title	emoji	colorFrom	colorTo	sdk	sdk_version	app_file	pinned	license
WizLedger - Monthly Statement Processor	💳	indigo	yellow	gradio	4.43.0	src/ui.py	false	mit

WizLedger - Monthly Statement Processor

Introduction

This project demonstrates a credit card statement assistant that can process PDF statements, perform OCR, and use LangGraph for AI-based transaction corrections.

Demo

FinKansis_edited3.mp4

Try it on HuggingFace

https://huggingface.co/spaces/chinkanai/wizledger

Technologies

• LangChain
• LangGraph
• Google Cloud Vision API Google Cloud Vision API
• OpenRouter API (for access Claude 3.5 Sonnet) OpenRouter API
• pdf2image (optional) pdf2image

LangGraph

Here is the graph of the workflow:

graph TD
    Start((Start)) --> A[ocr]
    A --> B[extract]
    B --> C[display_transactions]
    C --> D[get_human_input]
    D --> E[process_human_input]
    E --> F{check_if_done}
    F -->|continue| C
    F -->|done| G[store_csv]
    G --> End((End))

    style Start fill:#4CAF50,stroke:#333,stroke-width:2px
    style End fill:#FF5722,stroke:#333,stroke-width:2px

Loading

Prerequisites

• Python 3.10 or later
• OpenRouter API key (Get it here)
• Enable Google Cloud Vision API (Get service account keys here)
• Poppler (for pdf2image, optional)

Installation

1. Clone this repository:

   git clone https://github.com/chinkan/demo-ai-statement-extract
   cd demo-ai-statement-extract

2. Configure environment variables:

   • Copy .env.example to .env
   • Fill in the variables in the .env file

3. Run it

   • On Windows

   run.bat

   • On Linux or MacOS

   ./run.sh

Output

Processed data will be saved in the output folder.

Docker run from build local

docker build -t wizledger .
docker run -d -p 7860:7860 \
    -e OPENROUTER_MODEL="anthropic/claude-3.5-sonnet" \
    -e OPENROUTER_API_URL="https://openrouter.ai/api/v1" \
    -e OPENROUTER_API_KEY="<replace with your openrouter api key>" \
    -e GOOGLE_APPLICATION_CREDENTIALS="/app/secret/google-key.json" \
    -v ./secret:/app/secret \
    wizledger

Docker run from huggingface

1. Create a folder named secret and put google-key.json in it.
2. Run the following command:

docker run -it -p 7860:7860 --platform=linux/amd64 \
	-e OPENROUTER_MODEL="anthropic/claude-3.5-sonnet" \
	-e OPENROUTER_API_URL="https://openrouter.ai/api/v1" \
	-e OPENROUTER_API_KEY="<replace with your openrouter api key>" \
	-e GOOGLE_APPLICATION_CREDENTIALS="/app/secret/google-key.json" \
    -v ./secret:/app/secret \
	registry.hf.space/chinkanai-wizledger:latest python src/ui.py

API

import requests
import json
# Process a new statement
url = "http://localhost:7860/process"
files = {
    'statement': ('statement.pdf', open('path/to/statement.pdf', 'rb'), 'application/pdf'),
}
response = requests.post(url, files=files)
result = response.json()
print(result['result'])
thread_id = result['thread_id']

# Continue processing
url = "http://localhost:7860/continue_processing"
data = {
    'human_input': 'Some human input',
    'thread_id': thread_id
}
response = requests.post(url, json=data)
print(response.json())
result = response.json()

# Export transactions
url = "http://localhost:7860/export_transactions"
data = {'output': json.dumps(result)}
response = requests.post(url, json=data)

# Display the transactions
import pandas as pd
import io
df = pd.read_csv(io.BytesIO(response.content), encoding='utf-8')
display(df)

Contributing

We welcome contributions! To contribute to the project:

1. Fork the repository.
2. Create a new branch for your feature or bug fix.
3. Make your changes and commit them.
4. Push to your branch and create a pull request.

License

This project is licensed under the MIT License.