<?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 Usiere Moses Akpabio on Medium]]></title>
        <description><![CDATA[Stories by Usiere Moses Akpabio on Medium]]></description>
        <link>https://medium.com/@akpabio98?source=rss-886f8b80b243------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*m8JTeoLeTdNhGNDJKwWtcQ.jpeg</url>
            <title>Stories by Usiere Moses Akpabio on Medium</title>
            <link>https://medium.com/@akpabio98?source=rss-886f8b80b243------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 08:52:44 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@akpabio98/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[Five Agents, One Platform: Building Domain-Specialised RAG for 1,000+ Daily Researchers]]></title>
            <link>https://medium.com/@akpabio98/five-agents-one-platform-building-domain-specialised-rag-for-1-000-daily-researchers-c7e479d98c14?source=rss-886f8b80b243------2</link>
            <guid isPermaLink="false">https://medium.com/p/c7e479d98c14</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[data-engineering]]></category>
            <category><![CDATA[rags]]></category>
            <category><![CDATA[large-language-models]]></category>
            <dc:creator><![CDATA[Usiere Moses Akpabio]]></dc:creator>
            <pubDate>Tue, 28 Jul 2026 19:02:45 GMT</pubDate>
            <atom:updated>2026-07-28T19:02:45.845Z</atom:updated>
            <content:encoded><![CDATA[<p><em>How we replaced the idea of one all-knowing chatbot with five specialised agents, and what it took to make the whole thing survive production.</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*odxb4u7JIzjTjBrjaSx-jg.png" /></figure><p>Most RAG tutorials end where the real work begins. You chunk some documents, embed them, wire up a retriever and an LLM, and you have a demo that answers questions impressively for about a week. Then real users arrive with real queries, and the cracks appear.</p><p>This is the story of building an AI platform for the African Population and Health Research Center (APHRC), a research institution whose staff and partners needed conversational access to a sprawling knowledge estate: an institutional repository of reports and policy briefs, a corpus of academic publications, a database of grants and projects, an expert directory, and the platform’s own documentation. The system now serves over 1,000 researchers daily.</p><p>The central design decision, and the argument of this article, is that we did not build one RAG pipeline. We built five.</p><h3>The problem with one pipeline</h3><p>On paper, the task looks like a single problem: index everything, retrieve relevant chunks, generate answers. In practice, the five knowledge domains had almost nothing in common beyond being text.</p><p>The institutional repository held long-form reports and policy briefs organised in a deep community and collection hierarchy, with a meaningful subset of access-restricted material. The publications corpus was academic papers enriched with external metadata such as author identifiers and citation context. The grants domain was structurally different again: projects with budgets, phases, and links into a live ERP system, where the freshest answer matters more than the most semantically similar one. The expert directory and platform help content were smaller, but their queries had a completely different shape: “who works on maternal health in this region” is not a document retrieval problem in the same sense as “summarise our urbanisation research”.</p><p>A single pipeline forces a single set of compromises. One chunking strategy, one metadata schema, one retrieval tuning, one prompt style. Tuning that helps papers hurts grants. A recency boost that makes budget queries accurate makes literature queries worse. We tried to imagine the unified schema that would carry a budget phase, a repository hierarchy path, a citation record, and a staff profile at once, and concluded it would be the worst of all worlds.</p><p>So each domain got its own agent: its own vector store schema, its own chunking and indexing strategy, its own retrieval tuning, and its own prompt strategy. The grants agent carries project and budget metadata on every chunk. The repository agent carries hierarchy paths and access flags. The publications agent leans on enriched academic metadata. None of them has to pretend to be the others.</p><h3>Routing without a hard wall</h3><p>Five agents create an obvious new problem: users do not know or care about your architecture. Someone will ask the grants assistant about publications, and the wrong answer here is a dead end.</p><p>We solved this with a deliberately lightweight routing layer. Each agent declares the scope of queries it handles, expressed as keyword signals. When a query lands outside an agent’s scope, the user does not get an error or, worse, a confidently wrong answer stitched from the wrong index. They get a soft redirect: a brief, polite pointer to the agent that can actually help, along the lines of “for publications questions, try the research chatbot”.</p><p>Two things make this pattern better than it sounds. First, it is honest. A specialised agent declining gracefully beats a generalist hallucinating, and users learn the shape of the system quickly. Second, every redirect is logged. Those logs became an analytics goldmine: they tell you where users expect capabilities to live, which agent boundaries are drawn in the wrong place, and which cross-domain questions are common enough to deserve first-class support. The routing layer started as plumbing and ended up as a product research instrument.