uptime
Monitor websites uptime using Cloudflare Workers
Star
Fork
Watch
51 10 1
Files
.github
apps
config
deploy
docs
packages
scripts
.dockerignore
.env.example
.gitignore
.prettierignore
.prettierrc.json
AGENTS.md
docker-compose.yml
Dockerfile
Makefile
package.json
pnpm-lock.yaml
pnpm-workspace.yaml
README.md
tokens.css
tsconfig.base.json
README.md

Multi-region uptime monitor #

Uptime monitoring on a Node API and scheduler backed by PostgreSQL, with checks from up to nine regions run by Cloudflare probe Workers. Includes latency charts, request history, optional DNS diagnostics, notification destinations, public monitor views, and grouped status pages.

Environment Architecture #

Production and staging run the same way. The API and the scheduler are Node 24 services on a VPS next to PostgreSQL 18, and a reverse proxy on the VPS serves the built web UI. Cloudflare only hosts the regional probe Workers and the test target Worker.

Free hosted version #

The first version includes AT Protocol OAuth signup, dedicated tenant schemas, team access, and an operator dashboard. Use the hosted runbook for local OAuth development and deployment. Admission is controlled by the operator admission_open toggle with fixed product limits (3 monitors, 300-second checks, 3 regions, 3 members, 1 status page). The public landing currently collects interest through identity-only AT Protocol OAuth. Signups appear in the operator page and do not allocate tenant slots. The product release plan covers positioning, user journeys, operations, launch gates, and the broader roadmap. The platform research records infrastructure limits, costs, and requirements for a later paid release.

Architecture #

  • The API layer (apps/api-worker, Node 24 and PostgreSQL 18 only) handles authentication, configuration, and private history.
  • The coordinator / scheduler (apps/coordinator-worker, Node 24) runs as a single minute dispatcher. It claims due checks and sends signed regional probe batches.
  • The scheduler publishes reports to R2 through the API's configured storage integration.
  • Regional probe Workers perform the target HTTP requests across worldwide PoPs.
  • The VPS reverse proxy serves the React UI built with pnpm --filter @uptime/web build.

Development and verification #

Requires Node.js 24+, pnpm 11.24.0, a PostgreSQL 18 database, and a Cloudflare account for deploying probes.

pnpm install --frozen-lockfile
pnpm hash:admin

Copy the required API values from .env.example into an ignored .env file and fill in local secrets. Build and start the API, then start the UI in a separate terminal:

pnpm --filter @uptime/api-worker build:node
pnpm --filter @uptime/api-worker start:node
pnpm --filter @uptime/web dev

Vite serves the UI on port 5176 and proxies /api to the local API on port 8787. Set UPTIME_API_PROXY_TARGET if using another port. Background checks also require the scheduler, secrets, and reachable regional probes; see the deployment runbook.

Compiling VPS Bundles #

# Build Node 24 API bundle (apps/api-worker/dist/node/server.mjs)
pnpm --filter @uptime/api-worker build:node

# Build Node 24 Scheduler bundle (apps/coordinator-worker/dist/node/scheduler.mjs)
pnpm --filter @uptime/coordinator-worker build:node

Checks and Tests #

pnpm typecheck
pnpm test
pnpm build
scripts/tests/deploy-workers.sh
pnpm generate:wrangler -- --check

Deployment #

Validate the probe bundles without deploying:

scripts/deploy-workers.sh --dry-run

Operational behavior #

  • Region placement hints express affinity, not a guaranteed city or source IP.
  • Requests require an HMAC signature and a current timestamp.
  • Detailed observations default to seven days of retention; daily aggregates remain available. DNS diagnostics default to 30 days.
  • Public monitor reports include the available retained observation history. Closed latency history uses persisted hourly aggregates; the rolling 24-hour window (up to 25 hours from the floored hour) and partial oldest hours at the 7-day and 30-day boundaries use raw observations. Each aggregate keeps exact latency-value frequencies and 15-minute summaries, so percentiles and chart detail remain available.
  • Uptime charts use cached daily bars for closed days and aggregate the current day's observations separately.
  • HTTP notification providers are supported; SMTP delivery is unsupported in this application.
  • Monitor intentionally public targets. Literal-address and redirect checks do not provide DNS-to-socket pinning or replace origin access controls.
  • Staging budget policy: Budget enforcement is intentionally disabled on staging VPS (BUDGET_ENFORCEMENT_DISABLED=true). Production defaults remain strictly enforced.

See architecture for the runtime responsibilities, deploy/vps/staging/README.md for staging VPS operations, and operations for general configuration.