FinExtract AI

Tiny tool to read money from invoices. Fast. Friendly.

Quick demo

• You drop a PDF invoice.
• The app finds invoice number, date, vendor, subtotal, tax, and total.
• You click one big green button. It gives you a CSV.

Start in 3 minutes

Copy these 4 lines. Done.

bash
python -m venv .venv
# Windows
.\.venv\Scripts\pip install -U pip
.\.venv\Scripts\pip install -r requirements.txt
.\.venv\Scripts\python -m streamlit run app.py

Mac/Linux users:

bash
python -m venv .venv
source .venv/bin/activate
pip install -U pip
pip install -r requirements.txt && streamlit run app.py

Super simple setup (for beginners)

1. Install Python 3.10 or newer.
2. Open your terminal.
3. Paste the 4 commands above.
4. Put your PDFs ready.
5. Open the app link in your browser. Smile 🙂

How to use

1. Upload your invoice PDF.
2. Click the big button: "Extract & Download CSV".
3. Save the CSV to your computer.

UI / UX specs

• Put your mockup here:

  • 

• Streamlit Theme (copy to .streamlit/config.toml)

toml
[theme]
base = "light"
primaryColor = "#10B981"          # green
backgroundColor = "#FFFFFF"
secondaryBackgroundColor = "#F7F7F8"
textColor = "#111827"
font = "sans serif"

• Google Fonts (Inter / Poppins)

html

• Button style guidance

  • Color: Green (#10B981).
  • Corner radius: 8–12px.
  • Text: Bold, white.
  • Size: Big. Easy to click.
  • Focus ring: Clear, 2px outline.

• Big green button (HTML/CSS preview)

html

  Extract & Download CSV

Large, friendly, accessible

• ASCII button (if images not possible)

╔══════════════════════════════════════╗
║          Extract & Download CSV      ║
╚══════════════════════════════════════╝

• Accessibility
  • aria-label: "Extract and download CSV"
  • Keyboard: Enter/Space triggers button. Tab shows focus ring.

Prompt / LangChain / OpenAI usage

Short prompt for extraction (example):

text
You are a careful extraction bot.
Read the invoice text below and return ONLY valid JSON.
If a value is unknown, omit that key.

Expected JSON keys:
{
  "invoice_number": "string",
  "date": "YYYY-MM-DD",
  "vendor": "string",
  "subtotal": 0.0,
  "tax": 0.0,
  "total_amount": 0.0,
  "line_items": [
    {"description": "string", "qty": 0, "unit_price": 0.0, "amount": 0.0}
  ]
}

Document:
{{document_text}}

Short prompt for classification:

text
You classify this document. Return ONLY JSON:
{"type": "Invoice" | "Receipt" | "Purchase Order" | "Statement"}

Text:
{{document_text}}

Example JSON schema for extracted fields:

json
{
  "invoice_number": "INV-00123",
  "date": "2025-03-17",
  "vendor": "Acme Pvt Ltd",
  "subtotal": 1000.00,
  "tax": 90.00,
  "total_amount": 1090.00,
  "line_items": [
    {"description": "Widget A", "qty": 2, "unit_price": 500.00, "amount": 1000.00}
  ]
}

Export / CSV

Sample CSV header:

csv
doc_type,invoice_number,date,vendor,subtotal,tax,total_amount,_filename

Sample row:

csv
Invoice,INV-00123,2025-03-17,Acme Pvt Ltd,1000.00,90.00,1090.00,invoice_1.pdf

Local CSV download (Streamlit snippet)

python
import pandas as pd
import json
import streamlit as st

# fields is a dict with extracted values
df = pd.json_normalize(fields)
csv_bytes = df.to_csv(index=False).encode("utf-8")

st.download_button(
    label="Download CSV",
    data=csv_bytes,
    file_name="finextract.csv",
    mime="text/csv"
)

st.download_button(
    label="Download JSON",
    data=json.dumps(fields, indent=2).encode("utf-8"),
    file_name="finextract.json",
    mime="application/json"
)

Troubleshooting

• App does not start
  • Close old Streamlit tabs. Run the command again.
• Nothing happens after upload
  • Click the big button. Wait a few seconds.
• Wrong total
  • Check invoice has "Subtotal", "Tax", "Total". Try a clearer PDF.
• Missing invoice number
  • Look for "Invoice No" or "INV #" near the top. Try another file.
• OpenAI key issues
  • Create and set OPENAI_API_KEY in your environment, then restart.

FAQ

• Do I need internet?
  • Yes, for OpenAI. No for pure regex-only mode.
• Can I use this with Tally?
  • Export CSV and import to Tally or other tools.
• Does it read images?
  • Focus is PDF. Images may work, but PDF is best.
• Can I add my own fields?
  • Yes. Update the schema and UI.
• Is my data safe?
  • Keep keys local. Do not share sensitive docs.

Contributing / Contact / License

• Pull requests welcome.
• Open an issue to discuss features.
• License: MIT

MIT line:

MIT License — do what you want, just include the license.

Images & assets

• Put mockups here:
  • assets/ui-mockup.png
  • assets/button.png

Folder guide:

project/
  assets/
    ui-mockup.png
    button.png

Quick Theme (copy to .streamlit/config.toml)

toml
[theme]
base = "light"
primaryColor = "#10B981"
textColor = "#111827"
backgroundColor = "#FFFFFF"
secondaryBackgroundColor = "#F7F7F8"
font = "sans serif"

Minimal requirements.txt

txt
streamlit==1.39.0
openai>=1.40.0
langchain>=0.2.12
pytesseract>=0.3.10
pandas>=2.2.2
pillow>=10.4.0

What to change for production

• Use a secrets manager for OPENAI_API_KEY and any keys.
• Add document redaction and logging controls.
• Add tests with a small set of known PDFs and expected outputs.

If stuck, drop a message — I'll help. 🟢