</p><p>I would push back on the instinct to reach for a heavyweight orchestration framework here. Our router is configuration and logging, not an LLM deciding among tools on every request. It is cheap, debuggable, and fast, and for a system with well-separated domains, that is usually enough.</p><h3>The war story: getting to 100% repository coverage</h3><p>Every production system has one integration that fights back. Ours was the institutional repository, which runs on DSpace 8.</p><p>The naive approach, walking the paginated items API, quietly missed a large slice of the collection. Items lived in orphaned collections, deep community hierarchies, and corners of the system that no single API surface exposed. Our first pass indexed 364 items. We knew the true figure was substantially higher, and a knowledge platform that silently omits a third of the institution’s output is arguably worse than no platform at all, because users trust its silence.</p><p>Getting to full coverage took a five-layer discovery strategy, with each layer catching what the previous ones missed:</p><ol><li><strong>Paginated item listing</strong>, the standard walk, as a baseline.</li><li><strong>Community and collection crawling</strong>, descending the full hierarchy so that items were discovered through their organisational structure, not just the flat list.</li><li><strong>Discovery search sweeps</strong>, using the search API with broad queries to surface items the structural crawl missed.</li><li><strong>An orphaned collection sweep</strong>, targeting collections detached from the main hierarchy.</li><li><strong>Direct handle fallback</strong>, resolving known identifiers individually for the stubborn remainder.</li></ol><p>That strategy lifted coverage from 364 items to 658, the full corpus. The lesson generalises well beyond DSpace: when completeness matters, do not trust any single enumeration path of a complex system. Layer independent discovery methods and reconcile the union.</p><p>Authentication was its own small saga. DSpace 8 uses JWT with CSRF tokens rather than the older basic auth, with a session identifier that has to be extracted from an XML response, and tokens that expire mid-crawl. The fix was unglamorous and effective: re-authenticate every 50 items, unconditionally, rather than detecting expiry and recovering after a failure. Boring, predictable code beats clever error recovery in a long-running ingestion job.</p><p>The repository also contained access-restricted material, and this shaped one of the design decisions I would defend most strongly. Restricted items are indexed, but every chunk carries an access flag, and filtering happens inside retrieval itself, not as a post-processing step on generated answers. Filtering after generation means restricted content has already reached the LLM’s context, and you are one prompt injection or one careless summary away from a leak. Pushing the boundary down into retrieval means restricted content never enters the context of an unauthorised query in the first place. Enforce access control at the lowest layer that can express it.</p><h3>Surviving 1,000 researchers a day</h3><p>The gap between a working demo and a production system is mostly invisible in an architecture diagram. For us it came down to three things: caching, rate limiting, and boring operational discipline.</p><p><strong>Caching is the difference between a viable cost model and an alarming one.</strong> We cache at three tiers: full responses for repeated requests, embeddings so that identical text is never embedded twice, and top-K retrieval results for common query shapes. Together these cut LLM API spend by roughly 70% compared with the uncached baseline. In a nonprofit research context, that is not an optimisation, it is the condition under which the system is allowed to exist. Redis handles the distributed caching and search layer, with a local vector index doing the heavy similarity work.</p><p><strong>Rate limiting keeps one enthusiastic script from becoming everyone’s outage.</strong> Anonymous users get a modest hourly allowance, registered users a much larger one. Embedding calls to the upstream provider are rate-limited with exponential backoff, because provider limits are a fact of life and retry storms make everything worse.</p><p><strong>The unglamorous machinery does the rest.</strong> Sessions carry a 24-hour lifetime with lightweight context tracking. Nightly pipelines handle re-indexing, metadata enrichment, and analytics rollups, so ingestion load never competes with daytime traffic. Health checks and automatic restarts cover the failure modes nobody wants to be paged for. A recommendation layer analyses query history to suggest publications and topics, with short-lived caches invalidated as new queries arrive, which quietly increased engagement without a single prompt change.</p><p>None of this is novel. All of it is necessary.</p><h3>What broke in production</h3><p>Two incidents taught us more than any design review.</p><p>The first was a database constraint colliding with reality. A check constraint on a feedback table enforced a fixed set of values for a field that the frontend, after a later change, submitted as free text. The result was a stream of 500 errors on a feature users cared about. The fix involved backing up the affected data, removing the constraint, and verifying integrity afterwards. The deeper lesson: validation rules baked into the database and validation rules assumed by the frontend will drift apart unless something actively ties them together. Schema constraints are code, and they need the same change discipline as code.