Bank Statement Analyser

A comprehensive full-stack web application designed to automatically parse, redact, and analyse bank statement PDFs. By leveraging dual-layer personally identifiable information (PII) redaction and local AI-generated financial insights via Ollama (LLaMA 3), the platform provides secure, actionable intelligence on personal or business finances.

🎯 What Problem Does This Solve?

Analyzing bank statements manually is a tedious, error-prone, and time-consuming process. While AI makes extracting insights easier, uploading unredacted financial documents to third-party cloud services poses significant privacy and security risks.

Bank Statement Analyser solves this by:

1. Ensuring Absolute Privacy: Operating a dual-layer redaction pipeline. PII is redacted client-side in the browser before ever reaching the server. A secondary server-side pass ensures no sensitive strings slip through before the data touches the LLM.
2. Automating Financial Insights: Replacing manual spreadsheet data entry with an automated AI pipeline that categorises spending, identifies recurring transactions, flags unusual activity, and calculates net balances.
3. Keeping AI Local (or Controlled): Using Ollama (LLaMA 3) to perform the actual statement analysis, giving users the flexibility to keep their most sensitive financial analyses completely off big-tech cloud providers.

🚀 Key Features

• Secure PDF Upload & Parsing: Extracts text and bounding boxes from uploaded statements.
• Dual-Layer PII Redaction:
  • Client-side: Uses @openredaction/openredaction and pdf-lib to visually redact and remove PII from the PDF canvas in-browser.
  • Server-side: Uses Microsoft Presidio for a secondary pass on extracted text.
• AI-Powered Financial Analysis: Generates structured JSON insights (spending by category, recurring transactions, anomalies, etc.) utilizing LLaMA 3.
• Rich Dashboard & Visualizations: Interactive Vue 3 frontend utilizing Apache ECharts to display donut charts, income vs. expense breakdowns, and merchant analyses.
• Export Capabilities: Download your secure, categorized insights as a stylised PDF, Excel workbook, or raw JSON.
• Robust Authentication & Tagging: Secure JWT-based auth, user-specific statement tagging, and filtering.

🛠 Tech Stack

Frontend

• Framework: Vue 3 (Composition API) + Vite
• UI & Styling: Tailwind CSS 3, PrimeVue 4
• State & Routing: Pinia, Vue Router 4
• PDF & Canvas: pdfjs-dist, pdf-lib, @openredaction/openredaction
• Charts: Apache ECharts (vue-echarts)

Backend

• Framework: FastAPI (Python 3.11+)
• Database: PostgreSQL 15 + SQLAlchemy 2.0 (Async) + Alembic
• Task Queue: Celery + Redis for async PDF processing
• AI / LLM: Ollama Python SDK (LLaMA 3)
• Redaction: Microsoft Presidio (Analyzer + Anonymizer), pdfplumber, PyMuPDF
• Security: JWT (python-jose), bcrypt

For more in-depth technical details, please view the Backend Specification and Frontend Specification.

🏗 Project Structure

bank-analyser/
├── be/                 # FastAPI backend, Celery workers, and Python services
├── fe/                 # Vue 3 Vite application
├── Specification-docs/ # Detailed BE and FE architectural documentation
└── README.md

🏁 Quick Start

Prerequisites

• Node.js 20+
• Python 3.11+, uv installed
• Docker + Docker Compose (for Postgres and Redis)
• Ollama running locally or accessible as a cloud API (with llama3 model pulled: ollama run llama3)

Backend Setup

cd be
# Copy environment variables
cp .env.example .env
# Ensure you configure DATABASE_URL, REDIS_URL, etc., in .env

# Create virtual environment and install dependencies
uv venv
source .venv/bin/activate  # On Windows use: .venv\Scripts\activate
uv pip install -e .

# Run the FastAPI server
uvicorn app.main:app --reload

Make sure to run Alembic migrations (alembic upgrade head) and start Celery workers if processing background jobs.

Frontend Setup

cd fe
# Copy environment variables
cp .env.example .env
# Edit to point VITE_API_BASE_URL to your backend if different from default

# Install and run
npm install
npm run dev

📚 API Documentation

Once the backend is running, an OpenAPI (Swagger) UI is automatically generated. Visit: http://localhost:8000/api/v1/openapi.json or http://localhost:8000/docs to easily explore and test API endpoints.