<?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 Natesh Bhat on Medium]]></title>
        <description><![CDATA[Stories by Natesh Bhat on Medium]]></description>
        <link>https://medium.com/@nateshmbhat?source=rss-c2c2a0e489e7------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*fZVmBOlfvmy-3FKpqPTIgQ.jpeg</url>
            <title>Stories by Natesh Bhat on Medium</title>
            <link>https://medium.com/@nateshmbhat?source=rss-c2c2a0e489e7------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 17:34:35 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@nateshmbhat/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[1,500 Green Tests Didn’t Catch a Single One of These Bugs]]></title>
            <link>https://nateshmbhat.medium.com/1-500-green-tests-didnt-catch-a-single-one-of-these-bugs-76d7cb05292d?source=rss-c2c2a0e489e7------2</link>
            <guid isPermaLink="false">https://medium.com/p/76d7cb05292d</guid>
            <category><![CDATA[flutter]]></category>
            <category><![CDATA[flutter-app-development]]></category>
            <category><![CDATA[software-testing]]></category>
            <category><![CDATA[testing]]></category>
            <dc:creator><![CDATA[Natesh Bhat]]></dc:creator>
            <pubDate>Fri, 14 Aug 2026 06:28:21 GMT</pubDate>
            <atom:updated>2026-08-14T06:28:21.260Z</atom:updated>
            <content:encoded><![CDATA[<p>What building an in-app QA HUD for Flutter taught me about the limits of a test suite</p><h4><em>This is the tool at the end of the post. First, the three bugs it found in </em><strong><em>itself</em></strong><em> — past 1,500 green tests.</em></h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*TnoOKrPo_LhIOXhYQfwoSQ.png" /></figure><p>I shipped a Flutter package with about 1,500 passing tests.</p><p>Then I found a bug where <strong>every single tap in the app did nothing.</strong> Not a subtle bug. Not an edge case. The app was a photograph. And ~500 of those tests had been green the whole time.</p><p>It was the first of three, and they all had the same shape: three reproducible ways a test suite can be <strong>structurally incapable</strong> of seeing a bug. Each one below has the code, and the four-line test that would have caught it.</p><h3>1. The invisible sheet of glass</h3><p>I was building an overlay that renders above the host app. It needs its own Navigator — not for navigation, but because panels inside it use showDialog, DropdownButton, tooltips and text-selection toolbars, and <strong>all of those push routes and throw without one.</strong></p><p>So:</p><pre>Navigator(<br>  onGenerateRoute: (_) =&gt; PageRouteBuilder(<br>    opaque: false,          // app stays visible ✓<br>    barrierColor: null,     // no dimming ✓<br>    transitionDuration: Duration.zero,<br>    pageBuilder: (_, __, ___) =&gt; const MyOverlay(),<br>  ),<br>);</pre><p>Every knob set to <em>don’t get in the way</em>. The app underneath became completely uninteractable.</p><blockquote><strong><em>Every </em></strong><strong><em>ModalRoute — and </em></strong><strong><em>PageRouteBuilder is one — always contributes a full-screen </em></strong><strong><em>ModalBarrier.</em></strong></blockquote><blockquote><em>opaque: false controls whether routes below are </em><strong><em>painted</em></strong><em>. </em><em>barrierColor: null controls whether the barrier </em><strong><em>paints anything</em></strong><em>. </em><strong><em>Neither controls whether it absorbs pointers.</em></strong><em> It always does.</em></blockquote><p>An invisible, transparent, undismissible sheet of glass over the entire app.</p><p>The fix is OverlayRoute — the primitive underneath, which manages overlay entries and nothing else:</p><pre>class _ShellRoute extends OverlayRoute&lt;void&gt; {<br>  @override<br>  Iterable&lt;OverlayEntry&gt; createOverlayEntries() =&gt;<br>      [OverlayEntry(builder: builder, maintainState: true)];<br>}</pre><p>Routes pushed <em>on top</em> of this are still modal — correct, since a dropdown <strong>should</strong> capture input while open.</p><p><strong>Why ~500 tests missed it.