<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Daniel Abrão on Medium]]></title>
        <description><![CDATA[Stories by Daniel Abrão on Medium]]></description>
        <link>https://medium.com/@danielabro?source=rss-d4a3264e77d2------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*VLYhkH_zAKbo7_7D.jpg</url>
            <title>Stories by Daniel Abrão on Medium</title>
            <link>https://medium.com/@danielabro?source=rss-d4a3264e77d2------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 09:29:11 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@danielabro/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Scriptorium: Building a Contextual Bible Reader with Modern AI Tooling]]></title>
            <link>https://medium.com/@danielabro/scriptorium-building-a-contextual-bible-reader-with-modern-ai-tooling-649863a1d951?source=rss-d4a3264e77d2------2</link>
            <guid isPermaLink="false">https://medium.com/p/649863a1d951</guid>
            <category><![CDATA[architecture]]></category>
            <category><![CDATA[open-source]]></category>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[developer-tools]]></category>
            <dc:creator><![CDATA[Daniel Abrão]]></dc:creator>
            <pubDate>Sun, 08 Mar 2026 22:20:15 GMT</pubDate>
            <atom:updated>2026-03-08T22:20:15.950Z</atom:updated>
            <content:encoded><![CDATA[<p><em>A small open-source project exploring contextual reading, semantic search, and thoughtful AI integration.</em></p><h3>Building Scriptorium</h3><p>Over the past few weeks, I built a small project called <strong>Scriptorium</strong>.</p><p>It’s a contextual Bible reader that combines open datasets, semantic search, and optional AI capabilities to create a more exploratory reading experience.</p><p>But the project isn’t really about building a Bible app.</p><p>Scriptorium is better understood as an <strong>engineering artifact</strong> — a project that captures how I like software to feel: calm, accessible, respectful of users, and built with deliberate architectural boundaries.</p><p>Modern tools made building the project faster than ever. Code assistants handled much of the implementation work, allowing me to focus on system design, product decisions, and how the experience should behave.</p><p>The result is a small but complete system that brings together:</p><ul><li>open biblical datasets</li><li>semantic search with PostgreSQL + pgvector</li><li>optional local AI tooling</li><li>a reading interface designed for exploration</li></ul><p>The project is open source and available here:</p><p><strong>Live application <br></strong>https://scriptorium-5pbj.onrender.com/</p><p><strong>GitHub repository <br></strong>https://github.com/jungleBadger/scriptorium</p><h3>A small project that grew out of curiosity</h3><p>The project’s origin was fairly simple.</p><p>I started reading the Bible recently — not because of a sudden pursuit of faith, but because I wanted to understand the story and its historical narrative better.</p><p>Around the same time, I had a couple of days off work and decided to experiment with modern code assistants and local AI tooling.</p><p>Then something unexpected happened.</p><p>I received an email from someone at Voice.ai who had come across my GitHub and offered a month of API access to experiment with their voice model.</p><p>The message was informal and short — no pitch — just an invitation to try the technology and share feedback.</p><p>That combination of things sparked an idea.</p><p>If I were going to read the Bible anyway, it might be interesting to build a reader that allows <strong>contextual exploration</strong> of the text — something that lets curiosity guide the reading.</p><p>Scriptorium started as that small experiment.</p><h3>The core idea: contextual reading</h3><p>Most digital Bible readers focus primarily on presenting the text.</p><p>Scriptorium tries to go one step further.</p><p>Instead of only reading a passage, the interface allows exploration of <strong>entities and context</strong> connected to the text — people, locations, themes, and references derived from open datasets and enrichment pipelines.</p><p>The goal is to create a reading experience in which curiosity guides navigation through the narrative.</p><p>A passage leads to a name.</p><p>A name leads to a location.</p><p>That context leads to another part of the story.</p><p>Reading becomes a form of exploration.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Enz4AroW4kTgcHwUiVJTOw.png" /><figcaption><em>The Scriptorium reading interface is designed for focused reading with contextual exploration.</em></figcaption></figure><h3>Contextual exploration</h3><p>One of the most interesting parts of the experience is the contextual panel.