fauxbooks GitHub

Synthetic financial data, at scale,
that reconciles.

A deterministic CLI that generates fully reconciled, double-entry financial datasets. Give your long-running AI agents realistic books to work on, and a ground truth to verify their own output against.

$

reconcilable

Σ debits = Σ credits in every entry, a ground truth your agents verify against.

5,910.10 = 5,910.10

deterministic

Same seed, byte-identical books on every OS. Reproducible evals.

sha256: run 1 == run 2

at scale

A quick fixture or a full year of books, on demand. csv · jsonl · xlsx.

--volume small · medium · large

Examples

One run writes a complete double-entry close of 19 CSVs and a manifest.

gl.csv Σ debits = Σ credits ✓

Real rows from a seeded run. Click any file. Decimals are exact strings (9 fractional digits), never floats. One run writes 19 CSVs plus manifest.json (seeds, row counts, SHA-256 per file).


Reconcilable books,
in one command.

$ go install github.com/humzakhan/fauxbooks/cmd/fbx@latest