</p><p>The second was quieter: data quality debt. A meaningful share of indexed items had sparse metadata and correspondingly poor retrieval quality scores. Separately, duplicate chunks had accumulated through repeated ingestion runs, and cleaning them up reclaimed roughly 26% of storage. Neither problem announced itself. Both degraded answer quality in ways users experienced but could not articulate. The remediation, rescoring sparse items after enrichment and deduplicating chunks, was straightforward. Catching it earlier would have been better.</p><h3>What I would do differently</h3><p>An honest case study owes you this section.</p><p><strong>Ingestion validation should have existed from day one.</strong> The quality rescoring and duplicate cleanup were both symptoms of the same gap: we measured retrieval quality carefully but let content enter the index optimistically. A validation gate at ingestion, checking metadata completeness and detecting duplicates before indexing, would have prevented both incidents. Retrofit cost always exceeds upfront cost here.</p><p><strong>Dual LLM backends bought less than they cost.</strong> We ran the system so it could switch between two model providers. The optionality felt prudent, and it did provide negotiating comfort, but it doubled the prompt tuning surface, complicated evaluation, and meant every generation change was tested twice. If I built it again, I would pick one provider, keep the abstraction layer thin enough to swap later, and spend the saved effort on evaluation instead.</p><p><strong>Redirect analytics deserved first-class treatment sooner.</strong> We treated redirect logs as a byproduct for months before realising they were the clearest signal of user intent we had. Instrumenting the boundaries between agents from the start would have shaped the roadmap earlier.</p><h3>The takeaway</h3><p>If your knowledge estate has genuinely different domains, with different metadata, different freshness requirements, and different query shapes, resist the pull of the single unified pipeline. Specialised agents with their own schemas and tuning, joined by a routing layer that redirects gracefully and logs everything, gave us better answers in every domain than any compromise pipeline we sketched.</p><p>And when you get there, remember that the architecture is maybe a third of the work. The rest is coverage you can prove, access control enforced at retrieval, caching that makes the economics work, and the humility to instrument what breaks.</p><p><em>I’m a software engineer and AI systems architect working on RAG systems, multi-agent architectures, and data platforms for health and development organisations. You can reach me on Medium or at akpabio98[at]gmail.com.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c7e479d98c14" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Excitement, Uncertainty, and the AI Shift.]]></title>
            <link>https://medium.com/@akpabio98/excitement-uncertainty-and-the-ai-shift-b186904539c3?source=rss-886f8b80b243------2</link>
            <guid isPermaLink="false">https://medium.com/p/b186904539c3</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[jobs]]></category>
            <category><![CDATA[future-of-work]]></category>
            <category><![CDATA[gpt]]></category>
            <dc:creator><![CDATA[Usiere Moses Akpabio]]></dc:creator>
            <pubDate>Mon, 18 Aug 2025 20:14:46 GMT</pubDate>
            <atom:updated>2025-08-18T20:14:46.761Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*_vP3xxrHnecEGT55" /><figcaption>Photo by <a href="https://unsplash.com/@solenfeyissa?utm_source=medium&amp;utm_medium=referral">Solen Feyissa</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>This is a thought that’s been on my mind lately, and I felt it was worth sharing because it may help someone navigate the changes happening around us.</p><p>Over the past few years — and especially now — two words keep showing up in conversations about the future: <strong>excitement</strong> and <strong>uncertainty</strong>. The topic at the center of this tension is the rise of <strong>artificial intelligence</strong> and its integration into our daily work.</p><p>On one hand, people see AI as a wave of endless opportunities. On the other, many worry it’s going to take away their jobs. These debates are almost perfectly balanced, as though we’re standing at the edge of something massive — half terrified, half thrilled.</p><h3>Beyond Job Loss or Job Creation</h3><p>The truth is that both perspectives are valid, but they only scratch the surface. What we are experiencing now is just the beginning. A lot is going to change — industries, processes, systems, and even the very structure of work.</p><p>Think about it: most of today’s job structures still reflect the industrial era — <strong>9-to-5 schedules, hierarchical organizations, and siloed specialized roles.</strong> But AI doesn’t operate that way. It is fluid, adaptive, and boundary-crossing. This means our entire framework of work will inevitably shift.</p><p>The real story isn’t just about <strong>jobs being replaced</strong>; it’s about <strong>jobs being transformed</strong> on a scale we’ve never seen before. And that’s where the opportunity lies.</p><h3>Staying Ahead: Curiosity Over Anxiety</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*jRywRzXpUnlhgXgM" /><figcaption>Photo by <a href="https://unsplash.com/@josephandjosephandjoseph?utm_source=medium&amp;utm_medium=referral">Joseph Rosales</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>So, how can someone avoid being left behind? The key is not to panic, but to <strong>stay curious</strong>. Instead of obsessing over predictions or anxiously asking, <em>“What’s going to happen to me?”</em>, the better question is: <em>“How does this work?”