</p><p>When a passage or verse is selected, the reader can explore:</p><ul><li>summaries of the passage</li><li>people mentioned in the text</li><li>locations referenced in the narrative</li><li>cross-references and contextual connections</li></ul><p>This experience transforms the regular linear flow into something closer to navigating a <strong>knowledge graph</strong>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*NQxkpZ-ipLHe2-Cp1y7rxg.png" /><figcaption><em>Contextual exploration panel showing entities, locations, and AI-generated insights connected to a selected passage.</em></figcaption></figure><h3>Working with open datasets</h3><p>Much of the project’s richness comes from open data.</p><p>Scriptorium integrates multiple publicly available datasets, including:</p><ul><li>Bible translations</li><li>OpenBible resources</li><li>Hitchcock’s Bible names</li><li>additional open reference datasets</li></ul><p>These datasets are already available individually.</p><p>What becomes interesting is how they are <strong>combined, normalized, enriched, and indexed</strong> to support exploration.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*gCS7Yfz4rN_EWksqm7RBsQ.png" /><figcaption><em>Open datasets are processed, enriched, semantically indexed, and surfaced back into the reading experience.</em></figcaption></figure><h3>Architecture: keeping the system simple</h3><p>From an architectural perspective, the system intentionally avoids complexity.</p><p>The browser never communicates directly with external services.</p><p>Instead, the system uses a clear boundary:</p><ul><li>the <strong>browser client</strong></li><li>a <strong>Node.js application acting as the orchestration layer</strong></li><li>supporting data services and pipelines</li></ul><p>This approach simplifies several aspects of the system:</p><ul><li>security and API boundaries</li><li>integration management</li><li>control over external dependencies</li><li>experimentation with optional capabilities</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*HKEGIXXVds643dij1z4N2g.png" /><figcaption><em>Simplified architecture of Scriptorium, where the Node.js application acts as the boundary between the client and external integrations.</em></figcaption></figure><h3>Exploring modern AI workflows</h3><p>Scriptorium also became a small playground for experimenting with modern AI tooling.</p><p>Two areas were particularly interesting.</p><p><strong>Local models</strong></p><p>Running local models with tools like Ollama enabled experimentation with enrichment and processing workflows without relying entirely on external APIs.</p><p><strong>Semantic search</strong></p><p>Using PostgreSQL with pgvector enabled semantic indexing and contextual discovery across passages and entities.</p><p>These capabilities are integrated carefully so they enhance the reading experience without becoming the center of the product.</p><p>The goal was not to build an “AI platform,” but to explore how these tools can be used <strong>thoughtfully and economically</strong>.</p><h3>Accessibility and reading experience</h3><p>One of the most important goals of the project was to create a reading experience that feels calm and adaptable.</p><p>Scriptorium supports:</p><ul><li>keyboard navigation</li><li>screen reader compatibility</li><li>reduced motion options</li><li>responsive layouts</li><li>customizable typography and spacing</li><li>theme switching</li></ul><p>These details may seem small, but they shape how comfortable the reading experience feels over time.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/260/1*9G2RYVRGKbutypbVUaVRKw.png" /><figcaption><em>Reader customization options for typography, spacing, themes, and voice playback.</em></figcaption></figure><h3>Building with code assistants</h3><p>Another interesting aspect of the project was the development workflow itself.</p><p>I built Scriptorium heavily with the help of modern code assistants.</p><p>Rather than focusing on writing every line manually, the workflow shifted toward:</p><ul><li>designing system architecture</li><li>shaping the product experience</li><li>reviewing generated code</li><li>refining the interface and interaction details</li></ul><p>AI-assisted development doesn’t remove the need for engineering judgment.</p><p>It makes those decisions more important.</p><p>Someone still has to decide what the system should be.</p><h3>A different philosophy of product evolution</h3><p>I don’t intend for Scriptorium to evolve like a typical software product.</p><p>Many modern applications prioritize continuous feature expansion.</p><p>This project takes a different approach.</p><p>It should feel closer to something like a <strong>library</strong> — stable, predictable, and trustworthy.