</strong> They asserted things were <em>present</em>. find.byType(MyButton) succeeds perfectly well when an invisible barrier sits on top of that button. The widget is there. It is simply unreachable.</p><p>Presence and reachability are different claims, and only one of them is what you promised.</p><pre>// The highest-value test in the entire suite. Four lines.<br>await tester.tap(find.text(&#39;app button&#39;));<br>await tester.pump();<br>expect(taps, 1, reason: &#39;if this fails, the whole app is uninteractable&#39;);</pre><h3>2. The test that ran on a phone that doesn’t exist</h3><p>My panel opened at the top of the screen. Perfect on the simulator. On a real iPhone the tab labels looked dead — tap, nothing.</p><p>They were underneath the <strong>Dynamic Island</strong>, in the band where iOS consumes touches before your app ever sees them.</p><p>The reason no test caught it is almost funny:</p><blockquote><strong><em>Widget tests default to zero safe-area insets.</em></strong></blockquote><p>Every test was running on a phantom device with no notch. The assertion “the tab bar is visible” was <em>true</em>, and completely meaningless — it could never have been false.</p><pre>// Assert the inset EXISTS first, or the test can only ever pass.<br>final topInset = tester.view.viewPadding.top / tester.view.devicePixelRatio;<br>expect(topInset, greaterThan(0),<br>    reason: &#39;vacuous on a device with no notch&#39;);</pre><pre>expect(tester.getRect(find.byType(TabBar)).top,<br>    greaterThanOrEqualTo(topInset));</pre><p><strong>The generalisable version:</strong> any test that would pass on a device with no notch, a database with no rows, or a flag nobody flipped is not testing what you think. Assert your preconditions are real.</p><h3>3. A plugin with no host was recording user data — in production</h3><p>This is the one that still bothers me.</p><p>A plugin gated data capture on a _detached flag. Detach → capture stops. Sound design.</p><p>Except <strong>the flag started </strong><strong>false</strong>. So <em>&quot;drops changes while detached&quot;</em> was only true after an <strong>explicit</strong> detach. A plugin that had <strong>never attached</strong> captured on the very first change.</p><p>Now line that up with a release build:</p><pre>enabled defaults to !kReleaseMode  →  in release, the host never activates<br>                                   →  it never registers the plugin<br>                                   →  the plugin never attaches<br>                                   →  the flag stays false<br>                                   →  capture is ON</pre><p>Meanwhile the integrator has already wired the observer into ProviderScope(observers: [...]), <strong>exactly as the docs instruct.</strong> That observer is not the host&#39;s to stop.</p><p>Real user state, recorded into an in-memory log, in production.</p><p><strong>And here’s the part I want you to take away.</strong> I wrote the obvious test — <em>assert the log is empty</em> — and it passed. It would have passed even while the bug was live in a different form, because an empty log doesn’t prove nothing was <strong>read</strong>.</p><p>So I gave it an object that counts its own inspection:</p><pre>class _CountsItsOwnToString {<br>  int reads = 0;<br>  @override<br>  String toString() { reads++; return &#39;sensitive&#39;; }<br>}</pre><pre>final probe = _CountsItsOwnToString();<br>plugin.record(name: &#39;session&#39;, next: probe);</pre><pre>expect(plugin.log.changes, isEmpty);   // necessary<br>expect(probe.reads, 0,                 // ...but this is the real assertion<br>    reason: &#39;serialising real user state IS the leak&#39;);</pre><p>The value was being <strong>serialised on the way in.</strong> Reading the user’s data is itself the leak — retention is just the part you can see afterwards.</p><p><strong>Two lessons, and they generalise past Flutter.</strong> Fail-safe means <em>unarmed until something arms you</em>, never <em>armed until something stops you</em>. And when you test for a leak, assert on the <strong>work</strong>, not the <strong>residue</strong>.