fintl

Financial ETL CLI: parse, normalize, and explore your bank transaction data.

TL;DR

This tool helps you process, visualize and search your balance and transaction information that you have exported from your bank accounts.

Currently supports DKB, Postbank, GLS and Scalable Capital Broker.

Supported file formats: CSV, HTML, and PNG. PNG parsing uses a local ollama / llama-swap instance with a multimodal model — opt-in via fintl.toml (required only for Scalable broker PNG statements; gracefully skipped when not configured).

All your data stays on your machine. No need to trust another entity that is PSD2 certified.

How to install

As a cli tool

uv tool install fintl

As a dependency

From pypi

pip install fintl

or

uv add fintl

From a local repo

git clone https://github.com/eschmidt42/fintl.git
cd fintl
uv sync
uv tool install .

After installation, fintl should be available on your PATH:

which fintl
# e.g. /Users/YOURUSER/.local/bin/fintl

How to use

1. Configure your ~/.config/petprojects/fintl.toml. For details see here.
2. Go to your bank account.
3. Select your service, e.g. Giro.
4. Export csv file or similar to ~/Downloads, or directly your source dir for your bank / service.
5. Optionally, if you've stored your file in ~/Downloads, run cd ~/Downloads followed by fintl store (uses your fintl.toml from step 1).
6. Optionally, if you want to process PNG screenshots via Ollama, start Ollama.
7. Run the etl via fintl etl (also uses your fintl.toml from step 1).
8. Upon success visualize / search your data via fintl plot or fintl search.

Please see here and below for more usage details.

Repo structure

• src/fintl/accounts_etl/ — core ETL logic: schemas, parsers, registry, runner
• src/fintl/cli/ — CLI entry point and subcommands (etl, store, search, plot)
• tests/ — tests for packages of this repo

Development

Run tests:

uv run pytest -n auto tests

Type check:

uv run ty check src

Lint, format, type check, test and all the other good stuff:

prek run --all-files

Running extraction integration tests

Tests that require a live llama-swap / ollama service are marked ollama / llama_swap and skipped by default. To run them, set FINTL_OLLAMA_MODEL to the name of a multimodal model you have pulled (e.g. qwen3.6:latest) and pass -m ollama or -m llama_swap:

FINTL_OLLAMA_MODEL=qwen3.6:latest uv run pytest -m ollama tests/
FINTL_LLAMA_SWAP_MODEL=qwen-3.6-27b uv run pytest -m llama_swap tests/

FINTL_OLLAMA_BASE_URL can optionally override the default http://localhost:11434.

FINTL_LLAMA_SWAP_BASE_URL can optionally override the default http://0.0.0.0:8080.

Simulating CLI usage without real financial data

Set the FINTL_CONFIG env var to point at any TOML file instead of ~/.config/petprojects/fintl.toml. A dev-config.toml is included at the repo root and wires all CLI commands to the fixture files under tests/.

To try the CLI commands, run:

mkdir /tmp/fintl-dev

Then you can run one of

rm -fr /tmp/fintl-dev/*
FINTL_CONFIG=dev-config.toml fintl etl
FINTL_CONFIG=dev-config.toml fintl plot
FINTL_CONFIG=dev-config.toml fintl store --from-dir <dir> --yes --copy

Output lands in /tmp/fintl-dev. Edit dev-config.toml locally to change providers, paths, or target directory.

Note: FINTL_CONFIG=dev-config.toml fintl etl is expected to produce one warning, failing to parse Screenshot 2026-03-09 at 14.30.54.png, because it in fact is an insufficient png file to parse the requested information from.

Release

For release steps see here.