

# Choose the destination explicitly
sdi copy nvcr.io/nvidia/pytorch:24.05-py3 team/pytorch:24.05 --wait

# Sync every tag of a repository
sdi sync nginx team

# Wait, then pull the result locally
sdi copy redis:7 --wait --pull

# Inspect jobs
sdi list
sdi list --status failed
sdi status <job-id> --wait
sdi rm <job-id>

Scripting and agents

Every command accepts --json, which writes a single JSON document to stdout and sends all human-readable output to stderr:

IMAGE=$(sdi copy redis:7 --wait --json | jq -r '.destination')
docker pull "$IMAGE"

Exit codes are stable:

| Code | Meaning | | ---- | -------------------------------- | | 0 | success | | 1 | network, server or auth error | | 2 | bad arguments | | 3 | the sync itself failed | | 4 | --wait timed out (still running) |

sdi copy nginx:1.27 --wait --json > result.json
case $? in
  0) echo "synced: $(jq -r .destination result.json)" ;;
  3) echo "failed: $(jq -r .job.logs_url result.json)" ;;
  4) echo "still running, check later" ;;
  *) echo "error" ;;
esac

--wait polls every 5s and gives up after 30 minutes; tune with --interval and --timeout (both in seconds). The interval widens automatically as the wait goes on (10s after a minute, 30s after five) so long waits do not exhaust the server's GitHub API quota; passing --interval pins it. A failed poll is retried — only five consecutive failures abort the wait.

Commands

| Command | Description | | ------------------------ | -------------------------------------------- | | sdi login [url] | Save the web app URL and an API token | | sdi logout | Forget saved credentials | | sdi whoami | Show who the current token belongs to | | sdi config | Show the resolved configuration | | sdi copy <src> [dest] | Sync a single tag | | sdi sync <src> [dest] | Sync every tag of a repository | | sdi status <job-id> | Job status and step-by-step progress | | sdi list | List recent jobs | | sdi rm <job-id>... | Delete jobs |

Run sdi help for the full flag reference.

Environment

| Variable | Purpose | | ---------------- | ------------------------------------------ | | SDI_URL | Web app URL (overrides the config file) | | SDI_TOKEN | API token (overrides the config file) | | SDI_CONFIG_DIR | Directory holding config.json | | NO_COLOR | Disable coloured output |

License

MIT