Restarting from scratch: where is the data kept?

Data is kept in the files described below. Each of these can be removed, but removing does not take effect until the server is restarted. Except for settings.db, the names of these resources may have been changed in the settings.

This keeps settings that do not have their default value. You can freely edit this file if the server is not running. Note that the settings can change the locations of the items below.
This keeps the registered users and their permissions. You can freely edit this file if the server is not running. Passwords are stored as MD5-hashes, hashed with crypt/2.
This directory holds backup of the RDF data. Each named-graph is represented by two (optional) files. The base of this file is an encoded version of the graph-name. The extensions are trp for quick-load snapshots and jrn for journal files. See the SWI-Prolog Semantic Web package for details. You can freely remove a graph by removing both files if the server is not running. Similar, you can remove the journal to restore a graph to its initial state after loading it from the source-data. Finally, rdf_flush_journals/1 can be used to merge the journals into the trp quick-load snapshots.
Keeps a log of HTTP interactions. It is enabled by linking or copying cliopatria('config-available/') into config-enabled. Note that on heavily loaded servers this can be a significant slow-down and the file can become huge.

On Unix systems, this file can be moved or removed at any time. The server will start a new log-file. On Windows this will give a sharing violation.