Environment Setup: You will have to use your own API key in the .env file for complete functioning of the code. Working model will be demonstrated in video and live demo.
Annual reports are packed with valuable insights about a company’s financials, strategy, and performance. However, these reports are often long, complex, and hard to understand quickly — especially for investors, researchers, and business analysts.
Build an AI-powered app that can automatically:
- Read and understand uploaded annual report PDFs
- Process Text, Images, Tables
- Answer questions users ask in natural language (Q&A)
- Extract and summarize content from charts, tables, and text
- Help users gain key insights efficiently without reading the whole document
- Investors doing company analysis
- Financial journalists needing quick facts
- Business students learning report analysis
- Compliance officers reviewing key highlights
Automating this process reduces hours of manual reading and allows non-experts to access crucial insights in seconds.
Our app uses a RAG (Retrieval-Augmented Generation) pipeline to accurately answer questions grounded in the uploaded annual report.
- 🔹 PDF Text Extraction: Extract raw text and tables using PyPDF2.
- 🔹 Able to process images using Tesseract OCR
- 🔹 Chunking & Embedding: Split the text and generate vector embeddings using OpenAI's
text-embedding-ada-002. - 🔹 Semantic Retrieval: Use FAISS to fetch only the relevant chunks of the report.
- 🔹 Q&A Generation: A lightweight GPT model generates grounded answers based only on retrieved context and semantic search
This ensures users get fact-based answers derived from the annual report — no hallucinations.
Our app detects and extracts tables and figures from the PDF using PyMuPDF and image-based parsing. We ensure questions related to tabular data (like revenue by year) can be answered accurately, with fallback handling if charts are unavailable.