I think the answer depends on a lot of things.

First, do you have a complex frontend where using a framework like React, Vue, or Svelte would be helpful? If so, you’ll need to decide whether to keep your backend and frontend separate (e.g., Phoenix + Next.js/Nuxt/SvelteKit/Astro) or use a monolithic Phoenix setup.

If you’re sticking with a Phoenix monolith and still want to use a frontend framework, the current options that support SSR (server-side rendering of JS components) that I know of are LiveSvelte, LiveVue, and Inertia.js.

Now, for choosing between LiveSvelte/Vue and Inertia.js: I haven’t used LiveVue, so I can only speak about LiveSvelte and Inertia.js. LiveSvelte is specific to Svelte, while Inertia.js is framework agnostic and can be used with React, Vue, or Svelte. LiveSvelte integrates closely with LiveView, which means it includes all of LiveView’s features, like automatic socket connections and rendering based on socket state changes.

In contrast, Inertia.js doesn’t integrate with LiveView. Instead, it hands off the entire frontend to your chosen frontend framework. It provides SPA-like routing similar to LiveView’s live navigation but without using a socket. For real-time features, you would need to set up your own socket and channels.

I find LiveSvelte offers tighter integration, making the developer experience simpler overall. However, there are some known SSR performance issues and specific challenges (related to making LiveView work in an offline app) that led me to try Inertia.js.

The Elixir Inertia.js adapter has experimental SSR support, and while it sometimes crashes in development, it falls back to non-SSR if it fails. I haven’t encountered SSR issues in non-dev environments. Not integrating with LiveView also avoids some of the offline-related problems I faced.

Both have their pros and cons. I really like LiveSvelte, but Inertia.js solves the specific problems I have better at the moment. That said, I don’t think the issues I have with LiveView and offline functionality are unsolvable—I just haven’t dug into them further since I wanted to try out Inertia.js. Inertia.js isn’t perfect either, but so far, everything seems to just work where it needs to in my current project.

Sorry for the long-winded response! I hope this helps answer your question!

Hi @jam — I actually just hacked together a hackathon app this week using SvelteKit + Phoenix Channels, and I think I like it more than using LiveSvelte for PWAs that need to work offline. Specifically, I used SvelteKit in SPA mode. During development, I set up a Vite proxy to talk to Phoenix. In production, the SPA is compiled into Phoenix’s static folder, so I only need to run the Phoenix server — no separate SvelteKit server required.

There are some tradeoffs with this approach, though. Mainly, there’s no SSR unless you prerender pages. But for a PWA with a service worker caching files for offline use, that’s not a big deal — and having the SvelteKit router is great.

You can make LiveSvelte work as an offline PWA, but there are a couple of downsides:

1. Lack of client-side routing: Liveview has live navigation, but that only works when you’re online. If you want offline routing without page reload, you’ll need to build your own or use a third-party library, neither of which will likely be as full-featured as SvelteKit’s.
2. CSRF token issues: For offline apps, you’ll need to cache your HTML, which ends up caching the CSRF token used for the LiveView socket connection. This makes reconnecting after being offline for a while tricky with LiveSvelte. I had a very hacky workaround in a previous app, but I wouldn’t recommend it.

In summary, I wouldn’t recommend LiveSvelte for PWAs that need offline capabilities. It’s possible, but right now I don’t think it’s as clean as just using channels.

Naive question: why would I pick inertia over liveview?
what would be the rationale?