File size: 3,953 Bytes
c0cc6f7 64e5ee2 c0cc6f7 64e5ee2 c0cc6f7 64e5ee2 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 | ---
title: AI Financial Reconciliation Engine
emoji: 🧠
colorFrom: blue
colorTo: indigo
sdk: docker
app_port: 7860
pinned: false
---
# 🧠 AI Financial Reconciliation Engine
Automated Financial Auditing using Machine Learning and LLMs.
## 🚀 Overview
The **AI Financial Reconciliation Engine** is an intelligent system designed to automate the process of matching internal accounting records (Books) with external tax filings (GST). By combining **Fuzzy Logic**, **AI Semantic Embeddings**, and **LLM reasoning**, the system identifies discrepancies, detects fraudulent anomalies, and provides natural language explanations for auditors.
## ✨ Features
- **Intelligent Matching**: Combines basic matching with Fuzzy and AI semantic analysis to reconcile records even with typos or name variations.
- **Anomaly Detection**: Uses the `IsolationForest` algorithm to detect unusual transaction patterns and high-risk invoices.
- **AI Explanations**: Integrates Mistral LLM to provide human-readable audit comments for every discrepancy.
- **Interactive Dashboard**: A professional Gradio interface with summary metrics, risk-sorted results, and CSV export.
- **Graph Fraud Network**: Visualizes circular trading and multi-hop tax siphoning fraud rings using `NetworkX` and `Matplotlib`.
- **Persistent Vector Memory**: Uses C++ compiled `FAISS` algorithms to permanently remember vendor vector embeddings.
- **Deployment Ready**: Containerized with **Docker** and hosted on **HuggingFace Spaces**.
## 🛠 Tech Stack
- **Languages**: Python
- **AI/ML**: Scikit-Learn, Sentence-Transformers, RapidFuzz
- **Fraud Engine**: FAISS, NetworkX, Matplotlib
- **LLM**: Mistral AI API
- **Frontend**: Gradio
- **Infrastructure**: Docker, HuggingFace Spaces
## 📂 Installation (Local)
1. Clone the repository.
2. Install dependencies: `pip install -r requirements.txt`
3. Set your `MISTRAL_API_KEY` in a `.env` file.
4. Run the app: `python main.py`
### Prerequisites
- Python 3.11+
- Virtual Environment (venv)
### Setup
1. Clone the repository
2. Create and activate virtual environment:
```bash
python -m venv venv
source venv/bin/activate # Linux/macOS
venv\\Scripts\\activate # Windows
```
3. Install dependencies:
```bash
pip install -r requirements.txt
```
4. Configure environment variables:
- Copy `.env.example` to `.env`
- Add your API keys
## Usage
### Quick Start
```python
from utils import create_sample_data
from reconciliation import ReconciliationEngine
from anomaly import AnomalyDetector
# Create sample data
data = create_sample_data(num_records=100)
source_df = data['source']
target_df = data['target']
# Run reconciliation
engine = ReconciliationEngine(threshold=85.0)
result = engine.reconcile(source_df, target_df, 'VendorName', 'VendorName', 'Amount')
# Detect anomalies
detector = AnomalyDetector(contamination=0.05)
anomaly_result = detector.detect_anomalies(source_df)
```
### Web Interface
```bash
python main.py
```
Access the UI at `http://localhost:7860`
### Docker
```bash
docker build -t reconciliation-engine .
docker run -p 7860:7860 reconciliation-engine
```
## Project Structure
```
├── sample_data/ # Live CSV data and scenarios
├── main.py # Main FastAPI backend serving UI
├── reconciliation.py # Core reconciliation engine & FAISS Index
├── anomaly.py # Anomaly detection module
├── fraud_graph.py # NetworkX Circular Trading Detector
├── gst_api.py # Real-time Local Registry Gateway
├── generate_real_data.py # Script to generate 1800+ realistic rows
├── llm_explainer.py # LLM-powered explanations
├── utils.py # Utility functions
├── requirements.txt # Python dependencies
├── Dockerfile # Docker configuration
├── .env # Environment variables
└── README.md # This file
```
|