restore-backup

# Restore Open Walnut from an S3 backup **Goal:** get the user's Walnut data back, verified, with zero chance of making things worse. You are doing the restore *for* the user. Be **proactive**: run the commands yourself, verify integrity yourself, and only ask when the answer is genuinely theirs (which bucket, which restore point, whether to adopt). ## Two safety rules (never break these) 1. **Restore NEVER overwrites live data.** Always download into a fresh folder first, verify it, and only then (with the user's explicit go-ahead) swap it into place. If `~/.open-walnut` still exists, treat it as precious even if the user says it's broken: rename it aside, never delete it. 2. **Never delete anything in the bucket.** The bucket is the last copy. Read-only access is all a restore needs. ## What's in the bucket (so you can reason about it) Under the configured prefix (default `walnut`): - `<prefix>/backup-manifest.json` — the file list of the last complete backup. Written LAST, so if it exists, that backup finished. With bucket versioning on, each old version of this one key is an earlier restore point. - `<prefix>/data/<relative path>` — every backed-up file, mirroring the data folder layout. - `<prefix>/data/.sqlite-snapshots/<relative path>` — safe online snapshots of the SQLite databases (sessions, tasks, usage). The restore command puts these back at their real paths automatically. Search/index databases are absent on purpose: Walnut rebuilds them...