Full Universal Deploy

Goal

Deep & zero-config integration between (Vite) apps and deployment providers:

• Deep: support advanced features like SPA fallback, AI integrations, ...
• Zero-config: the deployment integration works out-of-the-box with no/minimal config.

This proposal is about extending Universal Deploy to support more common deployment features.

Note

Brainstorming: quick-wins & vision.

This isn't a fixed plan — it's brainstorming to develop quick-win opportunities for Universal Deploy as well as to develop the Universal Deploy vision.

Static deployments

Support for:

• URL rewrites:
  • Use case: SPA fallback (e.g. serve dist/client/product/index.html for route /proudct/:id i.e. URLs /product/42, /product/1337, ...)
  • Use case: 404 page (serve dist/client/404/index.html for route /* as a fallback)
• URL redirects

CLI

CLI integration:

• $ vite deploy logs => server logs
• $ vite deploy status => deployment status, and downtime status for past 30 days
• $ vite deploy ls => e.g. number of workers, system metrics (CPU usage, mem usage)
• $ vite deploy db => connect to DB and run queries, e.g. SELECT { title, release_date } FROM movies;

Note

Why? AI.

AI cannot (practically) work with UIs — the CLI enables AI to access deployment information (e.g. deployment logs for debugging).

Tasks

How to support tasks (cron jobs and queues) in a way that works for any deployment provider?

Is there a standard syntax for defining tasks & queues? The UNIX cron syntax?

Environment Variables

How can we best manage/support secrets?

Note

What are secrets?

Secrets are environment variables that hold senstive information such as DB_PASSWORD.

The goal here is that is that secrets are supported by the deployment provider without the user having to configure anything — the Vite-based framework and deployment provider talk to each other directly through Universal Deploy.

See also:

• SecretSpec

We can first focus on simple use cases for small teams, while maybe trying to be flexible for enterprises that have a rigid process for managing secrets.

AI integration

How will deployment integration with AI workflows look like?

Example: error detected in deployment => AI is automatically prompted => AI can access deployment logs and source code => AI proposes fix => human review => fix is merged & deploy — the only human intervention here is reviewing the bug fix.