Hono + Inertia + React Deployment Guide — Cloudflare Workers / Vercel / Bun: Choosing Your Stack and CI/CD [2026]

Because Hono runs everywhere, picking a deploy target is the first big decision. Three axes: - Runtime: Workers / Vercel Functions / Node / Bun - Location: edge (everywhere) or specific regions (Tokyo, Osaka, etc.) - Ops overhead: managed (Workers / Vercel), DIY VPS, or containers (ECS / Cloud Run) Combining these three narrows down to 3–5 plausible options quickly.

- Global SaaS / marketing-leaning sites: Cloudflare Workers - Domestic Japan business systems / internal tools: Vercel (Tokyo region) or Fly.io (Tokyo). Fly.io / Render / Railway are nice when you want Postgres alongside - AWS-heavy enterprise: ECS Fargate / Cloud Run - Small-scale, single-box: Bun on a VPS (Sakura, Vultr, Hetzner) - Enterprise SLA / audit needs: Vercel Enterprise or a dedicated cloud For Japan-domestic apps, "Workers + Tokyo Postgres via Hyperdrive" is increasingly viable in production.

Inertia supports SSR, but you don't always need it. - SEO-critical pages: SSR recommended - Behind login walls only: CSR is fine - First-paint speed matters most: SSR tends to win SSR puts a React renderer on your Hono server, raising build/operation overhead. Better to start without it and turn it on per-page when needed.

Minimum workflow on GitHub Actions: 1. trigger: push to `main` and PR 2. install: `bun install` 3. lint / typecheck: `bun run lint`, `bun run typecheck` 4. test: `bun test` 5. build: `bun run build` (client + server) 6. deploy: - Cloudflare Workers: `wrangler deploy` - Vercel: `vercel --prod` (or Vercel Git integration auto-deploys) - Fly.io: `flyctl deploy` Production deploys only on `main`; PRs go to preview environments.

Secret handling per target: - Cloudflare Workers: `wrangler secret put`; locally `.dev.vars` - Vercel: dashboard env vars; or `vercel env` - Fly.io: `flyctl secrets set` - VPS: `.env` encrypted via sops / age, or Vault / AWS Secrets Manager Common rules: never commit secrets, use distinct keys per env, define rotation up front.

Bare-minimum production observability: - Errors: Sentry on both client and server - App logs: Logpush / Logtail on Workers, Vercel Logs on Vercel; otherwise route to Datadog / Better Stack / Loki - Uptime: Better Stack / UptimeRobot - APM: Datadog APM or Cloudflare Workers Analytics, as needed - DB slow query log: especially valuable for business apps Mirroring summaries to a Slack/Discord channel pairs nicely with code-review culture.

- Workers size limits: Node-native modules in your bundle blow up — use the `nodejs_compat` flag or alternative libraries - Cookie / session stores: in-memory stores aren't usable on Workers — use KV / D1 / external Redis - File uploads: send large files directly to R2 / S3, not through the server - Cron: Workers Cron Triggers / Vercel Cron / external schedulers — pick per case - Timezones: explicitly set `Asia/Tokyo` for JP audiences and align server, DB, client

We mix Cloudflare Workers / Vercel / Fly.io / managed VPS based on project shape. For mid-sized internal-tool / business-system work, "Hono + Inertia + React on Vercel or Fly.io + Postgres" tends to hit the cost / ops / speed sweet spot. When AI features are involved, we combine with AI Consulting or OpenClaw for hybrid setups that take advantage of edge runtimes.

This concludes the series: - Part 1: stack overview & why - Part 2: setup guide - Part 3: forms & CRUD - Part 4: authentication - Part 5 (this post): deployment End-to-end, this is a working full-stack alternative to Next.js / Remix. To talk about applying it to your project or to scope a PoC, see Web Development / Software Development / Contact.

Q1: Workers or Vercel — what's the first cut? A: Workers for global + cost; Vercel for DX, polish, and team familiarity. For Japan-domestic apps, Vercel Tokyo or Fly.io Tokyo are the easy picks. Q2: Multi-region DB? A: Workers + Hyperdrive can proxy to external Postgres; Neon / Turso / D1 are also options. CRUD-heavy business apps often do fine on a single Tokyo region. Q3: Rollback? A: Both Vercel and Workers support instant rollback to the prior deploy. Just decide up front how DB migrations interact with rollback. Q4: Do we need a staging environment? A: Strongly recommended for business systems. Combine per-PR preview environments with a longer-lived staging environment for pre-release verification.

- Cloudflare Workers docs - Vercel docs - Fly.io docs - Hono — Cloudflare Workers guide - Inertia.js — Server-side Setup