From receipt to recorded transaction in one step
Most transactions still require at least a few seconds of manual entry: open the app, select a payee, enter the amount, pick a category. AI document processing removes that friction for transactions where you already have a physical or digital record of what happened.
You submit a receipt, invoice, email, or similar document. YAFFA reads it, extracts the relevant details — payee, date, amount, and line items — and builds a draft transaction pre-filled with everything it found. You review the draft, adjust any field that needs correcting, and save. The transaction is recorded and linked back to the original document for future reference.
What YAFFA extracts from a document
- Transaction date and amount
- Payee — matched against the payees you have already set up in YAFFA
- Account — matched against your existing accounts
- Individual line items with suggested categories — for receipts that break down the purchase into multiple goods or services
YAFFA learns from your past categorization choices. When the same item description appears again in a future document, it reuses the category you used before — so the suggestions become more accurate the more you use the feature.
Document sources
You can submit documents in three ways:
Manual upload
Upload a file directly from your device — PDF, JPG, PNG, or plain text. Useful for paper receipts you have photographed or scanned, and for invoices saved as PDF.
Email forwarding
Forward a receipt or order confirmation email directly to your YAFFA instance. The email body is processed automatically and a draft appears in your document queue without any additional steps.
Google Drive sync
Point YAFFA at a Google Drive folder. Any file you drop into that folder is picked up automatically and processed without you having to open YAFFA at all.
How text is extracted from documents
YAFFA uses different extraction methods depending on what you submit:
- Text-based PDFs — text is extracted directly, no AI image recognition needed.
- Images and scanned documents — processed using Tesseract OCR (self-hosted, free) or a vision-capable AI model such as GPT-4o or Gemini, depending on your configuration.
- Plain text files and emails — read directly.
Duplicate detection
Before showing you the finalization form, YAFFA checks whether a similar transaction already exists — same approximate date, amount, payee, and account. If a potential duplicate is found, a warning is displayed above the form so you can decide whether the new entry is genuinely distinct or whether the document was already recorded.
Setup requirements
AI document processing is optional. All other YAFFA features work without it.
To use it, you need to configure an AI provider in your YAFFA settings — currently OpenAI or Google Gemini. You supply your own API key, and your documents are processed using your account with that provider. YAFFA does not provide or resell AI access; you control the costs and the data sent to the AI service.
For image-based documents, you can optionally enable Tesseract OCR as a free self-hosted alternative to cloud vision models. This keeps all document processing local if that matters for your privacy setup.
You stay in control
AI processing in YAFFA is always assistive, never automatic. The AI produces a draft; you review and confirm it. No transaction is ever created without your explicit approval. If the AI gets something wrong — a misidentified payee, an incorrect category — you correct it before saving. The finalized transaction reflects your decision, not the AI's guess.
This matches YAFFA's broader philosophy: automation should help you record your finances faster, not replace your understanding of where your money goes.