👋 Welcome, Agent_Perry is at your service!

It's your turn: assign an important mission to the most famous platypus on the planet!

About this project

Agent_Perry is an autonomous agent created for the "Autonomous Software Agents" course at University of Trento, Italy.

The goal was to create an agent capable of playing the game Deliveroo.js: the project is open source, check it out on GitHub.

Installation instructions

Note

These instruction will build all the necessary components for the agent. However, a server for the agent is necessary. The latest version used for testing the agent has been forked into the following repository: https://github.com/AgentPerryASA/Deliveroo.js

After downloading the latest available release, run:

npm install

This will install all of the necessay dependencies.

Configuration instruction

The agent requires a .env file that needs to be placed into the root of the project. Such file must contain the following variables:

# Decide whether to use LLM functionality and to have another agent spawn.# This will unlock agent collaboration capabilitiesMATE=<enable or disable>LLM=<enable or disable># Address of the Deliveroo.js serverHOST=<http://localhost:8080 or proper address># Token to make the agent access the server.# Note: the admin token can be generated by inserting admin as a password# in the frontend of Deliveroo.js server. The name MUST also be admin.ADMIN_TOKEN=<admin token>TOKEN1=<token of the first agent>TOKEN2=<token of the second agent># LLM configurationLITELLM_BASE_URL=<address of LLM server>LITELLM_API_KEY=<api key of LLM server>LOCAL_MODEL=<llama-3.3-70b-lmstudio or other model># Key used by a pair of agents to let each other know of the respective presenceHANDSHAKE_KEY=<any string>

Usage instructions

To start the agent, simply run:

npm start

To start and log eventual output, run:

npm run log

To start and test the admin-issued task, run:

npm test

Authors

• GitHub: ErPreco
• GitHub: FireStoat3