Configuration

archive needs no configuration to read public data. Everything below is optional: where it keeps state, and how it finds credentials when you upload, delete, or read a task queue.

Where state lives

archive config show

Path	Default	Override
Data root	~/data/archive	--data-dir, ARCHIVE_DATA_DIR
Downloads	<data>/download	download -d
Cache	<data>/cache	follows the data root
Config	~/.config/archive	XDG_CONFIG_HOME
Credentials	~/.config/archive/credentials

config show prints the resolved values plus the effective client settings (workers, rate, timeout, retries) and whether credentials are loaded.

Credentials

Authenticated commands (upload, delete, wayback save --outlinks/--screenshot, and tasks on items you do not own) use an IAS3 access/secret key pair from archive.org/account/s3.php.

Store them once:

archive configure --access YOUR_KEY --secret YOUR_SECRET

With no flags, configure prompts (the secret is read without echo). You can also log in with your account email and password to fetch the keys:

archive configure --email [email protected]

Credentials are written to ~/.config/archive/credentials with 0600 permissions. Check what is configured (the secret is masked):

archive whoami

Resolution order

For any command that needs credentials, archive resolves them in this order, first match wins:

1. --access / --secret flags.
2. ARCHIVE_ACCESS_KEY / ARCHIVE_SECRET_KEY environment variables.
3. IA_ACCESS_KEY / IA_SECRET_KEY (the names the ia Python tool uses).
4. The ~/.config/archive/credentials file.

This lets you keep a stored default and override it per command, or run fully from the environment in CI without writing a file.

Networking knobs

Flag	Default	What it does
--rate	250ms	minimum delay between requests
--retries	5	backoff retries on 429/5xx (honours Retry-After)
--timeout	2m	per-request timeout
-j, --workers	8	concurrent downloads

The Wayback CDX and replay endpoints are rate-limited hard by the Archive; if you hit 429s, raise --rate (e.g. --rate 2s) and --retries.

Caching

Metadata, search pages, and availability lookups are cached on disk under the cache directory, keyed by request with a short TTL, so repeated commands are instant and gentle on the Archive. Bypass it for one run with --no-cache, and manage it with the cache command:

archive cache info     # size and entry count
archive cache clear     # empty it