Creating a new project

BiblioFetch.generate(path) drops a ready-to-edit project skeleton at path — the matching config/template/ directory inside the package is the single source of truth for what that skeleton contains. It's the one-liner version of "copy the config, write a job.toml, remember what fields go where" — instead you run:

bibliofetch init ~/my-bibliography

or from Julia:

julia> BiblioFetch.generate("~/my-bibliography")

Either form produces:

~/my-bibliography/
  job.toml       # edit the [doi].list, optionally tune [fetch] and [graph]
  README.md      # 30-line per-project how-to

After the skeleton is in place:

bibliofetch run ~/my-bibliography/job.toml

runs the job. Relative [folder].target values in the skeleton resolve against the project directory (see PR #37 / load_job), so the PDFs land right next to the job.toml regardless of the shell's cwd.

Guard rails

generate refuses to populate a non-empty directory unless force=true. That protects you from overwriting an in-progress project by typo, while still letting you re-seed a clean copy on top of old leftovers when you really mean it:

BiblioFetch.generate("existing-dir"; force=true)

Passing --force on the CLI does the same thing:

bibliofetch init existing-dir --force

Trying it from Julia

The example below creates a throwaway project under a tempdir, just so the docs build has a self-contained runnable demonstration.

using BiblioFetch

mktempdir() do parent
    dest = joinpath(parent, "demo-bibliography")
    BiblioFetch.generate(dest)
    (isdir(dest), sort(readdir(dest)))
end

(true, ["README.md", "job.toml"])

Customising the skeleton

The config/template/ directory inside the installed package is read verbatim — no placeholder substitution happens, so what's on disk over there is exactly what the user gets. Extending the skeleton is a one-file PR: drop a new file into config/template/, optionally update template/README.md to mention it, and every subsequent generate() call picks it up. No code changes required.

The onboarding loop end-to-end

# 1. Copy the machine-wide config template (if you haven't already)
cp "$(julia -e 'using BiblioFetch; print(pkgdir(BiblioFetch))')/config/config.toml" \
   ~/.config/bibliofetch/config.toml
# → edit email / store_root / any per-host [profiles.*]

# 2. Create a per-project skeleton
bibliofetch init ~/research/dmrg-bibliography
cd ~/research/dmrg-bibliography
# → edit job.toml: add DOIs / arxiv ids to [doi].list

# 3. Run
bibliofetch run job.toml

That's the full onboarding story — three shell commands and two edit sessions from zero to a first PDF in hand.