RedlineAI

• 

  Document Processing

• 

  Update a document

• 

  Analyze Risk

• 

  chat with your document

• 

  AI Chat assistance with your document

• 

  List of Alerts

• 

• 

  Login Screen

Inspiration

When I was signing my college apartment lease, I remember feeling overwhelmed by all the legal terms. I skimmed most of it because I didn’t fully understand what I was reading. Later, I realized there were hidden fees and responsibilities I hadn’t noticed — one being a $10 monthly charge for a third-party delivery company I never used.

That experience made me realize how inaccessible contracts are for everyday people. My mom faced an even tougher situation when signing her mortgage: pages of complex clauses about interest rates, penalties, and repayment schedules. Redline AI was born from the desire to make legal and financial documents understandable for everyone, not just lawyers.

What it does

Redline AI is an AI-powered contract reviewer and SLA (Service Level Agreement) copilot that helps individuals, small businesses, and organizations quickly understand, assess, and negotiate contracts. • 🧾 Upload any contract or SLA (PDF/DOCX) • 🤖 AI Agents extract text, detect clauses, and label sections (e.g., “Termination,” “Uptime,” “Jurisdiction”) • 🚨 Risk Detection flags high-risk or non-standard terms compared to policy benchmarks • ✍️ Smart Redlines generate suggested edits and fallback options • 📊 Executive Summary provides an easy-to-read dashboard with risk levels, key dates, and recommendations • 🔍 Search & Q&A lets users ask natural-language questions like “Where is the uptime defined?” • ⏰ Alerts & Reminders notify users before renewals or compliance deadlines

With Redline AI, what used to take a legal team 4+ hours can be done in under 5 minutes—with plain-English explanations for everyone.

How we built it

• Frontend: React + Tailwind (modern and intuitive UI for uploads and visualization) • Backend: FastAPI (Python 3.11) handling ingestion, classification, and redlining pipelines • Database: TiDB Serverless for hybrid storage (vector + relational + FTS search) • AI Stack: o Ingestion Agent: OCR + text extraction o Clause Classifier Agent: clause tagging with GPT-4 o Risk & Compliance Agent: compares extracted clauses to internal rules/policies o Redline Writer Agent: generates redline suggestions while maintaining legal tone o Summarizer Agent: creates concise executive summaries o Reviewer Copilot Agent: enables hybrid semantic + keyword search • Integrations: Slack/Email/Calendar notifications for clause-based reminders

Challenges we ran into

• Creating reliable clause detection for varied contract structures (some documents had tables or images instead of text). • Fine-tuning risk scoring so that AI explanations were both legally meaningful and user-friendly. • Managing vector + FTS synchronization in TiDB to make search results accurate and fast. • Maintaining tone control—the AI had to write redlines that sound professional, not robotic or casual.

Accomplishments that we're proud of

• Built a fully functional prototype that reviews and summarizes contracts end-to-end. • Successfully reduced contract review time from hours to minutes. • Designed an AI system that outputs plain-English legal summaries understandable by anyone. • Created a scalable architecture with TiDB’s vector and FTS capabilities for real-time querying.

What we learned

• The importance of explainability in AI for legal and civic use cases—users must trust the AI’s reasoning. • Combining multiple specialized agents is far more effective than a single large model prompt. • Civic tech projects thrive when they empower non-experts to make informed decisions with confidence.

What's next for RedlineAI

• 📱 Launching a mobile-friendly dashboard for faster uploads and alerts. • 🏛️ Partnering with local housing and consumer protection organizations to make lease and mortgage transparency accessible to all. • 🧩 Adding multi-language support for international agreements. • 📚 Building a public clause library of safe and risky examples to educate users. • 💬 Integrating Slack and email copilots for real-time review assistance.

Useful links: Project Demo => https://www.devphilip.com/projects/redlineaidemo/ Try it out => https://www.devphilip.com/projects/redlineai

Built With

• fastapi
• langchain
• langgraph
• llm
• python
• react
• sql
• typescript
• vectordb