Contract Answerer

An AI powered document question answering application that allows users to ask natural language questions about long PDF documents and receive accurate, source grounded answers.

This project demonstrates a production minded implementation of retrieval augmented generation using modern LLM tooling.

Demo

90 second walkthrough and setup video:

Why this project matters

Many real world workflows depend on large and complex documents that are difficult to search or reason about manually. This project explores how large language models, embeddings, and vector search can be combined to make long form documents immediately useful while minimizing hallucinations.

Key features

1. Ask natural language questions over arbitrary PDF documents
2. Embeddings based retrieval for precise context selection
3. Context constrained LLM responses to improve accuracy
4. Simple and fast UI designed for iteration and experimentation

Tech stack

1. Python
2. Streamlit
3. OpenAI API
4. ChromaDB
5. LangChain

Quickstart two minutes

Step 1 Clone the repository

git clone https://github.com/Aeh961/Contract-Answerer.git
cd Contract-Answerer

Step 2 Create and activate a virtual environment

python -m venv .venv
source .venv/bin/activate

Step 3 Install dependencies

pip install -r requirements.txt

Step 4 Add your documents

mkdir -p data/contracts

Place any PDF files you want to query into the data contracts folder.

Step 5 Set your API key

export OPENAI_API_KEY="your_key_here"

Step 6 Run the application

streamlit run src/app.py

How it works

1. PDF documents are loaded and split into semantic chunks
2. Each chunk is embedded and stored in a vector database
3. User questions retrieve the most relevant chunks
4. The language model generates answers using only retrieved context

Project structure

src
  app.py          Streamlit application
  load.py         Document ingestion and embedding
data
  contracts       User provided PDFs
README.md
requirements.txt

Future improvements

1. Inline citations with highlighted source text
2. Streaming responses for improved user experience
3. Document metadata filtering
4. Cost and latency optimizations

Author

Built by Abdallah Elhamawi as part of a broader exploration into practical and reliable AI systems.