RAG — Financial Document Intelligence

Ask questions to PDF documents and databases in natural language. No SQL. No digging through folders.

RAG.NET 8Semantic KernelOpenAIPostgreSQL

This demo showcases a RAG (Retrieval-Augmented Generation) system that allows querying financial PDF documents and database-stored data using natural language. The system processes documents, generates vector embeddings, and uses GPT to generate accurate responses grounded in the actual document content.

Live Demo

Open demo in new tab

How it Works

1

Ingestion

PDFs and SQL tables are processed and split into text chunks optimized for semantic search.

2

Embeddings

Each chunk is converted into a numerical vector using OpenAI's text-embedding-3-small model.

3

Query

The question is vectorized, the most similar chunks are found, and GPT generates a grounded response.

Tech Stack

.NET 8Semantic KernelOpenAI (text-embedding-3-small)OpenAI (gpt-4o-mini)PostgreSQLPdfPigRailway

Source Code

View on GitHub

Want this for your business?

Let's discuss how I can build a solution like this for your organization.

Get in touch