</em></p><p>Here are some practical ways to do that:</p><ul><li><strong>Experiment with AI tools.</strong> Play around with ChatGPT, Claude, APIs, Automation tools, AI niche products, or others — not just as users, but to understand how they function.</li><li><strong>Study the ecosystem.</strong> Watch the startups being built around AI. Ask yourself: <em>What problems are they solving?</em></li><li><strong>Follow the money.</strong> Pay attention to what venture capitalists and companies are investing in — it’s often a signal of where industries are heading.</li><li><strong>Look beyond your field.</strong> Read about automation in sectors you’ve never considered. Patterns often repeat in surprising places.</li></ul><p>This isn’t about becoming an AI expert overnight. It’s about building a mindset of curiosity and understanding the principles shaping this shift.</p><h3>The Takeaway</h3><p>History may not repeat itself, but it does rhyme. And one lesson we can draw is that in times of uncertainty, the people who thrive are rarely the anxious — they are the curious.</p><p>So my simple advice is this: <strong>move with the trends by understanding them deeply.</strong> Curiosity, not fear, will define who gets ahead in the age of AI..</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b186904539c3" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The AI Advantage: Unlocking Growth and Creativity in the Future of Work]]></title>
            <link>https://medium.com/@akpabio98/the-ai-advantage-unlocking-growth-and-creativity-in-the-future-of-work-6247bce37253?source=rss-886f8b80b243------2</link>
            <guid isPermaLink="false">https://medium.com/p/6247bce37253</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[careers]]></category>
            <category><![CDATA[future-of-work]]></category>
            <category><![CDATA[learning]]></category>
            <category><![CDATA[software-development]]></category>
            <dc:creator><![CDATA[Usiere Moses Akpabio]]></dc:creator>
            <pubDate>Wed, 13 Mar 2024 08:55:13 GMT</pubDate>
            <atom:updated>2024-03-13T08:55:13.992Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*e60nTiD9rfBoa0fS" /><figcaption>Photo by <a href="https://unsplash.com/@andrewtneel?utm_source=medium&amp;utm_medium=referral">Andrew Neel</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>In the ever-evolving landscape of technology, the discourse around artificial intelligence (AI) and its impact on the workforce has become a common theme. Conversations often revolve around the fear of job displacement as routine tasks become automated. While there is truth to the idea that AI will transform the nature of work, it’s essential to view these changes as opportunities for growth rather than threats.</p><p>The saying “standing on the shoulders of giants” applies aptly to the realm of technology. As new advancements arise, they serve as stepping stones for progress. Rather than expending energy on repetitive tasks, individuals can leverage existing solutions to streamline processes and free up time for more meaningful and creative endeavors.</p><blockquote>AI transforms work; embrace growth, stand on tech giants’ shoulders</blockquote><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fgiphy.com%2Fembed%2F3oz8xJSojoUrBcR5Ic%2Ftwitter%2Fiframe&amp;display_name=Giphy&amp;url=https%3A%2F%2Fgiphy.com%2Fgifs%2Fcolbertlateshow-3oz8xJSojoUrBcR5Ic&amp;image=https%3A%2F%2Fmedia0.giphy.com%2Fmedia%2Fv1.Y2lkPTc5MGI3NjExaTQ1ZXBvdzVtZTd2azUzMTBteWo0eDlld2F2cXcxb283aDNhMnk4diZlcD12MV9pbnRlcm5hbF9naWZfYnlfaWQmY3Q9Zw%2F3oz8xJSojoUrBcR5Ic%2Fgiphy.gif&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=giphy" width="435" height="244" frameborder="0" scrolling="no"><a href="https://medium.com/media/28b6861558f57ab47ddf60665f6cac98/href">https://medium.com/media/28b6861558f57ab47ddf60665f6cac98/href</a></iframe><p>The greater emphasis we place on less mundane tasks, the more productive we become and the more we aspire for improvement. While watching the movie “Hidden Figures,” which narrates the story of three African-American female mathematicians, it was fascinating to realize that, in the past, calculators were actually humans sitting in a room performing calculations before the introduction of the first IBM machine. Initially, this raised concerns, but soon the more crucial roles received increased attention and responsibilities.</p><p><em>Let analyse this applicability in 2024;</em></p><p>Consider the analogy of a software developer using a tool that simplifies coding from wireframes. This not only enhances efficiency but also allows the developer to dedicate more time to logic and innovation. The key is to identify routine tasks that consume time without yielding significant rewards and explore how AI can assist in those areas.</p><p>Imagine you are a marketing professional, and your daily routine involves analyzing heaps of data to create reports and charts. It’s a tedious task that consumes a significant portion of your time. By applying the principles of AI integration, you identify this data analysis as a potential area for improvement. You explore AI tools that specialize in data analytics and discover how they can automate the process, leaving you with more time to interpret the results creatively. The outcome? Your reports become more insightful, and you have the bandwidth to develop innovative marketing strategies.