Andalûh EPA Translation System and Chatbot 🇳🇬

This project implements an intelligent agent system using LangChain and LangGraph capable of maintaining a conversation (Chatbot) and automatically translating its responses to the Andalûh EPA (Êttandâ pal Andalûh) orthographic proposal.

It allows interaction both via command line (CLI) and through a modern web interface built with Streamlit, and supports multiple LLM providers (Google Gemini and Ollama).

Features

• Agent Architecture: Intelligent pipeline Chatbot -> Translator using LangGraph.
• Multi-Model: Native support for:
  • Google Gemini (via API).
  • Ollama (Local models like Llama 3, Mistral, etc.).
• Interfaces:
  • 🖥️ CLI: Simple terminal interface.
  • 🌐 Web App: Full application with Streamlit, including basic authentication.
• Dockerized: Easy deployment with Docker Compose (including Ollama).

Project Structure

• agents.py: LangGraph graph definition and agent logic.
• tools.py: Custom tools (wrapper for the andaluh library).
• app.py: Streamlit web application.
• main.py: CLI entry point.
• tests/: Unit test suite to verify translation rules.

---

🚀 Quick Deployment with Docker (Recommended)

The easiest way to run the full stack (Web App + Local Ollama) is using Docker Compose.

1. Start the services:

   docker compose up --build -d

2. Download model (First time only): If using Ollama, you need to download the model inside the container:

   docker exec -it andaluh-ollama ollama run llama3.2

3. Access: Open your browser at http://localhost:8501.

   • User: admin
   • Password: andaluh

---

🛠️ Local Installation and Usage (Development)

Requirements

• Python 3.11+
• (Optional) Ollama installed locally or Google API Key.

1. Environment Setup

1. Create virtual environment:

   python3 -m venv venv
   source venv/bin/activate

2. Install dependencies:

   pip install -r andaluh_chatbot/requirements.txt

3. Configure environment variables: Create a .env file in andaluh_chatbot/.env based on the following example:

   To use Ollama (Local):

   LLM_PROVIDER=ollama
   OLLAMA_BASE_URL=http://127.0.0.1:11434
   OLLAMA_MODEL=llama3.2:latest

   To use Google Gemini (Cloud):

   LLM_PROVIDER=google
   GOOGLE_API_KEY=your_api_key_here
   MODEL=gemini-2.5-flash

2. Run Application

Command Line Interface (CLI):

python -m andaluh_chatbot.main

Web Interface (Streamlit):

streamlit run andaluh_chatbot/app.py

---

✅ Tests

The project includes tests to verify that translation rules (seseo, ceceo, vowel opening, etc.) are applied correctly.

Run tests:

python -m unittest andaluh_chatbot/tests/test_translation.py