</p><p>If the project evolves, the focus will likely be on:</p><ul><li>improving feedback loops</li><li>making it easier to add new translations and languages</li><li>enriching contextual datasets</li></ul><p>The goal is refinement rather than constant expansion.</p><h3>Exploring the project</h3><p>If you’re curious to explore further:</p><p><strong>Live application <br></strong>https://scriptorium-5pbj.onrender.com/</p><p><strong>GitHub repository</strong></p><p><a href="https://github.com/jungleBadger/scriptorium">https://github.com/jungleBadger/scriptorium</a></p><p><strong>Interactive demo</strong></p><p><a href="https://app.arcade.software/flows/3iXoViqSU4oLFxiWI55V">https://app.arcade.software/flows/3iXoViqSU4oLFxiWI55V</a></p><p>Scriptorium is not meant to be a startup.</p><p>It’s simply a small engineering artifact — a project that captures how I like to build software and how modern tooling can be used thoughtfully to create useful experiences.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=649863a1d951" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How about a Developer perspective?]]></title>
            <link>https://medium.com/@danielabro/how-about-a-developer-perspective-1b8ae8aafa65?source=rss-d4a3264e77d2------2</link>
            <guid isPermaLink="false">https://medium.com/p/1b8ae8aafa65</guid>
            <category><![CDATA[developer-stories]]></category>
            <category><![CDATA[ux]]></category>
            <category><![CDATA[agile]]></category>
            <category><![CDATA[front-end-development]]></category>
            <category><![CDATA[freelancing]]></category>
            <dc:creator><![CDATA[Daniel Abrão]]></dc:creator>
            <pubDate>Fri, 23 Mar 2018 05:02:54 GMT</pubDate>
            <atom:updated>2018-03-23T11:12:20.714Z</atom:updated>
            <content:encoded><![CDATA[<p>Much is said about UX these days, and how user-centered design should be applied by all stakeholders of a project or even a given process, and this is awesome, right?!<br>Well, most of the time this is indeed very good and lead to a whole set of new ways of thinking and a bunch of crazy “how come we didn’t think of that earlier” situations.<br>The articles sharing concepts, experiences, and thoughts about UX are plenty, but today I want to talk about another type of experience often left on the side: <strong>the Developer Experience</strong>.<br>Considering the Developer in the process is <strong>not only asking him eventually how the mood is going on</strong> or <strong>if something is or is not doable</strong> in a given time frame.</p><p>I like to consider <strong>Software building as art like painting</strong> and sure I ain’t no Picasso, but I do have the passion! And I do believe that <strong>every Developer should have this proud</strong> on his daily basis tasks.<br>The developers are the ones who go to bed thinking about how to solve something on the project and putting a lot effort on the end user, and there are<strong> plenty of situations that the developer is treated as a cheap labor force</strong> only used to build something together and moving on to the next project. Well, this is kinda of actually true, you join a project, understand it, code it, and it ends sometime. The case is that <strong>we are paid to think and solve problems not only to build some blocks together</strong>, and this is rudely disregarded in some situations.</p><p>Developers often have to deal with malformed mockups with lacking info as at the best scenario, with no comprehension about Responsive guidelines, page effects, animations, transitions, application texts — even internationalized sometimes, which framework, icons guidelines are being used and plenty more examples which I could go on and on <strong>spiced just a bit by the pressure of the deadlines</strong> and being responsible for <strong>making it actually happen.</strong><br>So a kind of a different situation is created which if the developer does not think on the missing parts he will draw himself showcasing the project or he can join a sea of personal opinions which may or may not be great, they are mostly well intended but, again, does not consider the Developer Experience.<br>And this phenomenon starts to happen after some years of Agile methodology as a trend and very often misinterpreted which causes lack of documentation and well-planned processes and omitting certain types of experts specialized in one thing like Design or Test for instance.<br>Besides this multi-functional culture there is a bunch of startups and entrepreneurs which will need a Developer eventually and not care a bit about how you will plan to show off his brand, but a lot of opinions will be thrown, which, again, may or may not be great and mostly disregard the DX side of the story.