LexForge

LexForge is a privacy-first AI toolkit that analyzes legal contracts and fixes code using intelligent agents powered by Gemini.

Comment 1

Inspiration

LexForge: Full-Stack AI System Documentation

1. Project Overview

LexForge is a full-stack AI-powered system built to handle two critical tasks:

  • ⚖️ Contract Analysis (Lawyer Agent)
  • 🛠️ Code Debugging & Fixing (Coding Agent) What makes it different is not the features, but the privacy-first pipeline. Data is sanitized before it ever touches AI. ## 2. Technology Stack ### 🌐 Frontend
  • HTML → Structure
  • CSS → Styling & themes
  • JavaScript → Logic + API calls ### ⚙️ Backend
  • FastAPI (Python)
  • Async request handling
  • File upload system ### 🤖 AI Engine
  • Gemini Flash Lite ### 🛠️ Supporting Systems
  • PyPDF2 → PDF extraction
  • Redaction Engine → Privacy layer
  • Workspace Storage → File handling ## 3. System Architecture Frontend (Browser UI) ↓ HTTP Requests (Fetch API) ↓ FastAPI Backend ↓ ↙ ↘ Lawyer Agent | Coding Agent ↘ ↙ Gemini API ↓ Response Processing ↓ Frontend Rendering ## 4. Frontend Design ### Core Layout
  • Sidebar → Mode selection
  • Topbar → Tabs + model status
  • Workspace → Dynamic UI
  • Right Panel → Info + explanation ### Lawyer Mode Flow Upload contract → Processing → Redaction → Results You see:
  • 📊 Risk Scorecard
  • 🃏 Risk Cards (High / Medium / Low)
  • 💡 Fix Suggestions ### Code Mode Flow Upload code → AI Processing → Logs → Output Logs simulate:
  • 📤 Sending to AI...
  • 🔍 Analyzing code...
  • ✅ Fix applied...
  • 💾 Saved file.py ### Key Features
  • switchMode() → instant UI switching
  • Configurable backend URL (localStorage)
  • API Calls: /analyze-contract and /fix-code
  • ⚠️ Error display system
  • 📜 Real-time log simulation ## 5. Backend System ### Framework
  • FastAPI
  • Async endpoints
  • File handling ### Endpoints
  • /analyze-contract: Upload file, extract text, apply redaction, AI analysis, return structured output.
  • /fix-code: Upload code, send to AI, parse result, save file, return preview. ## 6. File Processing ### PDF Extraction
  • PyPDF2: Reads all pages and combines into text. ### Upload Rules
  • Max size: 5MB
  • Formats (Lawyer): PDF, TXT
  • Formats (Code): .py, .js, .cpp, .java, .txt ## 7. Privacy Layer (Core Strength) ### Redaction Pipeline Raw Text → redact_text() → PII Removed → Gemini API → restore_text_deep() → Original Restored ### ☑️ Removed Data
  • 👤 Names
  • 📞 Phone numbers
  • 📧 Emails
  • 🏦 IBAN
  • 📍 Locations AI never sees real user data. That's the whole point. ## 8. AI Integration
  • Model: Gemini Flash Lite
  • Call Flow: generate_content()
  • Retry System: 3 attempts
  • Handles: 429 (rate limit) with exponential backoff
  • Failure: 503 not handled → breaks system ## 9. Lawyer Agent
  • Input: Contract text
  • Output: JSON object with score, risks, and fixes.
  • Features: Risk scoring, clause detection, severity tagging.
  • UI Mapping: Score color, risks cards, fixes → suggestions. ## 10. Coding Agent
  • Input: Raw code
  • Expected Output: JSON with action: "write", filename, and content.
  • Processing Flow: AI Response → clean_json() → json.loads() → write_to_disk().
  • Safety: Prevents path traversal and protects critical files.
  • Response: Status: "saved", file path, and first 300 chars preview. ## 11. Error System
  • Backend Errors: Invalid file, file too large, empty file, AI failure.
  • AI Errors: ☑ 429 handled | ☑ 503 not handled
  • Frontend Errors: Displayed via UI alerts ## 12. Data Flow
  • Code Agent: Upload → /fix-code → Backend → AI → Parse → Save → UI
  • Lawyer Agent: Upload → Extract → Redact → AI → Restore → UI ## 13. Strengths
  • 🛡️ Privacy-first design
  • 🧩 Modular agents
  • ✨ Clean UI
  • 🔒 Secure file handling
  • 🌍 Real-world usability
  • 📈 Easily extendable ## 14. Conclusion LexForge is not just a basic AI tool. It's a structured multi-agent system combining legal intelligence and code debugging with a privacy-focused pipeline. ### Final Summary LexForge = Privacy-first AI system that analyzes contracts and fixes code using a FastAPI backend, Gemini AI, and a dynamic frontend UI. ₹ ## What it does

How we built it

Challenges we ran into

Accomplishments that we're proud of

What we learned

What's next for LexForge

Built With

Share this project:

Updates