ask.UOS

AI chatbot for applicants and students – get instant answers about studies, teaching, and campus life.

(Under development)

---

Table of Contents

• Overview
• Features
• Setup
  • Development with Docker
  • Production with Docker
• Configuration
• Milvus Vector DB
• Translation Mechanism

---

Overview

ask.UOS is designed to provide an interactive interface for users to communicate with a chatbot and access up-to-date university information. It leverages the LangChain library for natural language processing and Streamlit for the web interface. The system supports both German and English, and can be easily extended to other languages.

ask.UOS can answer a wide range of general questions about university life in just a few moments, including inquiries about the application process, academic requirements, and various study programs. The chatbot enables users to receive quick support in a conversational manner, without having to wait for office hours or search for the right contact person—available 24/7. For university departments such as student advisory services, the chatbot also helps reduce the volume of support requests.

---

Features

• Interactive chatbot interface for students and applicants
• Content summarization and document insights
• PDF reading and processing
• Web scraping and data extraction
• Multilingual support (German and English)
• Customizable prompts and responses
• Vector database (Milvus) for efficient unstructured data retrieval

---

Setup

Prerequisites

• Docker and Docker Compose
• (For development) See requirements.txt
• (For production) See requirements.prod.txt

---

Development with Docker

1. Clone the repository

   git clone <your-repo-url>
   cd <your-repo>

2. Create and configure environment file

   cp .env.dev-example .env.dev
   # Edit .env.dev as needed

3. Start the application

   docker compose up -d

4. Access the chatbot
   Open http://localhost:8501/ in your browser.

Note: If you run the application without Docker, you must set up a Milvus server manually.

---

Production with Docker

1. Create and configure environment file

   cp .env.prod-example .env.prod
   # Edit .env.prod as needed

2. Set up Nginx configuration (optional, for reverse proxy)

   cp ./nginx/nginx.conf.example nginx.conf
   # Edit nginx.conf as needed

3. Start the application in production mode

   docker compose -f docker-compose.prod.yml up -d

---

Configuration

This application uses a configuration file (config.yaml) to manage its settings. The configuration is validated at startup using Pydantic, ensuring all required fields are present and correctly formatted.

• Modify the example file as needed for your environment.

   cp config_example.yamo config.yaml
   # Edit config.yaml as needed

---

Milvus Vector DB

• A vector database (Milvus) is used for indexing and storage of unstructured data (e.g., pdf files).
• Milvus is automatically set up via Docker Compose and does not require manual installation.
• If you are already running a Milvus instance, you can connect ask.UOS with it. See config_example.yaml

---

Translation Mechanism

ask.UOS supports both German and English, allowing users to switch languages seamlessly. The translation mechanism is implemented using Python’s gettext library.

How It Works

1. Translation Setup

   • The system generates a base.mo file from locale/de/LC_MESSAGES/base.po (automatically during Docker build).
   • The project sets up German translation using gettext.translation and provides a function for dynamic text translation.

2. Language Initialization

   • Streamlit presents a radio button for language selection (default: German).
   • The selected language is stored in the session state and query parameters.

3. Usage in Streamlit

   • Throughout the app, the translation function is used to display text in the user’s chosen language.

---

License

MIT