cruncher is a core locus program. It creates your database index.
document part is a part
fulltext database is a database to which you put texts too long to read/remember and then ask which words are in them and where.
grazer is a core locus program. It searches in the database index and outputs (names and parts of) found documents.
query file (also called "input file") is a text file specifying what should grazer search for. It's optional - simple and/or one-time searches can be passed as command-line parameters.
locus.opt is a text file with database parameters (i. e. in which directories it is). cruncher and grazer expect this file in their working directory.
metrics measure how well a particular document matches some criterion. Treir relative importance defines soft operator.
soft operator maps a list of words to list of documents. You define and tweak them to find exactly what you looking for - i. e. "person's name", "exact term", "general topic".
templates specify formatting of grazer output. Each well-defined, regularly repeating piece of grazer output (i. e. document name) has its template string. Templates are similar to format strings for printf(), but they use numbered arguments (every template has its own list of them). Every template argument can be used multiple times or not at all. For example template for document name can be "%0\n" (just a file name), "%1File %0\n" (string in the appropriate color) etc.
user-defined objects file (also called "user-named objects file", "named objects file") is a text file with the same format as locus.opt (in fact, it can be locus.opt, if you name your objects differently from standard sections). You don't need it untill you have user-defined objects (soft operators and/or document parts).
wordlist is a list of all indexed words. Study it to understand and/or optimize your database.