</p><h3>The pattern</h3><p>Three bugs, one shape:</p><p>Where it livedWhy the suite couldn’t see itUntappable app<strong>Above</strong> the widget treeTests asserted presence, not reachabilityDynamic Island<strong>Outside</strong> the test’s fake deviceTests ran with no notch, so the check couldn’t failData leak<strong>Outside</strong> the host’s ownershipThe observer was wired by the integrator, not the host</p><p>Every one lived in a <strong>seam</strong> — the gap between two things that each worked fine alone.</p><p>Your test suite is a hypothesis about where bugs live. Mine said <em>inside units</em>. It was wrong, and it had 1,500 pieces of evidence for itself.</p><p>Three habits, near-zero cost:</p><ol><li><strong>Test the composition.</strong> Mount everything, then try to actually <em>use</em> the app through it. One test.</li><li><strong>Make assertions falsifiable.</strong> If it can’t fail, it isn’t a test.</li><li><strong>Assert on the work, not the result.</strong> An empty log doesn’t prove nothing was read.</li></ol><p><em>(Two more didn’t fit: a release-gate flag that threw </em><em>LateInitializationError on every transition because nobody had ever flipped it, and </em><em>dart pub publish --dry-run reporting 15/15 clean while three packages were unusable — it validates each package in isolation and never asks the server anything. Both are in the repo&#39;s changelog.)</em></p><h3>The tool these came from</h3><p><strong>Vantage</strong> — one floating bubble that turns <em>“it broke on my phone”</em> into a zip a developer can act on.</p><pre>void main() =&gt; runApp(<br>  Vantage(<br>    // Zero cost in release: `enabled` defaults to `!kReleaseMode`.<br>    plugins: [<br>      VantageDevicePreviewPlugin.withFrames(),<br>      VantageInspectorPlugin(),<br>      VantageFlightRecorderPlugin(),<br>    ],<br>    builder: (context) =&gt; const MyApp(),<br>  ),<br>);</pre><pre>class MyApp extends StatelessWidget {<br>  @override<br>  Widget build(BuildContext context) =&gt; MaterialApp(<br>    builder: Vantage.appBuilder,   // ← the only change inside your app<br>    home: const HomeScreen(),<br>  );<br>}</pre><p>No native setup. No build flavors. No wrapper around every screen.</p><p><strong>~40 devices</strong> with <em>real</em> safe-area insets and pixel ratios — a layout that collides with the home indicator fails in preview the way it fails on the phone. Plus a11y auditing, pseudo-localization, a deep-link launcher, a frame-time HUD, and live feature-flag overrides.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*iUm5ZQYFrPAwy-fpDz09ig.png" /><figcaption><em>Left to right — </em><strong><em>Inspector:</em></strong><em> tap any widget for size, constraints, padding and its ancestor chain. </em><strong><em>Network:</em></strong><em> every request, copy-as-cURL, auth headers redacted before they can reach a bug report. </em><strong><em>Flight Recorder:</em></strong><em> one tap zips breadcrumbs, a screenshot, the network log, the route stack and a device block to the share sheet.</em></figcaption></figure><p>Core depends on the <strong>Flutter SDK only</strong> — every integration ships as a separate federated package, and CI fails the build if that’s ever violated.</p><p><strong>Stated plainly, because you’d find out anyway:</strong> it’s 0.2.0, APIs will shift before 1.0, and screenshots in a bug bundle <strong>cannot</strong> be redacted — the export pipeline skips binary artifacts by design. The masking tool uses an <strong>opaque</strong> mask rather than a blur, because blurred PINs and card digits are recoverable.</p><p>📦 <a href="https://pub.dev/packages/vantage">pub.dev/packages/vantage</a> · 🧑‍💻 <a href="https://github.com/nateshmbhat/vantage">github.com/nateshmbhat/vantage</a> · MIT</p><p><strong>What’s the worst bug your green test suite ever hid from you?</strong> I collect these now.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=76d7cb05292d" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Building an AI Support Workflow for Zoho Desk — Key Learnings, Honest Mistakes, and What’s Next]]></title>
            <link>https://medium.com/swlh/building-an-ai-support-workflow-for-zoho-desk-key-learnings-honest-mistakes-and-whats-next-f11926c176f3?source=rss-c2c2a0e489e7------2</link>
            <guid isPermaLink="false">https://medium.com/p/f11926c176f3</guid>
