> Correct.  For CHICKEN 5, this is kept in the .egg file rather than
> the .meta file.

> In CHICKEN, we have Chickadee which can search the exports from
> all egg documentation: https://api.call-cc.org/5/doc/
> It even has autocomplete.  Try searching for "call-with-input" and
> you'll get the R5RS standard procedure call-with-input-file,
> the POSIX procedure from CHICKEN core call-with-input-pipe,
> the string port operator in CHICKEN core call-with-input-string
> (which r7rs also has I believe) and the http-client egg's
> call-with-input-request[*].

That's very convenient. IMHO this is already very close to what we need.
We could just adapt Chickadee.

> This is all coming from the wiki, which uses special markup like
>
> <procedure>(call-with-input-pipe cmdline proc #!optional mode)</procedure>

Seems like a reasonable approach, and the markup is lightweight.

> It should be possible to add different backends to Chickadee for parsing
> other kinds of documentation sources.

That would be worth investigating. I continue to advocate an approach
where we have a Scheme API. The server indexes everything, serves it to
the public over HTTP as S-expressions/JSON, and people can use that to
write websites/terminal programs/editor plugins/whatever else they want.
If I have understood correctly, Chickadee would provide both a starting
point for backend work, and an almost complete front-end.