Hello Erik

Good news. 

Getting a frugal Zotero-based citations is *definitely* possible. It is
just a matter of time. Your libraries already provide the necessary
plumbing to accomplish the job. 

Long story
==========

For my own understanding, I tried trimming down zot4rst to it's bare
essentials and the attached zotcite.py is the result.

What zotcite does is this: 

For two Zotero Items, it prints their Bibliogrpahic entries and their
Citation Reference, in "Text" format using Chicago-Author-Date style.

I believe a commandline interface could be built around this file along
the following lines. (Will you be interested in building this interface
for us?)

$ zotcite --style chicago --format <text> --items I1, I2 --print biblio
$ zotcite --style chicago --format <text> --items I1, I2 --print citeref

Once this is done, Emacs/Org can do invoke zotcite and get the required
Bibliographic definitions and references using "shell-command".

The assumption is that Emacs/Org somehow has captured zotero keys
through org-protocol or org-zotero.el or someother means.

Here is a output from zotcite.py.

--8<---------------cut here---------------start------------->8---
$ python zotcite.py
C.UTF-8
C.UTF-8
======== KEYS ========
I4AUIZ4S
AKJZBHRW
======== ITEM_IDs ========
[6, 40]
======== BIBDATA ========
[u'Brin, S. 1999. \u201cExtracting patterns and relations from the world wide web.\u201d The World Wide Web and Databases: 172\u2013183.\n', u'Jambunathan, K. On Choice of Connection-Polynomials for LFSR-Based Stream Ciphers. In Progress in Cryptology \u2014INDOCRYPT 2000, ed. Bimal Roy and Eiji Okamoto, 1977:9-18. Berlin, Heidelberg: Springer Berlin Heidelberg. http://www.springerlink.com/content/n27yjr5eqhabux0g/.\n']
===== CITATION REFERENCE ===== 
(This Brin 1999 That)
--8<---------------cut here---------------end--------------->8---

IIRC, I made two changes to the export.js in your Javascript backend.

1. I had trouble understanding the organization of citation clusters. So
   I modified getCitationBlock as below. Note the use of "true" as a
   second param of the appendCitationCluster. This is OK as
   registerItemIds already does updateItems().

--8<---------------cut here---------------start------------->8---
function getCitationBlock (citation) {
    var results;
    var str = "";
    try {
	results = zotero.reStructuredCSL.appendCitationCluster(citation, true);
    } catch (e) {
	zotero.debug("XXX  oops: "+e);
    }
    // var index = citation['properties']['index'];
    // for (var i = 0 ; i <= results.length ; i++) {
    //     // if (results[i][0] == index) {
    //         return escape(str + results[i][1]);
    //     // }
    // }
    return escape(results[0][1]);
}
--8<---------------cut here---------------end--------------->8---

2. There should be way to set the output format form python side of
   things. I had to modify instantiateCiteProc by hand to set the output
   format to "text". You know what output format that LibreOffice plugin
   uses? I see "text", "html" and "rtf" as output formats. But not ODT.
   See https://bitbucket.org/fbennett/citeproc-js/src/tip/src/formats.js

--8<---------------cut here---------------start------------->8---
function instantiateCiteProc (styleid) {
	// Suspenders and a belt.
	try {
		if (!styleid) {
			styleid = "chicago-author-date";
		}
		if (styleid.slice(0,7) !== 'http://') {
			styleid = 'http://www.zotero.org/styles/' + styleid;
		}
		zotero.debug("XXX does this exist?: " + styleid);
		var style = zotero.Styles.get(styleid);
		zotero.reStructuredCSL = style.csl;
		zotero.reStructuredCSL.setOutputFormat("text");
	} catch (e) {
		zotero.debug("XXX instantiateCiteProc oops: " + e);
	}
};
--8<---------------cut here---------------end--------------->8---

Btw, I found that with my 2-day old zotero database and for printing the
above 2 keys, the whole machinery takes a perceivably a lot of time. Is
this your experience as well? I am not a netbook.