The bash offers control over the behavior of autocompletion.
The most primitive example is this (just run it in your bash; if you want it available everywhere, put the complete ... line into your .bashrc):
> complete -W "list of all words for an automatic completion" command_to_be_completed
> command_to_be_completed a<TAB>
all an automatic
With complete you define how the specified command shall be completed. For basic needs, -W (as in "word list") should be enough, but you may also specify a function, a glob pattern and many more. complete -p gives you a list of currently defined autocompletions. Behold, thou might not define multiple completions for one command.
I recently built a script that takes a project name and then boots a development environment. The project name is taken from a directory holding all projects, so I created the following completion to save tedious project-spelling:
complete -W "$(ls ~/dev/projects )" devenv
Resources:
