Configuration

There is no project file. Every command is configured by its flags, and the database is a connection string passed with --database or read from DATABASE_URL. That is the whole configuration surface.

Not supported yet. A besu.config.* file, config discovery and per-environment profiles do not exist.

#The connection string

--database takes a URL. The scheme chooses the backend, and the rest is handed to that backend's pool:

besu migrate --database sqlite:app.db
besu migrate --database sqlite://app.db
besu migrate --database postgres://user:pass@host/db
besu migrate --database postgresql://user:pass@host/db
besu migrate --database mysql://user:pass@host/db

An unknown scheme fails before connecting:

unsupported database scheme `mssql`; expected sqlite, postgres or mysql

TLS, socket paths and pool parameters are part of the URL, because the driver passes it straight to sqlx. There are no CLI flags for them.

#Environment variables

--database is the only option with an environment fallback, and the only variable is DATABASE_URL:

export DATABASE_URL=postgres://user:pass@host/db

besu pull --out schema.json
besu generate --schema schema.json --name add_email
besu migrate --dir migrations

An explicit flag beats the variable. In --help the variable is shown with its current value, so [env: DATABASE_URL=] means it is unset.

besu studio reads it the same way, so a shell that already has DATABASE_URL set needs no --database at all.

#Defaults

There is no flag for a snapshot path on migrate, no input directory for generate (it diffs against the database), and no way to point the CLI at a named environment.

#What is not configurable

• The dialect. It comes from the URL scheme; there is no override.
• The journal table name. It is always _besu_migrations.
• The statement splitter. It is fixed, and its limits are described in custom migrations.
• Logging. The CLI installs no subscriber, so there is no RUST_LOG behaviour to set; besu studio does install one.
• Migration naming. generate uses NNNN_name.sql and does not consult a template.

#Gotchas

• A missing DATABASE_URL is a clap error, not a runtime error. The command prints its usage and exits before any connection attempt.
• DATABASE_URL is shared with everything else in the environment. A shell that exports a production URL will use it for any besu command that omits --database.
• A SQLite typo creates a database. sqlite:app.db opens or creates the file, so a wrong path silently produces an empty database.
• Defaults are relative to the working directory. --dir migrations and --out migrations mean the directory below wherever you invoked the binary, not the crate root.