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/296/1*dmWfH8-jDl7bUOI8O2o0HQ.jpeg" /><figcaption>Yada Yada Yada</figcaption></figure><p>So what can we do to make the ecosystem more healthy and collaborative? I suggest people who make the decisions and gather a team to respect team skill singularity bringing every expertise together and forming the missing ones. There is kind of stereotype that people that like to draw can automatically fill a design role and someone who struggles to code can fill a tester position.<br>Besides most of the times resulting in nothing good long-term speaking because those people not necessarily WANT to follow this career path and move on to something later on it also reflects all said above, making a poor developer experience.</p><p>Let’s <strong>value expertise</strong> and let people be good at <strong>what they feel they should</strong>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1b8ae8aafa65" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Vue.js bundled by Gulp.js + Browserify]]></title>
            <link>https://medium.com/@danielabro/vue-js-bundled-by-gulp-js-browserify-7a125e818a96?source=rss-d4a3264e77d2------2</link>
            <guid isPermaLink="false">https://medium.com/p/7a125e818a96</guid>
            <category><![CDATA[gulp]]></category>
            <category><![CDATA[frontend]]></category>
            <category><![CDATA[vuejs]]></category>
            <category><![CDATA[browserify]]></category>
            <dc:creator><![CDATA[Daniel Abrão]]></dc:creator>
            <pubDate>Wed, 24 Jan 2018 13:08:53 GMT</pubDate>
            <atom:updated>2018-01-24T13:11:41.095Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*A6J1Mq29jrp4oR05mKIgsw.png" /><figcaption>Simple gulp.js task execution</figcaption></figure><p>Homogeneous javascript stack power</p><p>Today we have a variety of tools and frameworks to build up our stack and gain productivity within our projects. The goal of this post is not point which is better or not. My main thought about this subject is that we should always rather learn the language itself before diving in into a full featured framework. If you do that you will most likely enjoy good parts of every framework and detect situations where one is better suited than other.</p><p>Anyway I’ve been working with Vue.js lately and I’m enjoying a lot. It reunited good parts from Angular that I had a little experience at the time and also matches my personal style back when I was coding only VanillaJS. It also supports very well CommonJS, a pattern that I’m a fan.</p><p>I’ve worked with Gulp.js before starting with Vue and the combination of automating things in a node.js environment, working with Node.js on backend and exporting frontend modules with CommonJS`s require/module.exports (❤) was easy peazy. It is just enjoyable to me work with this Stack. So homogeneous.</p><p>Well everything was fine until I noticed soon enough that every tutorial, post, official docs was kinda promoting Webpack which is an awesome tool as well. But I was not willing give up Gulp and I will share some caveats both working with Vue.js apps and Vue.js exportable components in a Build tool perspective.</p><p>First of all we need to load up some dependencies, the list is quite long but you know… Plugins.</p><h4>Gulp plugins</h4><pre>npm install gulp gulp-babel gulp-browserify gulp-cache gulp-cond gulp-envify gulp-eslint gul<br>p-plumber gulp-postcss gulp-rename gulp-sass gulp-sourcemaps gulp-uglify gulp-util gulp-watch — save-dev</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/228/1*dCqLeREmPWn4ApHPmzdxdQ.png" /></figure><h4>Vue dependencies</h4><pre>npm install vue vue-template-compiler eslint-plugin-vue vueify — save-dev</pre><h4>Other dependencies</h4><pre>npm install fs-extra babel-core babel-plugin-transform-runtime babel-preset-env babel-eslint<br> babel-runtime babelify browserify eslint envify cross-env dotenv — save-dev</pre><h4>Putting together</h4><p>Now you need to create a gulpfile.js and set it up. For sake of brevity I will share a Github link to the example file instead of declaring it all here: <a href="https://gist.github.com/jungleBadger/95290b6333aa0409ec17ca55c086bf59">Gist</a>. Declaration are on top of the file.</p><h4>Problem #1 — Applying correct env</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/772/1*zHr5QBo_qp1jA-Lrl8lwlg.png" /><figcaption>Vue.js dev mode warning</figcaption></figure><p>The first problem that I’ve faced was the correct way to declare production/test/beta environment within Vue.js transforms and builds. E.g remove warnings and Vue browser plugin without explicitly setting Vue global properties to false.