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fgiphy.com%2Fembed%2FQNv4mKyspa8j6%2Ftwitter%2Fiframe&amp;display_name=Giphy&amp;url=https%3A%2F%2Fgiphy.com%2Fgifs%2Fmad-men-madmen-peggy-dondraper-don-sterling-cooper-cannes-oneshow-dandad-clio-droga5-QNv4mKyspa8j6&amp;image=https%3A%2F%2Fmedia4.giphy.com%2Fmedia%2Fv1.Y2lkPTc5MGI3NjExenZ0bWVpdXE4bTNxaGplbzQ3cHQ0aGlma3dsZGl4c3dwaTQ5eXFzcSZlcD12MV9pbnRlcm5hbF9naWZfYnlfaWQmY3Q9Zw%2FQNv4mKyspa8j6%2Fgiphy.gif&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=giphy" width="435" height="243" frameborder="0" scrolling="no"><a href="https://medium.com/media/3f6c6459c1747b087d5e4312defed237/href">https://medium.com/media/3f6c6459c1747b087d5e4312defed237/href</a></iframe><p>Or consider a project manager overwhelmed with scheduling tasks, emails, and routine communications. By identifying these as redundant and time-consuming aspects of the workflow, the manager decides to leverage AI-powered project management tools. These tools help automate scheduling, streamline communication, and even predict potential roadblocks. As a result, the project manager becomes more efficient, focusing on strategic planning and problem-solving, ultimately contributing to the success of the projects.</p><p><em>All these are testaments to how AI can truly transform our entire workflow, but how then can we address it if starting from the basics?</em></p><p>Incorporating this concept into our personal workflow isn’t straightforward initially, but with deliberate structure and planning, it becomes more manageable. To seamlessly integrate AI into your workflow, we can adhere to these steps:</p><figure><img alt="A visual representation of the feeback loop to become proficient in integrating AI to our workflows." src="https://cdn-images-1.medium.com/max/1024/1*5tEMBKuJtDlZAdihDICqJA.jpeg" /><figcaption>A visual representation of the feeback loop to become proficient in integrating AI to our workflows.</figcaption></figure><ol><li><strong>Learn (start with Basics): </strong>Acquire fundamental knowledge of AI by taking a recommended course to understand its capabilities and applications. https://www.coursera.org/learn/introduction-to-ai, <a href="https://www.coursera.org/learn/introduction-to-generative-ai">https://www.coursera.org/learn/introduction-to-generative-ai</a></li><li><strong>Identify Your Workflow: </strong>Assess your daily tasks and pinpoint areas that are repetitive, time-consuming, and less enjoyable.</li><li><strong>Draw a Persona for Creativity:</strong> Envision the person you could become if freed from mundane tasks, focusing on more creative and impactful work.</li><li><strong>Explore Automation Opportunities:</strong> Investigate existing tools, such as ChatGPT, Gemini, GitHub copilot, IBM watson, etc, to automate tasks that align with your identified areas of improvement.</li><li><strong>Transform into an AI Hero:</strong> Implement AI solutions strategically, enhancing your productivity and creativity in the process.</li><li><strong>Regular Assessments: </strong>Conduct quarterly or monthly evaluations to identify new opportunities for AI integration and continuous improvement.</li></ol><p>The transition to becoming your better self in the age of AI is a journey that requires proactive steps. Embrace the transformative power of AI, turning potential challenges into opportunities for personal and professional growth. As technology evolves, so too can our ability to harness its capabilities for the betterment of our work and creativity.</p><p><em>After digesting this information, should you find yourself feeling overwhelmed, I’ve prepared a sample Notion template to facilitate the initiation of this process smoothly. If you’d like access to the template, kindly reach out to me via </em><a href="/service/mailto:akpabio98@gmail.com"><em>email</em></a><em>.</em></p><p>I acknowledge that AI may present challenges in the future, and I’m open to discussing them. While I have some reservations, I also recognize the significant positive impact it brings to our daily workflow in the workforce.</p><p>The concept of AI has long been intriguing, but its significance has heightened, especially in 2024 and beyond. While concerns may arise, the positive contributions it offers are equally significant.</p><p>Let me know your thoughts on this.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6247bce37253" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Why Architecture is Crucial in Software Applications.]]></title>
            <link>https://medium.com/@akpabio98/why-architecture-is-crucial-in-software-applications-ac41203c17b8?source=rss-886f8b80b243------2</link>
            <guid isPermaLink="false">https://medium.com/p/ac41203c17b8</guid>
            <category><![CDATA[programming]]></category>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[python]]></category>
            <category><![CDATA[web-development]]></category>
            <category><![CDATA[vuejs]]></category>
            <dc:creator><![CDATA[Usiere Moses Akpabio]]></dc:creator>
            <pubDate>Mon, 12 Feb 2024 03:57:42 GMT</pubDate>
            <atom:updated>2024-02-12T03:57:42.218Z</atom:updated>
