load-db

Load Database

Instructions

When this skill is invoked — either via /load-db or through a natural language request like "clone the live database from december 12" — download a database backup from Pantheon and load it into the local development environment.

Usage

/load-db                           # Interactive: choose environment and backup
/load-db live                      # Use latest backup from live environment
/load-db dev 2026-02-10            # Use backup from dev matching the given date
"clone the live database"          # Natural language: infer environment=live, pick latest
"load the db from test dec 12"     # Natural language: infer environment=test, date=dec 12
"pull down the production database"# Natural language: default to live, pick latest

Interpreting Natural Language

When invoked through conversation rather than a slash command:

• Look for an environment name in the request (live, dev, test, qa, training, or any multidev name). If none mentioned, proceed to environment selection interactively.
• Look for a date reference in the request (e.g., "december 12", "feb 3", "yesterday", "last tuesday"). Convert it to a date string like 2026-02-10 for matching.
• If the user says "latest" or "most recent", skip backup selection and use the newest available.
• If the user says "production", treat it as the live environment.

Steps

1. Parse Arguments and Context:

   • From slash command: first argument is environment, second is date
   • From natural language: extract environment and date from the user's message as described above
   • If both environment and date/latest are known, skip interactive selection steps

2. Select Environment:

   • If an environment was already determined, use it directly — skip interactive selection
   • Otherwise, list available environments using terminus:

     terminus env:list ${TERMINUS_SITE} --format=list --fields=id
     

   • Use the AskUserQuestion tool to let the user pick an environment from the list
   • Present the most commonly used environments first: live, test, dev, qa, then others
   • Include up to 3 options in the question, plus the user can type "Other" to specify a different environment
   • Recommend live as the default choice since it has production data

3. Download Backup:

   • Latest backup (no date specified): Skip the backup:list step entirely. Use terminus backup:get without --file to get the latest database backup URL directly, then download it in one command:

     curl -L "$(terminus backup:get ${TERMINUS_SITE}.{environment} --element=db)" --output /tmp/pantheon-db-backup.sql.gz
     

     This avoids fetching and parsing the full backup list JSON, which is large and wastes context tokens.
   • Specific date requested: Only in this case, list backups to find the matching one:

     terminus backup:list ${TERMINUS_SITE}.{environment} --element=db --format=json
     

     Filter to entries whose filename contains the target date string (e.g., 2026-02-10), then download:

     curl -L "$(terminus backup:get ${TERMINUS_SITE}.{environment} --element=db --file={backup_filename})" --output /tmp/pantheon-db-backup.sql.gz
     

   • Interactive (no environment or date): List backups using the same --element=db filtered command above and use AskUserQuestion to let the user pick:
     • Present backups sorted by date (newest first)
     • Show the filename and size for each option
     • Recommend the most recent backup
     • Include up to 3 most recent backups as options, with "Other" available for older backups
   • If no database backups exist for the environment, inform the user and stop
   • If the download fails, inform the user and stop
   • Verify the downloaded file exists and has a non-zero size

4. Import Database:

   • IMPORTANT: The db-rebuild.sh script prepends $(pwd)/ to positional file arguments (see line 80 of the script). You MUST either:
     • cd /tmp before invoking the script and pass the relative filename, OR
     • Pass the file using the -- separator to avoid the path prepending
   • Run the import:

     cd /tmp && /usr/local/bin/db-rebuild.sh pantheon-db-backup.sql.gz
     

   • This script will:
     • Drop and recreate the local database
     • Detect compression format and decompress
     • Import the SQL into the local database
     • Run drush deploy to apply config and migrations
   • Stream the output to the user so they can see progress
   • This command takes a long time to run — use a generous timeout (at least 10 minutes)

5. Clean Up:

   • Remove the downloaded backup file:

     rm -f /tmp/pantheon-db-backup.sql.gz
     

   • Inform the user that the database has been loaded successfully
   • Report which environment and backup were used

Error Handling

• If TERMINUS_SITE is not set, inform the user they need to set it (should be myeap2 for this project)
• If terminus authentication fails, suggest running terminus auth:login
• If no backups are found for the selected environment, inform the user and suggest trying a different environment or creating a backup first with terminus backup:create
• If the download or import fails, preserve error output and present it to the user