ClauseSense — Vendor Risk & Contract Copilot

Teammates: Sara Malik, Tanish J

ClauseSense is a Streamlit-based AI agent that reviews vendor contracts like a Legal Ops assistant. Upload a contract PDF and get an explainable risk report (with clause evidence), a plain-English summary, negotiation language, and a searchable clause library across vendors.

Disclaimer: This is a student project and not legal advice.

Features (MVP)

• Risk score + flagged clauses (with evidence)
• Plain-English summary of key terms
• Negotiation email + suggested fallback language
• Clause library search across uploaded contracts (semantic search)

Demo flow (what to show in class)

1. Upload a contract PDF (text-based PDFs work best)
2. Click Index Contract (builds the clause library)
3. Click Analyze Risk + Summary
4. Go to Clause Library and search for: auto renewal, termination, liability cap
5. Generate a Negotiation Draft and copy the email

Tech Stack

• Frontend: Streamlit
• LLM: Ollama (local model: llama / mistral)
• Agent orchestration: LangChain Community
• Embeddings: Sentence-Transformers
• Vector DB: ChromaDB (local)
• PDF parsing: pdfplumber
• Optional OCR: pytesseract + pdf2image

Setup (Mac)

1) Create venv + install deps

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

2) Install & Run Ollama

brew install ollama
ollama serve

In a new terminal:

ollama pull llama3.1:8b
# or: ollama pull mistral

3) Run App

python -m streamlit run app.py

Repo Structure

app.py
requirements.txt
core/
assets/

Screenshots