            <content:encoded><![CDATA[<p>Embarking on the journey of creating a groundbreaking web project is an exhilarating experience for any developer.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*O7un9ucffuKjbTJt" /><figcaption>Photo by <a href="https://unsplash.com/@nandovish?utm_source=medium&amp;utm_medium=referral">Fernando Brasil</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>The enthusiasm to incorporate numerous features is palpable, but a common challenge emerges as the project evolves — increasing complexity. As features pile up, pinpointing and addressing individual issues becomes increasingly daunting. Many developers have faced this dilemma, and while solutions like refactoring exist, the question remains: does it truly resolve the underlying problem?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*NwudJJ8ef6Hi1bLt" /><figcaption>Photo by <a href="https://unsplash.com/@brett_jordan?utm_source=medium&amp;utm_medium=referral">Brett Jordan</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>The organization of your application holds the key to its long-term structure and maintainability. Although it may seem elementary at first glance, the way you structure your application can profoundly impact its ultimate outcome.</p><p>Exploring the concept of architecture, I will delve into the modular architecture explanation as a transformative technique. Picture an open office space with three sub-departments but lacking compartmentalization or segmentation. The potential chaos is evident. Now, envision an office space where each department is meticulously organized and designated. The difference is striking and extends beyond mere aesthetics.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*Zti9MT5ZZRKzjPAO" /><figcaption>Photo by <a href="https://unsplash.com/@frantic?utm_source=medium&amp;utm_medium=referral">Alex Kotliarskyi</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>In the realm of software applications, adopting modular architecture involves categorizing functionalities or features into distinct modules or folder sets. This could involve Module A, Module B, and so forth.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/653/1*7YOM39K9udl76bDlqqLYkw.png" /></figure><p>Let’s illustrate this with an example using a Vue.js application:</p><p>Consider a Vue.js application for an e-commerce platform. Without a proper architecture, all components related to product display, shopping cart, and user authentication might be intertwined, making it challenging to update or modify one aspect without affecting the others. However, by employing modular architecture, you can have separate modules for Product Display, Shopping Cart, and User Authentication. Each module functions independently, allowing developers to enhance or modify specific features without disrupting the entire application.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/311/1*Au5KHoBbSb9TzfLLWS1eTA.png" /><figcaption>Folder structure showing a modular application in vueJs</figcaption></figure><p>Let’s illustrate this with an example using a Python application:</p><p>Imagine developing a Python application for data analysis. Without proper architecture, various tasks such as data collection, processing, and visualization might be intertwined in a single script, leading to a convoluted and hard-to-maintain codebase. However, with modular architecture, you can create separate modules for Data Collection, Data Processing, and Data Visualization. Each module operates independently, allowing developers to enhance or modify specific data-related tasks without affecting the entire analysis pipeline.</p><p>Consider the scenario where you need to update the data processing logic. With modular architecture, you can focus solely on the Data Processing module, ensuring that changes made here don’t impact the data collection or visualization components.</p><p>By adhering to architectural principles, developers can compartmentalize and organize features independently. This approach prevents interference and the ensuing chaos that can arise when various features intertwine.</p><p>In conclusion, the implementation of architectural principles is paramount, allowing developers to develop features independently and maintain order amidst the complexity of a growing project.”</p><blockquote>Architectural elegance is not merely a design choice; it’s the compass guiding developers through the evolving landscape of complexity, enabling them to shape groundbreaking projects with order and maintainability.</blockquote><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ac41203c17b8" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Basic Form Validation using Vue.js: A Simple Guide]]></title>
            <link>https://javascript.plainenglish.io/basic-form-validation-using-vuejs-ad9489628617?source=rss-886f8b80b243------2</link>
            <guid isPermaLink="false">https://medium.com/p/ad9489628617</guid>
            <category><![CDATA[javascript]]></category>
            <category><![CDATA[coding]]></category>
            <category><![CDATA[vuejs]]></category>
            <category><![CDATA[programming]]></category>
            <category><![CDATA[frontend-development]]></category>
            <dc:creator><![CDATA[Usiere Moses Akpabio]]></dc:creator>
            <pubDate>Fri, 22 Oct 2021 06:05:27 GMT</pubDate>
            <atom:updated>2021-10-22T16:26:36.915Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*kRjuRbZ8uGE-wt0HoweVNg.png" /></figure><p>When developing web applications that require users to fill in forms, it is highly essential to perform validations for the data that’s entered as input before submission. This helps in ensuring that the needed data is received in the right format.</p><p>Form validation can be performed both on the frontend and the backend. Though in most cases, basic form validations such as ‘ensuring that all fields have been entered as input’ can be done only on the frontend.</p><p>The major benefit of validation on the frontend is to give the user immediate feedback on the data that was entered.</p><p>In this article, I will explain how to perform basic form validation on the frontend using<strong> Vue.js</strong>.</p><p>I shall create a simple web form page with 3 fields: ‘First name’, ‘Last name’ and ‘Email’. I will be using Bootstrap for my input fields.