What I did for the original intermedia search was to store a copy of the indexed content in the site_wide_index table. This makes it simple to do queries and pull out headlines for objects since everything is in one place. This should be working the code is in the intermedia-driver package.
You can see in the search package there is a search-convert-procs.tcl http://cvs.openacs.org/cvs/openacs-4/packages/search/tcl/search-convert-procs.tcl?rev=1.2&view=log which could be modified to allow diffrent services to extract the text. We based our work on Oracle 8i and using intermedia was not easy on that version.
