OntoAgent

Overview
Architecture
Example
Installation of the sample
Sample OntoAgent session
Download

Overview

OntoAgent is a small Agent Platform basing on the Semantic Web standards RuleML and RDF Schema. Using our OntoSQL tool, the RuleML rules are converted to SQL views. Using event condition action (ECA) rules, the agent can react to messages from other agents and perform actions, such as sending other messages in turn. 

Architecture

OntoAgent bases on a relational database. Agent derivation rules are mapped to SQL views. ECA rules are implemented using a set of Java modules which eventually call the Command Library. The messaging interface bases on HTTP and Servlet technology. See our paper for details.

Example

The example comes from the domain of document retrieval. We have a simple rule that recommends documents based on keyword matches (rule1.rule). Recommendations are simply printed to the screen (rule3.rule). The idea is that every user is running a document recommendation agent, rather than relying on a single central repository. Every user will also have his or her personal bookshelf. Using a feedback machanism for instance, the users' agents could decide to invoke specific agents for queries relating to specific topics (rule2.rule).

Installation of the sample

We tested the sample using MS SQL Server. Unfortunately it does not run using Access, since transaction support is required.

• Create two logical databases db1 and db2
• Create two ODBC connections called ontoagent8080 and ontoagent8081 to db1 and db2
• execute the SQL scripts (generated by OntoSQL) agent1.sql on db1 and agent2.sql on db2
• insert the test data (data1.sql on db1 and data2.sql on db2)
• launch the agents using the run.bat script in the demo/agent1 and demo/agent2 directories
• point your browser to http://localhost:8080/servlet/ontoagent.MessageIn?subject=http://localhost:8080/servlet/&predicate=searches&object=JDBC to start the interaction

Sample OntoAgent session

The initial request is made via the browser. We are sending a message to Agent1 indicating that Agent1 searches for information on JDBC. The successful processing on the message is acknowledged by the "ok" return message.

This is the console output of agent1. It works against the ODBC datasource ontoagent8080. After the message is received, rule 1 fires causing agent1 to recommend the general document (second to last block of lines). Furthermore, rule 2 fires, which forwards the search request to agent2. As a result of this message, agent2 recommends the specific doc to agent 1 (bottom lines).

This is the console output of agent 2. It receives the forwarded search message from agent 1. It then recommends the specific doc to agent 1.

Download

The download includes

• documentation
• source code
• the small example shown below
• the necessary libraries