After a workstation failure, restore the smallest set that can safely recreate and operate the site: code, dependency locks, configuration, databases, media, deployment scripts, and required AI assets. An AI backup is not a special backup category; it is a backup that also preserves AI-related files, settings, prompts, data, and integrations. The correct restore order depends on what the workstation handled. A developer may need local source code and tooling, while a WordPress or Drupal administrator may also need database exports, uploaded files, configuration secrets, and deployment access.
Table of Contents
- Restore the Site's Operating Core First
- Which Files Matter for WordPress and Drupal?
- What AI-Related Assets Should Be Included?
- What Should Not Be Restored Blindly?
- How Can You Rebuild and Verify the Replacement Workstation?
- Frequently Asked Questions
Restore the Site's Operating Core First
Restore these items before rebuilding optional tools or downloading cached files: The database and uploaded files deserve separate attention. Code can often be recovered from version control, but content, form submissions, user accounts, settings, and media may exist only in database or storage backups.
Do not assume a complete workstation image is the best restore source. It may contain obsolete dependencies, temporary files, cached credentials, or damaged data. A clean system plus verified project backups often produces a safer result.
- Source code, themes, plugins, modules, and custom scripts
- Dependency manifests and lock files, such as `package-lock.json`, `yarn.lock`, `composer.lock`, or equivalent files
- Database backups, including WordPress or Drupal content and configuration data
- Uploaded media, document libraries, and other user-generated files
- Deployment scripts, infrastructure definitions, and build configuration
Which Files Matter for WordPress and Drupal?
For WordPress, prioritize the database, `wp-content/uploads`, custom themes, custom plugins, and any code that does not exist in the official WordPress repository. The `wp-config.php` file contains important connection settings, but its credentials should come from a secure vault or deployment system when possible. For drupal, restore the database, the `sites/default/files` directory, custom modules, custom themes, and the project's Composer files.
The `settings.php` file may contain database credentials, hash salts, and other sensitive values, so treat it as secret material rather than an ordinary source file. A backup of the application directory alone may not restore the site. For example, restoring a Drupal codebase without its database can leave the site without content, users, permissions, and active configuration. Restoring a WordPress database without its media directory can produce broken images and missing downloads.
What AI-Related Assets Should Be Included?
Preserve AI assets that affect repeatability, quality, cost, or compliance. Depending on the project, that may include prompt templates, system instructions, model settings, tool definitions, retrieval configuration, evaluation examples, approved reference documents, and generated assets selected for publication. Also record the model provider, model identifier, API parameters, preprocessing rules, and post-processing steps. These details help explain why a workflow produced a particular result, even when the original model or service later changes.
Vector indexes and embeddings require special handling. They can be large and may be reproducible from the original documents, but rebuilding them can change results if the embedding model, chunking rules, or source files have changed. keep the index when fast recovery matters, and keep the source documents and build instructions when long-term reproducibility matters. Do not casually back up private customer data, unpublished strategy documents, API keys, or personal information inside AI logs. Store sensitive material in an encrypted system with controlled access, and retain only the logs needed for troubleshooting, auditing, or approved reuse.
What Should Not Be Restored Blindly?
Some workstation files consume space without helping recovery. Reinstall or regenerate them unless the project specifically depends on them: Credentials need a separate decision. Restore access through a password manager, secrets manager, SSH-key process, or hosting provider rather than copying every credential from an old disk.
Rotate keys that may have been exposed during the failure, especially API keys used by advertising, analytics, hosting, payment, or AI services. Review backup contents for secrets before sharing an archive with a contractor or uploading it to a new storage service. A `.env` file, database dump, debug log, or browser profile can reveal more access than expected.
- Dependency directories such as `node_modules` or vendor caches
- Operating-system caches and temporary files
- Build outputs that can be recreated from source
- Browser profiles containing unrelated personal data
- Duplicate exports and unverified database snapshots
How Can You Rebuild and Verify the Replacement Workstation?
Use a restore sequence that proves each layer works before depending on the next: Test the restored site before changing production. Check authentication, permissions, scheduled tasks, contact forms, payment or advertising integrations, analytics, search indexing controls, and backups themselves.
The strongest backup is one that has been restored successfully. A file timestamp or successful upload proves that data exists; it does not prove that the site, CMS, deployment process, or AI workflow can use it. For example, a prompt archive without its referenced documents may be readable but unable to reproduce the workflow.
- Install the supported runtime, package manager, database client, and version-control tools.
- Clone the project and install dependencies from its lock files.
- Recreate environment variables from a documented template and secure secret store.
- Import a database copy into an isolated environment.
- Restore media and confirm that application paths match the imported records.
Frequently Asked Questions
Should I back up the entire workstation?
A full image can help with forensic recovery, but it should supplement separate, documented backups of the project, CMS data, media, and secrets.
Do AI prompts alone provide a recoverable backup?
No. Prompts may depend on model settings, source documents, tools, retrieval indexes, credentials, and application code.
Can I restore production directly onto a replacement computer?
Rebuild and test in an isolated environment first, then deploy through the normal controlled process.
