This repository implements a complete stack for an educational chatbot powered by Generative AI, using a Retrieval-Augmented Generation (RAG) architecture, Docker Compose, Streamlit, LangChain, GitHub Actions, and automated deployment to a VPS using Nginx + HTTPS via Certbot.

The AI assistant, codenamed Luiza, answers questions based on a local knowledge base built from Markdown files, combining vector search with OpenAI-generated responses.

• LangChain Retriever using embedded ChromaDB
• Embeddings are generated on boot from .md files in the /docs folder
• Integration with RetrievalQA and OpenAI (configurable model)

• Dockerfile based on python:3.13-slim
• docker-compose.yml orchestrates app + Nginx
• HTTPS reverse proxy via Certbot
• Persistent volume for CSV-based structured logs
• Automated deployment to a VPS using CI/CD

• GitHub Actions pipeline performs:

  • Code linting with ruff, bandit, mypy
  • Docker image build
  • Push to Docker Hub
  • Remote deployment via SSH + docker compose up

.
├── .github/workflows/        # CI/CD Pipelines
├── /docs/                    # Markdown knowledge base (indexed with ChromaDB)
├── /logs/                    # CSV logs of user queries
├── retriever.py              # Document loader and embedder
├── streamlit_app.py          # Main application interface
├── docker-compose.yml        # Service orchestration
├── Dockerfile                # Main Docker image
├── redeploy.sh               # Manual restart script for VPS
└── .env                      # OpenAI API key

1. Clone the repository:

git clone https://github.com/92username/langchain-quickstart.git
cd langchain-quickstart

2. Create a .env file with your OpenAI key:

OPENAI_API_KEY=sk-xxxxx

3. Run with Docker Compose:

docker compose up --build

The application will be available at http://localhost:8501

The production environment runs on a VPS (Hostinger) and is accessible via:

https://estudamais.tamanduas.dev

• Secure HTTPS requests
• Reverse proxy via Nginx
• Automated deployment via GitHub Actions

• All user questions are logged in logs/conversas.csv
• Volume is persistent across builds
• Logs may be used to generate a data-driven FAQ in the future

• RAG with LangChain + OpenAI
• Automated deployment with GitHub Actions
• Persistent interaction logging
• Implement caching for frequently asked questions and repeated answers.
• Auto-generated FAQ module
• Admin dashboard with usage metrics

• bandit for security scanning at build time
• ruff and mypy for linting and static typing
• CI pipeline fails on security violations