Agent Chaining

Sometimes a single agent is not enough to accomplish a more complex information mining job, and this control panel lets you set up agents to chain from one to another, based on the results of the online search. You can also use this feature to conditionally start outside software applications when the agent finds information on the web that would justify it. This control allows users to Surf3D operate as forward chaining proposition network of agents.

Magnifying Glass icon on the main display Toolbar: Agent Configurations control panel for start and finish conditions, capture actions, email alerts, and Agent Chaining.

How can I use the agent Chaining feature in the Configuration Panel? (advanced users only)

The button titled Chain opens the Agent Chain and shell execution window. One valuable feature of the software is the ability to rapidly and automatically update shell application programs and change the priorities of subsequent agent module activity by chaining agents. This can act as a rapid update process using live search on the web to provide maximum recency of information to a shell environment applications. This effects better specificity, currency and relevance of information. The chaining and shell application execution permits an additional operational capability at the multi-agent scale.

A collection of agents can be conditionally chained together with different conditional flow logic, in order to accomplish more complex tasks and collection operations than a single agent can provide. Such multi-agents and agents chains comprise a propositional system.

Agents may be chained to and embedded into conventional rule and or framed based inference systems. Am agent can comprise the execution of business rules that represent key application logic such as rules about conditions associated with the diagnosis and corrective maintenance of a product, or rules associated with an transactional exchange system relating to price, quantity and delivery. Such rule based systems already common within industry describe and facilitate specification of rules and rule sets to be applied to support decision making and management of information.

Using the chaining function, agents can be subordinated to other agents, and clusters of agents can operate in a compound manner to derive complex activities. Agents are defined as active objects within an environment. Agents, as active objects, are classed within as different types of agents. Each agent type has associated individual attributes and capabilities. Agents also have starting and finishing states and variables. Agents undergo transitions from one state to another state depending on their interaction with the environment and the behavior engaged. Agents can operate as an aggregate collection.

Agents can be sequenced and associated according to inter-agent rules of interaction and discovered information due to search activity on a network. An individual agent, when launched, engages in an activity without the aid of other agents. An individual agent, upon completion of it's run-time task, can then trigger other agents through chaining. Multiple agents can be triggered from a single agent. Agent outcome states are preconditions to the initiation of other agents based on the Boolean state of the first Agent results. Agent causing other agents to activate is a form of "forward chaining" propagation. The forward chaining is data driven from the data set discovered by an agent during search or retrieval task execution on web or network.

The Chain Execute window provides a means to conditionally chain agents to one another based on the outcome of their search and retrieval activity returning a TRUE or FALSE result. The On TRUE text entry line is where the user can specify another agent to commence activity if the results of the prior agent returned a TRUE result. The On FALSE text entry line is where the user can specify another agent to commence activity if the results of the prior agent returned a FALSE result.

Both the On TRUE and On FALSE text entry lines can individually specify more than one agent to start, or can specify the computer to launch any application program, or combination thereof of agents, application programs, using a batch file.

The Browse buttons associated with each On TRUE and On FALSE sections can be used to capture a file name and path on the users computer that are to be executed. The full file path must be present, not just the name of the agent to be triggered.

When an Agent chains to another agent, it can pass a URL to it. The URL that is passed to it is used as the Starting URL for the new Agent. This is similar to the GET URL command button on the Search set up window (-X- icon on main toolbar), except it gets it from the preceding Agent instead of the Browser.

There are entry lines for adding the URL to the Agent that is being chained. The URL which is passed to the next Agent in the chain can be either a web address or an HTML document located on the local hard disk or in local device memory.

By setting the URL to point at the HTML report for the Agent project that is on the hard disk, a group of URLs that are discovered on the network by one Agent can be passed to another Agent by looking at the Report. To use this feature you must click on Report in the Final Actions section of the Configuration window.

By limiting the size of the Report in terms of items per page and number of pages, a limited number of URLs can be passed between agents. By using the metric sorting, a limited set of URLs with high-ranking value for successive agent chaining can be passed.

If the Chain feature is being used to launch an application instead of an Surf3D agent, the text which holds the URL data for an Surf3D agent is used as the standard command line argument for the application. When an Agent is launched and there is a URL being passed to it, (which will only happen when an Agent is launched via the Chain feature), it forces the AGENT START action to be set to run (use NEW Agent Start checkbox). This insures that it will execute a NEW search while preventing the Open, Continue, and Get URL+New actions. These other Agent Starts are usually inappropriate for a 'chained' agent.

Do chained agents preserve Password protection?

The agent chaining setup accommodates the password protected mode of an Agent being used as a screen saver module. It is configured so that any agent that is chained by another Agent that is password protected will inherit the password and the full screen mode. In other words, if the first Agent in the chain is launched as a screen saver module, all subsequent Agents that may be chained to it will also act as screen saver modules.

Copyright © 2001-2008 Navagent, Inc.