Vantora, personal finance software that keeps your accounts at home
Your statements, receipts and emails: read on your computer, stored in an encrypted vault.
Project at a glance
- The need
- Track personal finances with dashboards, without handing one’s accounts to an online service.
- My contribution
- Designed and built a local desktop app: import of statements, receipts and emails, an encrypted database, dashboards and AI on the computer.
What was built
- PDF, receipt and email import
- Encrypted vault
- Dashboards and budgets
- Local AI advisor
Project on hold since April 2026. No screenshot published, as the local database holds real data; the visual is a diagram composed from the code.
- Nature
- Desktop software (Python, Streamlit) · personal project on hold
- Period
- March 2026 – April 2026
- Client
- VantoraPersonal project, desktop software
The project
Personal finance software designed so that no banking data leaves the computer. It answers a simple question: how do you get dashboards on your accounts without handing them to an online service?
What the software does
- Import of PDF bank statements (pdfplumber and AI), photos of receipts (EasyOCR, in French and English), financial emails (IMAP), and manual entry with AI categorisation.
- Dashboards: balances per account, flows (Sankey diagram), a spending heatmap, a monthly waterfall and budgets.
- Wealth: stock portfolio, loans with an amortisation table, savings goals and debts between individuals.
- An AI advisor that relies on your local data and spots subscriptions.
- Seven themes, French and English, exports to PDF and CSV.
Privacy, in the code
The SQLite database is encrypted (Fernet, AES-128) and the PIN is hashed (Argon2id). The AI runs on the computer, with Ollama and the Qwen 2.5 (7B) model; without Ollama, it switches off and the software keeps working. Each outgoing connection serves one specific feature: the prices you follow (Yahoo Finance), exchange rates, the update check (GitHub) and reading your emails (IMAP), when you use it. The project’s privacy policy rules out any telemetry.
Evidence
Every figure carries its source. No commercial result is claimed.
42
Automated tests passing
SourceProject roadmap (step 3.3), re-read on 13 October 2026
16
Commits from 21 March to 9 April 2026
SourceProject repository history, re-read on 13 October 2026
7
Display themes
SourceProject README, version 1.0.0 (23 March 2026)
Where the project stands
Version 1.0.0 dates from 23 March 2026 and the last commit from 9 April 2026: the project is on hold. It served as the testing ground for writing my method for working with AI, presented in another case study.
What is not claimed
No users and no sales are claimed. No screenshot is published, because the database on this machine holds real data: this file’s visual is a diagram composed from the code. The project’s README announces two network calls; its privacy policy and its code show more, and this file follows the code’s list.
Technical details
The technologies used for this project. Features and their limits are detailed in the narrative.
- Python
- Streamlit
- SQLite
- Ollama
- EasyOCR
- pdfplumber
Expertise involved
Gallery
The path of a piece of data, from the code (a diagram, not a screenshot).
A similar need?
Let's discuss your context and what you would like to simplify.