Screaming Meemies: Multi-Agent ghosts that gossip, lie and whisper alibis behind your back
This is a submission for the Hacktoberfest Open-Source AI Challenge Week 4: Ghost in the Machine What I Built I interpreted the theme as Halloween horror and ghosts. If we think of ghosts as agents, then ghosts communicating with each other and with people (in an eerie way!) becomes a multi-agent architecture. My inspiration is The Haunting of Hill House, my favorite show ever, by Mike Flanagan…
Screaming Meemies is a Halloween-themed mystery game played in the terminal. The game features four AI agents, each with its own personality, secrets, lies, truth, and memory. The agents include a petty housekeeper, a sarcastic gardener, a snobby butler, and a great-great-aunt who merely wants candy. The game is designed to be played in the terminal, where you interrogate the ghosts and attempt to identify the thief who stole your grandmother's dentures.
Each game presents a different thief, making it challenging to remember the answer. The game uses the Qwen 3.8 27B open-weight model hosted on Groq for its AI agents. PydanticAI was used to manage the agents, with each response structured in a fixed format consisting of an action, a statement, and an optional haunting. The ghosts communicate through a shared channel, allowing them to gossip and whisper secrets to each other, ensuring the game's eerie and mysterious atmosphere.
The game's memory feature allows you to resume a game after a break, with the same thief and the same number of wrong guesses. The game's backend is powered by MongoDB Atlas, enabling the storage of game states and conversations. To prevent leaks, a leak checker was implemented to ensure that the thieves don't reveal their hiding places directly.
The project's open-source nature allows anyone to build upon and modify the code, making innovation accessible to everyone.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
