Skip to content

Overview

Jump to bottom
github-actions[bot] edited this page Mar 17, 2026 · 2 revisions

Spendify — Your unified bank statement, on your computer

Informational document — basis for web page / landing page

The problem everyone knows but nobody solves

You have three accounts: a current account, a credit card and a savings account. Every month you download three files from the bank, open them in Excel, try to paste them together — and every time you get lost among duplicates, dates, amounts with random signs, and the certainty that something doesn't add up.

Then there is the problem of all problems: the card charge on the current account. The supermarket purchase appears both in the card statement (as an individual transaction) and in the current account (as a monthly aggregated debit). Adding everything up, your expenses look twice as high as they really are.

Spendify solves exactly this.

What Spendify is

Spendify is a personal financial register that automatically aggregates bank statements from different banks into a single chronological ledger, without duplicates, without sign errors, without monthly subscriptions and without sending your data to anyone.

It runs on your computer. Your data stays on your computer.

How it works in three steps

1. Download the files from your bank

Export statements as CSV or XLSX from your bank's portal. No special steps required — Spendify automatically recognises the format.

2. Drag them into Spendify

Select all the files together (even from different banks, even different years) and click Process. Spendify:

  • Automatically detects what type of document it is (current account, credit card, prepaid card, savings account)
  • Corrects the sign of amounts if needed (some banks export expenses as positive numbers)
  • Eliminates double-counting between card and account
  • Classifies each transaction with a category label

3. See where your money goes

The unified ledger shows you everything in one place. With charts, filters, export, and the certainty that every euro is counted only once.

The things no one else does automatically

Card–current account reconciliation

When your credit card charges the monthly amount to the current account, Spendify recognises the relationship and removes the double-counting automatically. The algorithm uses a time window of ±45 days and three matching phases (sliding window + subset sum for fractional amounts).

Internal transfer detection

A bank transfer from your current account to your savings account is neither an expense nor an income: it is an internal transfer. Spendify recognises it by comparing amounts, dates and account holder names in the descriptions — even if the two files were imported at different times.

Idempotent deduplication

Each transaction has a unique code calculated on its content (SHA-256). If you import the same file twice, nothing happens. You can re-import your entire history without worry.

Hybrid classification

Categorisation uses a four-level cascade system:

  1. Your rules — patterns you have defined yourself (e.g.: "CONAD" → Food / Supermarkets)
  2. Static regex — predefined patterns for the most common categories
  3. LLM — for everything else, the language model assigns a category and subcategory
  4. Fallback — if everything else fails, the transaction is placed in "To review"

Privacy: your data stays yours

Offline-first mode

By default, Spendify uses Ollama locally: an AI engine that runs on your computer, without an internet connection. Your bank statements never leave your disk.

If you want to use OpenAI or Claude

You can, but Spendify first automatically removes all identifying data:

  • IBAN → <ACCOUNT_ID>
  • Card numbers → <CARD_ID>
  • Tax identification number → <FISCAL_ID>
  • Your name → a fictional name

Only after sanitisation is the text sent. If for any reason the check fails, the call is blocked — not silently degraded.

No cloud, no subscription

The data is in a SQLite database on your computer. You can copy it, move it, back it up like any other file.

Who Spendify is for

For those with multiple accounts at different banks

If you use more than one bank account (current account + credit card + savings account + trading account) you know how difficult it is to have a unified view. Spendify does exactly this — without having to do anything manually.

For those who take their finances seriously

If you use Excel to track expenses, Spendify can replace that routine: you import the files once, Spendify unifies and classifies them, you check and correct only the exceptions.

For those who do not trust the cloud

Spendify has no mandatory remote backends, does not require an account, and sends nothing to anyone by default. Your banking data stays where it belongs.