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*V62127Xg4KJRShaXb1v3-w.png" /></figure><pre>&lt;template&gt;</pre><pre>&lt;div&gt;</pre><pre>&lt;div class=&quot;main&quot;&gt;</pre><pre>&lt;h1&gt;Kindly fill in your details&lt;/h1&gt;</pre><pre>&lt;b-form-input v-model=&quot;firstName&quot; placeholder=&quot;First name&quot; class=&quot;input&quot;&gt;<br>&lt;/b-form-input&gt;</pre><pre>&lt;b-form-input v-model=&quot;lastName&quot; placeholder=&quot;Last name&quot; class=&quot;input&quot;&gt;<br>&lt;/b-form-input&gt;</pre><pre>&lt;b-form-input v-model=&quot;email&quot; placeholder=&quot;Email&quot; class=&quot;input&quot;&gt;<br>&lt;/b-form-input&gt;</pre><pre>&lt;button class=&quot;btn1&quot; @click=&quot;submit()&quot;&gt;Submit&lt;/button&gt;</pre><pre>&lt;/div&gt;&lt;/div&gt;</pre><pre>&lt;/template&gt;</pre><p>For the data, the ‘Errors’ array is where the errors will be stored. While ‘firstname’, ‘lastname’, and ‘email’ will be received from the input data.</p><pre>&lt;script&gt;</pre><pre>export default {</pre><pre>name: &quot;&quot;,</pre><pre>data() {</pre><pre> return {</pre><pre>   errors: [],</pre><pre>   firstName: &#39;&#39;,</pre><pre>   lastName: &#39;&#39;,</pre><pre>   email: &#39;&#39;,</pre><pre>}}}};</pre><pre>&lt;/script&gt;</pre><h3>Steps for Validation</h3><h4>For basic form validation, there are 3 important steps, namely;</h4><ul><li>Getting the errors.</li><li>Looping through the errors and displaying them.</li><li>Success logic.</li></ul><h4>1. Getting the errors</h4><p>This is a very important step. In the ‘Errors’ array, you will want to store the errors as they appear.</p><p>This will be achieved when the user clicks the submit button. Hence, I shall create a method called ‘submit’ that checks if the user has entered a value as input. If there isn’t an input value in a particular field, the designated error message is pushed into the errors array.</p><pre>methods: {</pre><pre>submit(){</pre><pre>// form validation</pre><pre>this.errors = [];</pre><pre>if (!this.firstName) {</pre><pre>this.errors.push(&#39;First name required&#39;);</pre><pre>}</pre><pre>if (!this.lastName) {</pre><pre>this.errors.push(&#39;Last name required&#39;);</pre><pre>}</pre><pre>if (!this.email) {</pre><pre>this.errors.push(&#39;Email required&#39;);</pre><pre>}}<br>}</pre><h4>2. Looping through the errors and displaying them</h4><p>I am using a ‘Bootstrap Alerts’ placeholder that will be displayed only when there is an error. Each error in the array is to be checked and displayed if available via looping.</p><pre>&lt;b-alert show dismissible variant=&quot;danger&quot; v-if=&quot;errors.length &gt; 0&quot;&gt;</pre><pre> &lt;span v-for=&quot;error in errors&quot; :key=&quot;error&quot;&gt;</pre><pre>  {{error}},</pre><pre> &lt;/span&gt;</pre><pre>&lt;/b-alert&gt;</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*x9PVGR6u4mEk60lH9YILsA.png" /></figure><h4>3. Success logic</h4><p>So finally, when all the conditions are met (all fields are filled), what do you want to happen? In most cases, this is where the database logic goes comes in.</p><p>But for the simplicity of this article, I shall only be displaying a success message to the user, using the V-if attribute, to change displays.</p><pre>if (this.firstName &amp;&amp; this.lastName &amp;&amp; this.email ) {</pre><pre>// hidding the form <br> this.showMain = false;</pre><pre>// showing the success message<br> this.showSuc = true</pre><pre>}</pre><p><em>Changing what is shown to the user:</em></p><pre>&lt;div class=&quot;main&quot; v-if=&quot;showMain&quot;&gt;</pre><pre>// the form</pre><pre>&lt;/div&gt;</pre><pre>&lt;div class=&quot;succcess&quot; v-if=&quot;showSuc&quot;&gt;</pre><pre>//success message</pre><pre>&lt;/div&gt;</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ypwUR8mi0SfkC3khJ0XUzg.png" /></figure><p>So there you have it! Performing form validations on the frontend was not much of an issue. Hope this article has helped, thank you.</p><p><em>More content at </em><a href="http://plainenglish.io/"><strong><em>plainenglish.io</em></strong></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ad9489628617" width="1" height="1" alt=""><hr><p><a href="https://javascript.plainenglish.io/basic-form-validation-using-vuejs-ad9489628617">Basic Form Validation using Vue.js: A Simple Guide</a> was originally published in <a href="https://javascript.plainenglish.io/">JavaScript in Plain English</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Simplifying Middleware in Express for Newbies and Pros Alike]]></title>
            <link>https://javascript.plainenglish.io/simplifying-middleware-in-expressjs-baad2d6595f3?source=rss-886f8b80b243------2</link>
            <guid isPermaLink="false">https://medium.com/p/baad2d6595f3</guid>
            <category><![CDATA[javascript]]></category>
            <category><![CDATA[programming]]></category>
            <category><![CDATA[nodejs]]></category>
            <category><![CDATA[web-development]]></category>
            <category><![CDATA[software-development]]></category>
            <dc:creator><![CDATA[Usiere Moses Akpabio]]></dc:creator>
            <pubDate>Sun, 29 Aug 2021 21:56:11 GMT</pubDate>
