FinancialQA-Assistant

A local, privacy-preserving question-answering system for financial documents using FastAPI, sentence-transformers, and Ollama Mistral. The system extracts content from raw PDFs, performs semantic search over chunked embeddings, and returns precise answers with a clean HTML UI.

---

Features

• FastAPI-based backend for real-time inference
• PDF ingestion using pdfplumber
• Semantic search using sentence-transformers
• LLM integration using Ollama Mistral for local answer generation
• Simple UI using HTML/CSS/JS (chat-style interface)
• Fully offline and privacy-respecting

---

Interface Preview

Main Chat Interface

Example Query 1

Example Query 2

---

Setup Instructions

1. Clone the repo

git clone https://github.com/Amaan-developpeur/FinancialQA-Assistant.git
cd FinancialQA-Assistant

Create a virtual environment (optional but recommended)

python -m venv .venv
.\.venv\Scripts\activate

Install dependencies

pip install -r requirements.txt

Start Ollama and pull Mistral (if not done) [Manually Install Ollama from Official Website]

ollama run mistral

Run the Application

uvicorn app.main1:app --reload

Then open your browser and visit:
http://127.0.0.1:8000

Project Structure

financial-qa/
├── app/
│   ├── main.py               # Swagget.UI can be openned
     ├── main1.py             # FastAPI app entrypoint
│   ├── query_engine.py         # Embedding search logic
│   ├── templates/chat.html     # Jinja2-based UI
│   ├── static/                 # CSS & JS files
│   └── utils/
│       ├── local_generate.py   # LLM inference using Ollama
│       └── system_prompt.py    # System prompt builder
├── data/
├── scripts/                      
├── Images/                     # Screenshots for documentation
├── requirements.txt
├── .gitignore
└── README.md

LICENSE

This project is licensed under the MIT License.