For developers

Spendify is an open source Python project with a modular architecture and a complete test suite. It is an interesting starting point for those who want to:

  • Experiment with LLM pipelines on structured data
  • Build integrations with specific banks
  • Extend the data model or taxonomy
  • Deploy on a server for family or small team use

Main features
Multi-bank import CSV and XLSX from any Italian bank (and beyond)
Auto-detect format No manual configuration for each file type
Unified ledger All transactions in chronological order, filterable by date / account / category / context
Automatic classification 15 expense categories + 7 income categories with subcategories
Customisable taxonomy Add / modify categories and subcategories without restarting the app
Deterministic rules Create "ESSELUNGA always → Food" rules applied retroactively
Interactive analytics 7 Plotly charts: monthly trend, cumulative balance, expense pie chart by category, subcategory drill-down, top 10 merchants
Life contexts A dimension orthogonal to category: segment expenses by Work / Holiday / Daily life
Check List Month × account pivot table: see at a glance which months you have not yet imported
Export Standalone HTML (with charts), CSV, XLSX
Configurable LLM Local Ollama, OpenAI, Claude, Groq, Google AI Studio, LM Studio, any compatible API
PII sanitisation Automatic IBAN / PAN / tax ID / name protection before any remote call

Technology stack

Spendify is built on mature technologies with no exotic dependencies:

  • Python 3.13 + pandas for the data pipeline
  • Streamlit for the web interface (accessible from the browser, without installing anything special)
  • SQLite + SQLAlchemy for persistence (a single, portable file)
  • Pydantic v2 for schema validation
  • Plotly for charts
  • uv as package manager (installation in ~30 seconds)

No LLM framework (no LangChain) — the AI backends use the official SDKs directly, with a common abstract interface.

Installation in one command

The only prerequisite is Docker Desktop.

Mac / Linux:

curl -fsSL https://raw.githubusercontent.com/drake69/spendify/main/install.sh | bash

Windows (PowerShell):

irm https://raw.githubusercontent.com/drake69/spendify/main/install.ps1 | iex

The script downloads the image from GitHub Container Registry, starts the container and opens the browser at http://localhost:8501 automatically.

Go to Settings, enter your name, add your accounts — and import.

For native installation (Mac with Ollama) or Linux/Windows server with Docker+local LLM → Full guide.

For developers: what's inside

Modular pipeline

CSV/XLSX file
    → Schema classifier  (column SHA-256 fingerprint → LLM if new schema)
    → Normaliser          (encoding, parse_amount Decimal, SHA-256 tx_id)
    → RF-03 Reconciler    (card–account, 3 phases)
    → RF-04 Detector      (internal transfers, symbolic matching)
    → Description cleaner (LLM, noise → canonical text)
    → Categoriser         (rules → regex → LLM → fallback)
    → Database            (SQLAlchemy, idempotent)

Extensibility

  • New LLM backends: implement LLMBackend (3 methods) and register it in BackendFactory
  • New bank formats: Flow 2 recognises them automatically via LLM without code changes; the schema is saved and reused in subsequent imports
  • New categories: from the Taxonomy page, without touching the code
  • REST API: the process_file() pipeline is completely separate from the UI — it can be exposed via FastAPI without modifications

Test suite

uv run pytest tests/ -v          # 184 tests, zero external dependencies
uv run pytest tests/ --cov=core  # coverage of the business logic layer

All tests use SQLite in-memory: no files, no external services, no LLM mock required for the base suite.

Roadmap

  • PDF support (native parsing of bank PDF statements)
  • Monthly budget per category with alerting
  • Automatic import via Open Banking (PSD2)
  • REST API for external integrations
  • Multi-user version with authentication
  • Mobile app (ledger and analytics viewing)

Contributing

Spendify is open source. The areas where contributions are most useful:

  • New bank formats — if your bank is not recognised automatically, you can open an issue with an anonymised sample
  • Tests — the suite covers the business logic layer but not yet the UI
  • Internationalisation — the architecture already supports multiple languages for descriptions; the UI is in Italian
  • Performance — batch categorisation is the bottleneck with a local LLM; there is room for parallelisation

Spendify is not a cloud service. It is a program that runs on your computer. Your banking data is never sent to third-party servers, unless you explicitly choose a remote LLM backend — in that case, all identifying data is removed before sending.

Clone this wiki locally