Project Internship – Intelligent Agents

Setting up the required environment

In the project internship we use Python with multiple packages. To simplify the setup, we provide a Docker-Container based on a prebuilt image containing all necessary dependencies for the project internship. The entire Docker-Container is prepared to cover all three exercises of the project internship, you will not need all packages and software for the first exercise.

Used Software

• Python 3.10 with the following packages
  • spade: Agent framework used to build our agent
  • nltk: Natural language toolkit may be used for preprocessing and other language related tasks
  • gensim: Topic modelling library used to train topic models
  • pyLDAvis: Visualization for topic models
• Additionally, we need the following software for the previously mentioned packages to work
  • Prosody IM XMPP-Server for SPADE

Provided Package

We provide a package, in which all exercises of the project internship may be solved. The packages contains a test scripts to check the setup and also some utility functions/ classes. The package is meant as a structure to fill with the solutions of all three exercises, however, each group may also create its own different structure.

The package is available as zip file from Moodle.

./project
├── docker-compose.yml # contains the configuration for the Docker-Container provided by us
├── run.sh # runs the Docker-Container starting a shell inside it
└── src # the main Python package
    ├── exercise1 # folder for exercise 1
    │   ├── agent_auctioneer.py # stub for exercise 1 
    │   ├── agent_bidder.py # stub for exercise 1
    │   ├── data # a folder containing data for exercise 1
    │   │   ├── corpus.txt
    │   │   └── sell.txt
    │   └── main.py # stub for exercise 1
    ├── test # contains test scripts 
    │   ├── moin_agent.py # one of the agents used by test_agent.py
    │   ├── test_agent.py # runs a simple agent, testing SPADE and the XMPP server
    │   └── visitor_agent.py # second agent used by test_agent.py
    └── utils # contains some util classes and functions
        ├── external.py
        ├── functions.py # write and read (json) files
        ├── random.py # seeded random
        └── wikipedia.py # class downloading articles from wikipedia  (uses a cache-file to speed up recurrent fetches of same article)

The package is meant to run as a Python package called src. If not using Docker, you may have to change the imports, e.g., change the lines 9-13 in ./project/src/test/test_agent.py.

Setup

We strongly recommend to use the Docker-Container provided by us. The container contains all dependencies and is already configured to work with our provided package and source code. However, it is also possible to install the packages on your own and use your own setup.

Setup using Docker

We prepared an Docker-Image on Docker Hub here. Using Docker and Docker-Compose, the Docker-Container starts as a shell inside the prepared environment.

1. Install Docker
   • macOS
   • Linux
   • Windows – Based on WSL 2
2. Download our provided package from Moodle
3. Open a terminal in the folder of the provided package
4. Run the ./run.sh (or docker-compose run ia)
   • It may take some time to download the Docker-Image (ca. 425 MB download, ca. 1.2 GB on disk)
   • A bash shell running inside the container will start
5. Check the container using the scripts in the test section of our provided package
   • Run them inside the container via the shell by
   • ./test/test_agent.py: Tries to run a simple setup of three SPADE agents saying or listening for a “Moin!”
6. Check out the Docker-Container
   • The folder ./project/src/ on the host machine is mounted at /home/user/src/ inside the container
   • I.e., one may create a file in ./project/src/exercise1/my-file.py on the host machine and run it inside the container via python ./exercise1/my-file.py
   • The entire folder src is installed as Python package inside the container
   • Thus, importing the Wikipedia class in the utils directory is possible with from src.utils.wikipedia import Wikipedia

We have tested the Docker-Image and Docker-Container together with the scrips on macOS and Linux.

Attention: As everything is running in a stateless Docker-Container, only changes in the folder /home/user/src/ are stored on the host machine after the container has been terminated!

Manual Setup

Install all needed dependencies manually and run the code from our provided package outside of a Docker-Container.

1. Install Python 3.10
2. Install the needed packages
   • Use a virtual environment if desired
   • pip3 install spade requests nltk gensim pyLDAvis scikit-learn
   • Make sure to download resources for NTLK, e.g., python3 -m nltk.downloader punkt stopwords wordnet omw-1.4
3. Install an XMPP-Server (see XMPP-Server)
4. Download our provided package from Moodle
5. Check the setup using the scripts in the test section of our provided package
   • python3 ./project/src/test/test_agent.py: Tries to run a simple setup of three agents saying and listening for a “Moin!”
     • If the provided package is not installed as package src, change the lines 9-13 in test_agent.py
     • Comment 9,10 and uncomment 12,13