</p><pre>//js task<br>gulp.task(“js”, function () {<br> browserifyInstance = browserify({<br> “entries”: YOUR_ENTRY_POINT.js,<br> “noParse”: [“vue.js”],<br> “plugin”: argv.w || argv.watch ? [watchify] : [],<br> “cache”: {},<br> “packageCache”: {},<br> “debug”: !isProd<br> }).transform(“envify”, {<br> “global”: true,<br> “NODE_ENV”: process.env.NODE_ENV<br> })<br> .transform(babelify)<br> .transform(vueify)<br> .on(“update”, methods.bundleJS);<br> return methods.bundleJS();<br>});</pre><pre>//bundleJS<br>“bundleJS”: function () {<br> if (isProd) {<br> fse.remove(path.join(modulePath, “dist/js/bundle.js.map”));<br> }<br> browserifyInstance.bundle()<br> .on(“error”, function (err) {<br> gutil.log(err);<br> })<br> .pipe(source(path.join(modulePath, “js/main.js”)))<br> .pipe(cond(!isProd, plumber()))<br> .pipe(buffer())<br> .pipe(rename(“bundle.js”))<br> .pipe(cond(isProd, uglify()))<br> .pipe(cond(!isProd, sourcemaps.init({“loadMaps”: true})))<br> .pipe(cond(!isProd, sourcemaps.write(“./”)))<br> .pipe(gulp.dest(path.join(modulePath, “dist/js/”)));<br> return isProd ? gutil.log(“FINISHED PRODUCTION BUILD”) : gutil.log(“FINISHED DEV BUILD”);<br>}</pre><p>When we run `gulp js` task it will look for the path(s) declared on entries property to resolve requires.</p><p>We are also telling to browserify do not parse vue.js library, and apply watchify plugin if suited (not covered on this topic but you just need to pass a -w flag).</p><p>Cache and packageCache are to handle the hot reload and patch build strategy.</p><p>Now we start applying the first env variable, V<em>ue.js does not accept simply running browserify+gulp passing a PROD/TEST env flag on command line</em>, so we use envify transform applying as a global property and controlling NODE_ENV as a parameter from a command line argument or a environment variable file.</p><p>After envify apply its declaration we then apply <a href="https://github.com/babel/babelify">babelify</a> and <a href="https://github.com/vuejs/vueify">vueify</a>. For specification of these tools please refer to their documentation.</p><p>To finish this routine we just minify it if it is production or generate sourceMaps if test env. It has full support of vue components including scss/stylus transformations.</p><p>Seems straightforward right? Well took me some time to figure this config out. Notice that you will still need the .babelrc and .eslintrc config files.</p><h4>Problem #2 Working with Vue.js runtime only build.</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/566/1*dvEl4DActATP5WhkoJVp3w.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/570/1*9w3gOuJ4ctj7cEUIoDbu-A.png" /><figcaption>Comparing Vue.js bundle builds</figcaption></figure><p><a href="http://(https//vuejs.org/v2/guide/installation.html">Vue.js has two builds</a> and it can be confusing to make it work with the most lightweight version with this stack.</p><p>First of all we need to do a configuration on our `package.json` file and set the browser property as following (assuming that Vue.js installation is done through NPM):</p><pre>//package.json<br>“browser”: {<br> “vue”: “vue/dist/vue.runtime.common.js”<br>}</pre><p>After declaring it on package.json you should set your main application point (declared as entries on js task) should import a root component following the &lt;template&gt;&lt;script&gt;&lt;style&gt; pattern</p><pre> “use strict”;<br> const Vue = require(“vue”);</pre><pre>new Vue({<br> “el”: “#app”,<br> “name”: “admin”,<br> “render”: function (h) {<br>    return h(require(“./components/app.vue”));<br>  }<br> });</pre><p>This will load your component on the anchored el property. The required files can contain another requires from both .js and .vue files and make a complex application.</p><h4>Problem #3 Working with exported .vue components</h4><p>On this one I almost gave up and use webpack. <em>The problem is that I was not able to generate a valid UMD/CommonJS file to be required and used as intended</em>. But turns out it was just a misconfiguration of Browserify and a little tweak make it possible, so now I’m finishing my first component to be exported and distributed.</p><pre>gulp.task(“js”, function () {<br> browserifyInstance = browserify({<br> “entries”: path.join(modulePath, “src/js/main.js”),<br> “noParse”: [“vue.js”],<br> “read”: false,<br> “standalone”: “JngVueChat”,<br> “plugin”: argv.w || argv.watch ? [watchify] : [],<br> “cache”: {},<br> “packageCache”: {},<br> “debug”: !isProd<br> }).transform(“envify”, {<br> “global”: true,<br> “NODE_ENV”: process.env.NODE_ENV<br> })<br> .transform(babelify)<br> .transform(vueify)<br> .on(“update”, methods.bundleJS);<br> return methods.bundleJS();<br> });</pre><p>The only change on this script comparing from the one above is that I&#39;ve added a ‘standalone’ property passing a module name and it this task will make the module available to be exported as UMD/AMD and Browser.</p><p>Hope it can help you guys save some time. Any feedback is appreciated.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7a125e818a96" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>