Receipt & Document Agent
Upload a receipt or PDF → extract vendor, date, amount, lines → validate → exception queue.
Structured fields with validation and exception queue for human review.
30-second recap
- Upload a receipt or PDF → extract vendor, date, amount, lines → validate → exception queue.
- Structured fields with validation and exception queue for human review.
Receipt & Document Agent
The interactive console loads on demand. Upload your own file or edit the example input. Every result comes back live from the API.
The problem
Receipts arrive as photos and PDFs; manual entry is slow and error-prone.
How it works
Upload → extract fields → validate schema → high confidence auto-draft; low confidence to exception queue.
System design
Vision-capable LLM path (or OCR + LLM) returning typed fields plus confidence per field via LLM structured JSON.
Human review & control
Reviewers correct fields inline; escalations flag missing tax invoices or unreadable scans.
Integration scope
Stores drafts for downstream coding / bill creation (mock).
Stack
- Vision LLMs
- OCR
- Next.js
What the live demo shows
Live structured LLM output returning typed fields with per-field confidence. File parse runs in the browser, no server storage.
Current limitations
No real OCR pipeline or accounting system push. Validation and downstream bill creation are simulated.