Bruno Silva

Windows Live Agents - Data Sources and Input Matching

After I wrote about getting started with Windows Live Agents, it’s time to go deeper into BuddyScript, the script language that is used in agents development.

To make your code easier to manage and even to make it reusable you can create package files. These are files that aggregate functions, procedures and data sources that are strongly connected. These packages can be referenced in your scripts using the package statement in the beginning of your script file. To create a new package right click you solution and add a new item based on the PKG File template. The name you give to your package file will be also the package name.

I’ve created a package named DataSources.pkg in which I will write sample code related to data sources usage. You can have several kinds of data sources which are listed in the official documentation. Today I’ll talk about 2 of them. First of all the HTTP data source. It Allows you to download data from the web to use in your scripts. When you declare a data source you must give it a name, return type and if you wish you can add parameters. The example bellow is called GetRSSFeedDataSource, has no parameters and the result type is composed by three fields: title, description and link of a RSS feed entry.

In the data source body I define the source file on the web, the timeout for the HTTP request and the expiration time (for caching purposes). You can even define HTTP headers like the user-agent in the example bellow. Now, lets take a look at the source code.

The data that is downloaded by this HTTP data source is parsed using a simple xml filter (there are other like script-based, XSLT, etc). This kind of filter allows you to filter and format your data source results. In this case you can dive into the returned XML and return particular parts of it. This data source only returns the first entry on my RSS feed, but as you can see in the commented code, you can iterate over the XML using loops.

Now lets see an example of usage of my data source.

I defined a procedure where I call the data source, extract the title and link, and send it to the default output device (which will be the Windows Live Messenger conversation window).

Another nice kind of data source that can be configured and used is based on SOAP web services. The example bellow uses a silly web service that I wrote in .NET that sums two numbers and return the result. This data source has 2 parameters (the operands) that are used as input to the web service. I had to define the proxy parameter (URL of the web service), the name of the operation to be performed (and the respective action) and also the namespace of the web service.

SendSum() is a simple procedure that outputs the sum of two parameters. I’ll use it later. The demo web service is packaged with the rest of the project that I made as an example for this post. Download project source code and take a look at it.

Now that we have defined these two data sources, I’ll show you how to use them during a conversation of an agent with an user. The next lines of code were added to the Main.ddl script file. First of all the the reference to the new package, adding the “package DataSources” line to the beginning of the file.

I have chosen to show the first entry of my blog when the user types “news” into the conversation window. When I recognize that pattern, I call the SendLatestRSSEntry procedure that I’ve defined before.

The second example is great to show how simple it is to recognize patterns and split them into variables. The second expression matches with messages that have the format sum <integer> and <integer>. And store the two integer in two variables: OP1 and OP2, which are used as parameters to the procedure SendSum. You can learn more about matching expressions in MSDN.

Feel free to download the project source code and change it at will.