            <atom:updated>2021-09-10T05:05:31.789Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*0-Nqr2rIDCojCYio" /><figcaption>Photo by <a href="https://unsplash.com/@jstrippa?utm_source=medium&amp;utm_medium=referral">James Harrison</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>In writing Express applications, it is paramount to truly understand the concept of <strong>Middleware</strong> as they are the building blocks for building Express applications. Though implemented by a lot of developers in Express projects, many still do not fully understand the overall concept.</p><p>In this article, I will do my best to explain middleware using the most simplistic analogy for newbies and experienced developers to understand it better.</p><h3><strong>The Analogy of the Secretary</strong></h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*kEoeGsOsRdu21AeZK02rGQ.png" /></figure><p>As seen in most modern corporations that have a standard organisational culture, at the right hand of the top management is a reliable <strong>secretary</strong>.</p><p>Any request that is made to the top management of the company will most definitely have to pass through the <strong>secretary</strong>. The major daily duty of the secretary for the respective organisation is <strong>communication</strong> and <strong>correspondence</strong>. Requests that are made to the top management usually goes through the secretary, who analyses the information and seeks the most appropriate method to pass the information across.</p><p>Information given to the secretary is analysed and scrutinised for non-conformance to certain policies. In addition, the secretary performs other tasks such as documentation, duplication of information and many other processes that may be added with respect to the organisation. It is thus rare that any information/ request sent to the organization, will not pass through this process.</p><p>Lastly, when a response is made from the top management to staff or the public, it is done with the secretary functioning in between. He/ she (the secretary) usually delivers the information in the best format that the receiver will quickly understand.</p><p>In this scenario, we can see that the secretary basically acts as the <strong>middleman</strong> to which all requests and responses to and from top management are made.</p><p>In summary, <strong>Express</strong> M<strong>iddleware </strong>can be represented as <strong>secretaries that take the form of callback functions to which all HTTP client requests and server responses are made.</strong> In contrast to the analogy, more than one middleware can appear in a client request.</p><p><strong>Enough of the story.</strong></p><p>Basic example:</p><pre>const express = require (&#39;express&#39;);<br>const app = express();</pre><pre>middle1(req, res, next) =&gt; {<br>      console.log(&#39;You have made a request&#39;)<br>}</pre><pre>app.get(&#39;/&#39;, middle1)</pre><p>Illustrating the above code, the function middle1 will act as a middleware to which all GET requests made to the server will pass through. Therefore, for all GET requests,<em> ‘You made a request’</em> will be logged in the console.</p><p>Another useful comparison is that Express middleware acts as a layer of functions for every client request made to the server.</p><h3><strong>Definition.</strong></h3><p>The most reliable definition (from the express official documentation) is thus :</p><blockquote><strong><em>Middleware</em></strong> functions are functions that have access to the <a href="https://expressjs.com/en/4x/api.html#req">request object</a> (req), the <a href="https://expressjs.com/en/4x/api.html#res">response object</a> (res), and the next middleware function in the application’s request-response cycle. The next middleware function is commonly denoted by a variable named next.</blockquote><h3><strong>Categories of Middleware in Express</strong></h3><p>In dealing with express middleware, there are 3 main categories to understand. These categories are:</p><ol><li><strong>Application-level middleware:</strong></li></ol><p>They represent middleware that are bound to the app object by using app.use and app.method().</p><p>app.use will encompass all requests made, while app.method will be specific to the request type. (GET, POST, DELETE and PUT).</p><p><strong>Example:</strong></p><pre>const express = require(&#39;express&#39;)<br>const app = express()</pre><pre>//for all requests made<br>app.use(function (req, res, next) {<br>  console.log(&#39;I am an application level middleware&#39;)<br>  next()<br>})</pre><p><strong>2. Router-level middleware:</strong></p><p>Same as Application-level middleware, but the main difference is that it is bound to an instance of the Express router.</p><p><strong>Example:</strong></p><pre>const express = require(&#39;express&#39;)<br>const app = express()<br>const router = express.Router()</pre><pre>// this example is respective only to get requests<br>router.get(&#39;/user/:id&#39;,(function (req, res, next) {<br>  console.log(&#39;Time:&#39;, Date.now())<br>  next()<br>})</pre><p>3<strong>. Error handling middleware</strong></p><p>Works in the same way as other middleware functions, except that it has 4 arguments.</p><p><strong>Example:</strong></p><pre>app.use(function (err, req, res, next) {<br>  console.error(err.stack)<br>  res.status(500).send(&#39;Something broke!&#39;)<br>})</pre><p>So there you have it, hope this has helped in briefly explaining the concept of <strong>Express</strong> <strong>Middleware</strong> better. For a more in-depth explanation of middleware in Express, I recommend you read the official documentation:</p><p><a href="https://expressjs.com/en/guide/using-middleware.html">Using Express middleware</a></p><p>Thanks for reading.</p><p><em>More content at </em><a href="http://plainenglish.io/"><strong><em>plainenglish.io</em></strong></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=baad2d6595f3" width="1" height="1" alt=""><hr><p><a href="https://javascript.plainenglish.io/simplifying-middleware-in-expressjs-baad2d6595f3">Simplifying Middleware in Express for Newbies and Pros Alike</a> was originally published in <a href="https://javascript.plainenglish.io/">JavaScript in Plain English</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>