Skip to content

Latest commit

 

History

History
157 lines (115 loc) · 5.96 KB

File metadata and controls

157 lines (115 loc) · 5.96 KB

🇵🇪 IndraLex: AI Compliance Assistant for Indra Perú

Python LangChain Streamlit Demo License

A functional RAG prototype that analyzes Peruvian regulations to accelerate project viability and mitigate risk. Built for Indra/Minsait.

▶️ Live Demo & Source Code

(No installation required. Deployed on Streamlit Community Cloud.)

(Repository includes instructions for local execution if desired.)

🎯 Projected Business Impact

Metric Traditional Process IndraLex (Estimated) Potential Impact
Analysis Time 6-8 hours* <1 minute ~480x faster
Cost per Analysis $150-200* $0.02-0.05 Significant reduction
Coverage Variable 5 core regulations Standardized

*Based on typical legal review cycles in enterprise projects

🚀 What IndraLex Does

IndraLex is a functional prototype that demonstrates how RAG (Retrieval-Augmented Generation) can accelerate regulatory compliance analysis for technology projects in Peru. Built specifically for Indra's context.

Core Capabilities

  • Processes 5 key Peruvian regulations: Ley 29733 (Data Protection), DL 1412 (Digital Government), SBS Circulars, and more
  • Identifies specific articles relevant to your project description
  • Maps regulatory entities (SEGDI, SBS, RENIEC) and their requirements
  • Generates actionable compliance roadmaps in seconds

🏗️ Technical Implementation

Architecture

PDF Laws → PyPDFLoader → Text Splitter → OpenAI Embeddings → FAISS Vector Store → GPT-4o-mini → Compliance Report

Core Stack

  • Framework: LangChain 0.3.27 (same architecture as Indra's enterprise RAG chatbots)
  • LLM: GPT-4o-mini (easily upgradeable to GPT-4o)
  • Vector DB: FAISS 1.12.0 with intelligent caching
  • Frontend: Streamlit 1.48.1 for rapid prototyping
  • Performance: <3 seconds with cache, processes 260+ legal chunks

💼 Real Use Cases for Indra Projects

🏦 Banking Sector (SBS)

Input: "Mobile banking platform with biometric authentication and real-time transactions"
Output: Specific SBS circular articles, LPDP requirements, technical controls needed

🏛️ Government Digital Transformation (PCM/SEGDI)

Input: "Interoperability platform for state entities processing citizen data"
Output: DL 1412 compliance points, SEGDI standards, integration requirements

🆔 Digital Identity (RENIEC)

Input: "Electronic DNI validation system with biometric verification"
Output: Law 27269 requirements, certificate specifications, RENIEC standards

📦 Installation (For Local Development)

Prerequisites

  • Python 3.9+
  • OpenAI API Key

Quick Start

  1. Clone the repository
git clone https://github.com/ncarrerakevin/IndraLex.git
cd IndraLex
  1. Install dependencies
pip install -r requirements.txt
  1. Configure environment
echo "OPENAI_API_KEY=your_api_key_here" > .env
  1. Add legal documents
mkdir legislacion
# Add Peruvian law PDFs to this folder
  1. Run the application
streamlit run app.py

📚 Supported Regulations

Regulation Coverage Relevant For
Ley 29733 Personal Data Protection All sectors
DL 1412 Digital Government Public sector
Circular SBS G-140 Cybersecurity Banking
Ley 30096 Computer Crimes All sectors
Ley 27269 Digital Signatures e-Government

📈 Strategic Vision for Indra Integration

This MVP is designed to evolve within Indra's ecosystem:

Phase Architecture Value for Indra
MVP (Current) Streamlit + Local FAISS High-impact demo, concept validation
Pilot FastAPI + Azure Functions Internal API for project teams
Production Docker + Azure Kubernetes Enterprise-scale solution

Why This Matters for Indra

  • Immediate value: Can be used tomorrow for real client proposals
  • Scalable architecture: Same RAG pattern used in production systems
  • Peru-specific: Not a generic tool, but tailored for local regulations
  • Integration-ready: Can connect with existing Indra platforms

🛠️ Technical Details

  • Chunk Strategy: 1500 chars optimized for legal articles
  • Retrieval: 8 most relevant chunks per query
  • Caching: MD5 hash validation for document changes
  • Model Config: Temperature 0.2 for consistent outputs
  • Dependencies: LangChain 0.3.27, OpenAI 1.99.9, FAISS 1.12.0

📊 Next Steps

  1. Expand coverage: Add more regulations (OSIPTEL, SUNAT, etc.)
  2. Enhance accuracy: Fine-tune prompts for specific sectors
  3. API development: RESTful interface for system integration
  4. Metrics tracking: Implement usage analytics and accuracy monitoring

👨‍💻 Author

Kevin Navarro C.

📄 License

MIT License - Open for enhancement and integration


Developed as a proactive demonstration of the initiative, technical skills, and business-oriented mindset I would bring to the AI R&D team at Indra.