Every month I was doing the same manual work: export bank statements, clean messy CSV/PDF formats, decode ugly descriptions, categorize transactions, then update spreadsheets by hand.
So I built Spectra.
Spectra is a file-based Python pipeline that ingests bank exports (CSV/PDF), normalizes them, optionally categorizes transactions with an LLM (OpenAI/Gemini), detects recurring payments, converts historical FX rates, and updates a Google Sheets dashboard automatically.
Core parsing/normalization/dedup runs locally; categorization can be cloud-based depending on configuration.
If you’ve dealt with bank exports before, I’d love feedback on the worst edge cases you’ve seen and what you’d prioritize next.
Repo: https://github.com/francescogabrieli/Spectra
Top comments (0)