LatamLex 🧠🌎

Fintech & Crypto Regulatory Assistant for LATAM using LLM + RAG

LatamLex is an intelligent assistant that answers questions about financial and crypto regulations in Argentina, Brazil, Mexico, and Chile. It uses a Retrieval-Augmented Generation (RAG) pipeline with semantic embeddings, FAISS search, and answer generation via LLM (like GPT-4).

🚀 Features

• ✅ Query laws, regulations, and resolutions by country
• ✅ Ingest legal PDFs and HTMLs
• ✅ Chunking and embeddings using sentence-transformers
• ✅ Semantic search with FAISS
• ✅ Contextual answers with source references and full fragments
• ✅ Simple interface via Streamlit
• ✅ Docker container for local demo

📦 Requirements

• Python 3.10+
• OpenAI API Key
• (Optional) Docker

🛠️ Local Installation

Clone the repository and create a virtual environment:

git clone https://github.com/0xregulus/latamlex.git
cd latamlex
make venv

Create a .env file with your OpenAI credentials:

OPENAI_API_KEY=your_openai_key_here
MAX_CONTEXT_CHARS=12000
TOP_K_RETRIEVAL=5

Install dependencies and prepare the data:

make venv
make all

Launch the app:

make app

Or run everything manually:

source .venv/bin/activate
python app/ingest.py
python app/chunker.py
python app/embedder.py
streamlit run app/interface.py

🐳 Using Docker

docker build -t latamlex .
docker run -p 8501:8501 --env-file .env latamlex

Or with Docker Compose:

docker-compose up --build

Or using Makefile:

make docker      # builds and runs the container
make compose     # runs docker-compose up --build

🧪 Running Tests

To run all automated tests:

make test

This will execute all unit tests inside the tests/ directory using pytest.

📁 Project Structure

LatamLex/
├── app/                   # Core code: ingestion, chunking, embeddings, retrieval, QA, interface
│   ├── ingest.py
│   ├── chunker.py
│   ├── embedder.py
│   ├── qa_chain.py
│   └── interface.py
├── data/                  # Input data organized by country
│   ├── argentina/
│   ├── brasil/
│   ├── mexico/
│   └── chile/
├── vectorstore/           # FAISS indexes and metadata
├── requirements.txt
├── Dockerfile
├── docker-compose.yml
├── Makefile
└── README.md

🧪 Example Question

What does Mexican law say about Financial Technology Institutions?

The app will retrieve the most relevant fragments from the Mexican legal documents, generate an answer, and display the sources used.

👨‍💻 Credits

Project created by Facundo Rodríguez as a demo of using LLMs and RAG in the fintech-regulatory domain.