TECH NOTES

PROXY R&D has focused on the design and implementation of a Java-based Multi-Agent Management (MAM) system, made to interface to a MOO in order to provide a sense of context for agent interaction. The MAM system emphasizes agent personification, and the ability to autonomously and asynchronously fulfill specific tasks while exploring the notion that ongoing interaction does not necessarily require or even benefit from ongoing communication. Agents catalyze community formation by coupling utilitarian function with playfully emergent and unpredictable behavior.

One begins by submitting a request to activate an agent. This process also assigns unique personality attributes that will be keyed to the requested agent. Once profiled, the "phat-client" application interface can be used to ingest data as the play begins. This data can be Web sites, documents, imagery, or virtually anything from your local drive retrievable by a Web browser (or even material that's not). The agent system also automatically facilitates this process on your behalf. For more information see the phat-client tutorial.

CONTENT MANAGEMENT of the agent system includes the filtering, storage, retrieval, and reccomendation of an agent's holdings, whether those holdings be local or remote. Content is anything that can be referenced via a URL, and may range from static items such as documents, sounds, images and applications, to dynamic and real-time sources such as news feeds or other multimedia streams. This content is accessed via a search interface capable of simple keyword queries (see phat-client tutorial). Search results consist of listings of other agents with content relevant to the query.

Ingested Information gets visualized various ways (as ASCII, HTML, JAVA, VRML, or through custom Plugins witten in GL4Java) based upon the participant's preference (see phat-client tutorial). A key role of the agent system is to provide dynamic rendering of content to different users, engaged via multiple modes of interaction.

The most basic mode of connection is possible via Telnet, though not much direct agent-interaction is supported through this method. A more media-rich though far less dynamic connection happens via HTML, though again without the benefit of much direct agent interaction. The phat-client provides the richest interaction, allowing full access to the agent system. With the phat-client, participants can change and add metadata about content, making their agent a better source for queries from other agents.