diff --git a/website/docs/reference/cli-commands.md b/website/docs/reference/cli-commands.md index 869c3cde6c..ff2c9a1d46 100644 --- a/website/docs/reference/cli-commands.md +++ b/website/docs/reference/cli-commands.md @@ -1106,6 +1106,7 @@ The backup uses SQLite's `backup()` API for safe copying, so it works correctly - `*.db-wal`, `*.db-shm`, `*.db-journal` — SQLite's WAL / shared-memory / journal sidecars. The `*.db` file already got a consistent snapshot via `sqlite3.backup()`; shipping the live sidecars alongside it would let a restore see a half-committed state. - `checkpoints/` — per-session trajectory caches. Hash-keyed and regenerated per session; wouldn't port cleanly to another install anyway. - `models/`, `runtimes/`, `node/` at the root of `~/.hermes` (and of each `profiles//`) — regenerable runtime downloads, often tens of GB. Deeper directories with the same names (a skill's `models/`) are kept. +- Browser profiles: `browser-profile/` (the real-profile snapshot — copied Cookies / Login Data), `browser-profiles/` (live CDP profiles) at any depth, and `browser_profiles/` (the Browser Use CLI backend's Chromium user-data dir, with its own Login Data / Cookies) at the root of `~/.hermes` and of each `profiles//`. Credential stores that must never enter an archive; all are regenerated on the next launch. - Regenerable entries of `cache/` at those same roots — model/plugin catalogs, stamps, browser profiles, tool-output spill. Durable artifacts stay in: `cache/images`, `cache/audio`, `cache/videos`, `cache/documents`, `cache/screenshots` (media delivered to or received from you) and `cache/citations` (the grounded-citations ledger). A deeper `cache/` (inside a skill) is kept whole. - Unix sockets, devices, and symlinks — a zip cannot hold them; before they were excluded, a stray `gateway.sock` made every full backup report `Backup incomplete`. - The `hermes-agent` code itself (this is a user-data backup, not a repo snapshot).