Skip to main content
Loading...

Dagent Messenger

· 2 min read
Gus
Mascot
Dagent Messenger

Dagent (Dual AGENT) is a two-user chat room messenger for users who are exclusively two artificial agents (i.e., "Dagents") selected by the human user.

The agents are inspired by the non-generative AI agents of the 1960s (e.g., ELIZA). It allows a human user to select, trigger, observe two Dagents messaging each other in the chat room.

Its design draws inspiration from old school instant messenger application (e.g., Yahoo Messenger) and operating systems (e.g., Microsoft Windows 98).

SFX via Vinidoz Synth.

Videos

1960s AI - Mira vs Perry in Dagent Messenger

This is a demo of the Mira and Perry agents chatting in Dagent Messenger. These agents are based on the original 1960s ELIZA chatbot.

Helping Perry Have Better Chats

Perry's AI has been updated to provide more interesting responses in Dagent Messenger

ELIZA vs Mira R-AI chatbots (with rebuilt engines)

The ELIZA agent has been added to behave similarly to the original 1960s ELIZA chatbot. The Dagent agent engine (Dagent v2) has also been rewritten to support new agents, including ELIZA, and other rule-based AI (Dagent R-AI agents) like Mira and Perry. R-AI agents have similar functionality to ELIZA but are now easier to create and customize with unique personalities. These simpler agents will also gain new abilities without compromising their original personalities ion future updates. Since R-AI have the same issue of getting stuck in loops (similar to 1960s AI), the Dagent chat engine was also updated to intervene and redirect chats, which reduces loops whilst keeping agents purer and more faithful to their original personalities and capabilities.

Lex vs Perry Port (classic non-generative R-AI chat)

Perry has been ported to the new engine, and chats with Lex. Both R-AI agents are using classic non-generative to chat to each other in Dagent Messenger.

Comments

Community Rules
This is a community. Please be constructive and polite in your comments and interactions. No toxic negativity.
Comments use GitHub via giscus. By interacting with comments, you consent to GitHub's data processing. See our Privacy Policy for more information.