Scheme-complete provides a single function, scheme-smart-complete,
which you can use for intelligent, context-sensitive completion for
any Scheme implementation. The intelligent completion comes from the
use of simple type inference. In many cases this can drastically prune
the list of possible completions. For example, given the text
(string-ref (n^
where the cursor is represented by ^, typing tab (or whatever you bind the completion function to) would know that in the default R7RS environment the only possible completion of a procedure returning a string and beginning with "n" is number->string and would complete that for you automatically.
Given
(let ((len (string-length str)))
(string-ref str (- ^
completing would fill in "len" as the only possible completion since a number is required as an argument to - and all the standard R5RS bindings are procedures and syntax.
Relying on this completion for known type procedures is a handy way to avoid type errors, even before the compilation. In more general cases it's just nice to have basic pruning, such as not completing syntax in a non-operator position.
Currently completion inside strings does filename completion, though this may be made more flexible in the future.