AMPLE

SAFE REDEPLOYMENT

How to redeploy after changing an AI-built app

Release source or configuration changes with Ample, understand unchanged no-op deploys, and inspect failures before retrying.

By Ample ComputerUpdated July 30, 2026

DIRECT ANSWER

The short version

After changing an AI-built app, run the same Ample deploy workflow again. Ample hashes the packaged files, build configuration, and environment, then compares them with the live release. A real change creates a new release. No change returns unchanged: true and the current live URL without rebuilding.

WHO THIS IS FOR

Before you start

This guide is for builders shipping a second or later version of an app that is already live on Ample.

STEP BY STEP

How to do it

  1. 1

    Make and verify the change

    Ask the coding agent to run the smallest relevant local check before deployment. Commit the change if that is part of your repo workflow.

  2. 2

    Reuse the recorded plan

    If the repo has ample.toml, keep it aligned with any service, database, volume, domain, or environment changes.

  3. 3

    Deploy once

    For a single service, run ample deploy. For a planned multi-service topology, run ample up so dependencies are applied in order.

    ample deploy
  4. 4

    Read unchanged correctly

    An unchanged result is a successful no-op. It means the live release already matches the packaged content and deployment configuration.

  5. 5

    Force only for intentional reactivation

    Use force when unchanged source must be rebuilt and reactivated, such as recovering a live release that is misbehaving at runtime.

    ample deploy --force

FIT CHECK

What this supports

Good fit

  • Content-addressed release comparison
  • No-op results for unchanged live source
  • Saved deployment targeting and configuration
  • Intentional forced rebuilds

Know the limits

  • A blind retry does not fix a failed release
  • Force rebuilds unchanged source and should be intentional
  • Infrastructure changes should be reflected in the plan

TROUBLESHOOTING

Common failure modes

The deploy says unchanged but you expected new code

Confirm the edited file is inside the deploy root and is not excluded from packaging.

The new release fails

Read its build or runtime logs, fix the cause, and deploy the changed source.

A prior start command disappeared

Check the saved config or ample.toml and pin the intended start command when detection is not sufficient.

KEEP READING