Keep Me Alive

A small Go binary that periodically checks a configured list of websites — some run locally via docker compose, some remote — restarts local ones that go down, and notifies a Signal group via signal-cli-rest-api when a site goes down or recovers.

Features

• Local sites (checked over HTTP, restarted with a configurable shell command) and remote sites (checked only).
• Signal notifications via signal-cli-rest-api, behind HTTP Basic Auth.
• Notifications fire only on state transitions (healthy→down, down→healthy) — no repeat spam while an outage continues. Local-site restarts are still retried every cycle while a site stays down.
• Check history persisted to a local SQLite database, viewable live via a status TUI.
• Four run modes: one-time check, daemon, verbose daemon (debug), and a live status view.

Installation

Requires Go 1.25+ (or GOTOOLCHAIN=auto, the default, to fetch it automatically).

git clone https://github.com/LycheeOrg/Keep-Me-Alive.git
cd Keep-Me-Alive
go build -o keep-me-alive .

Prebuilt binaries for Linux and macOS (amd64/arm64) are attached to each GitHub release.

Configuration

Configuration lives in a YAML file (default config.yaml, override with -c/--config).

Field	Type	Description
check_interval	duration	Daemon-mode poll interval (e.g. 60s)
restart_recheck_delay	duration	Wait after restarting a local site before rechecking it
http_timeout	duration	Per-request timeout for site checks and Signal calls
state_db_path	string	SQLite file for check history, read by -s
history_size	int	Max recent checks retained per site (ring buffer)
signal.base_url	string	URL of the signal-cli-rest-api instance
signal.username / signal.password	string	HTTP Basic Auth credentials for the Signal API
signal.sender_number	string	Registered sender number
signal.recipients	[]string	Recipient numbers or group.<id> group IDs
sites[].name	string	Unique site name
sites[].type	local | remote	Whether the site is restarted locally
sites[].url	string	URL checked with a GET request; 2xx = up
sites[].restart_command	string	(local only) Shell command run via sh -c to restart the site
sites[].working_dir	string	(local only) Directory the restart command runs in

Example (see config.yaml):

check_interval: 60s
restart_recheck_delay: 15s
http_timeout: 10s
state_db_path: "keep-me-alive.db"
history_size: 50

signal:
  base_url: "http://signal-host:8080"
  username: "proxyuser"
  password: "proxypass"
  sender_number: "+15551234567"
  recipients:
    - "group.abcdef123456=="

sites:
  - name: "lychee"
    type: local
    url: "http://localhost:8081/api/health"
    restart_command: "docker compose restart lychee"
    working_dir: "/opt/lychee"

  - name: "example-remote"
    type: remote
    url: "https://example.com/"

Usage

keep-me-alive -o             # one-time check, prints a results table, exits
keep-me-alive -d             # daemon: runs forever at check_interval, quiet logging
keep-me-alive                # same as -d, but with verbose/debug logging
keep-me-alive -s             # live status TUI, one line per host
keep-me-alive -c other.yaml  # use a different config file (any mode)

-o, -d, and -s are mutually exclusive.

Down-handling behavior

• Remote site down → a Signal notification is sent immediately (once, on the down transition).
• Local site down → the configured restart_command is run in working_dir, then after restart_recheck_delay the site is rechecked. If it's still down, a Signal notification is sent. If it recovered, nothing is sent (just logged). While a local site remains down across daemon cycles, restarts keep being retried every cycle, but the notification only fires once per outage.
• Recovery → a Signal notification is sent once when a previously-down site comes back up.

One-time mode (-o) has no state carried over between invocations — each run is independent.

Status view (-s)

Reads state_db_path and shows one line per configured site: current status, last-checked time, latency, and a sparkline of recent checks. It doesn't perform any checks itself, so it works alongside a running daemon or after one has stopped, live-refreshing every 2 seconds. Press q to quit.

Development

make build       # go build -o keep-me-alive .
make vet         # go vet ./...
make fmt-check   # gofmt -l . (fails if anything is unformatted)
make test        # go test ./...
make test-race   # go test -race ./...
make all         # build + vet + fmt-check + test

Tests are self-contained (using httptest servers and temporary SQLite files) and don't require Docker or a real Signal server.