            <category><![CDATA[automation]]></category>
            <category><![CDATA[zoho]]></category>
            <category><![CDATA[support]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Natesh Bhat]]></dc:creator>
            <pubDate>Fri, 20 Mar 2026 08:32:11 GMT</pubDate>
            <atom:updated>2026-03-20T08:32:11.792Z</atom:updated>
            <content:encoded><![CDATA[<h3>Building an AI Support Workflow for Zoho Desk — Key Learnings, Honest Mistakes, and What’s Next</h3><p><em>A building-in-public breakdown of how we approached ticket automation, what surprised us, and where we’re taking it.</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*1UfcLQb3JVESg4WWUm_Mrw.jpeg" /></figure><p>We’ve been quietly building an AI-assisted support automation system for a guided meditation app. The support queue gets a steady volume of tickets every day: login issues, meditation playback problems, streak resets, OTP failures, feature requests, and everything in between.</p><p>For a while, the team was handling this entirely manually. That’s not a complaint — it’s just the reality of a growing product with a small support team. But as volume grew, so did the patterns. The same 10 types of questions accounted for most of the queue. And every repeat question answered manually is time taken away from the ones that actually need attention.</p><p>So we built something. This post is an honest account of the decisions we made, what we got wrong first, and what the system actually looks like now — including a memory layer built on top of <strong>Memvid</strong> that we think is the most interesting part of the whole thing.</p><h3>The Problem with Starting Simple</h3><p>Our first attempt was exactly what you’d expect: feed the ticket to an LLM, ask it to categorize and draft a reply, post the reply. Done.</p><p>The problem showed up almost immediately. The LLM was confident about everything. A user writes <em>“my progress has disappeared”</em> — the system sees “meditation progress” and cheerfully drafts a troubleshooting guide. But that ticket is describing data loss. It needs a human, an investigation, and probably a personal response. Not a canned reply.</p><p>The issue isn’t that LLMs are bad at support. It’s that <strong>confidence without calibration is dangerous in support contexts</strong>. The cost of a wrong auto-reply isn’t just a missed resolution — it’s an upset user who now feels like they were fobbed off by a bot.</p><p>So we scrapped the naive approach and built around a different principle: the system’s first job is to know when NOT to respond.</p><h3>What We Actually Built: Confidence-Scored Routing</h3><p>The core of the system is a categorization engine that scores every ticket before deciding what to do with it. Every ticket gets a confidence score, and that score determines the route:</p><pre><strong>≥ 95% + approved category  →  AUTO_RESPOND<br>90–94%                     →  AI_REVIEW_NEEDED (draft only, human approves)<br>&lt; 90%                      →  MANUAL_REQUIRED (flagged, Slack alert sent)</strong></pre><p>Scoring starts at a base of 70% and adjusts based on signals:</p><pre><strong>+15%  3+ strong keyword matches<br>+10%  2 keyword matches<br>+10%  Category AND subcategory fields populated in Zoho Desk<br>+5%   Similar historical tickets resolved the same way (memory boost)<br>+5%   Repeat user with prior tickets in same category<br><br>-10%  Vague subject (&quot;Help&quot;, &quot;Problem&quot;, &quot;Issue&quot;)<br>-5%   No category/subcategory from Zoho Desk</strong></pre><p>But here’s the key — even at 100% confidence, a ticket touching data loss is <strong>never</strong> auto-responded to. The check isn’t just the score. It’s a hard flag on the ticket type. Streak resets, shield misuse, disappearing progress, multiple account anomalies — these always go to a human, full stop. The words “lost”, “disappeared”, “reset”, “gone” in any ticket trigger this block regardless of what the score says.</p><p><strong>What we learned:</strong> Confidence scoring is only useful if you also maintain a hard “never auto-respond” list that sits outside the scoring logic entirely. The two need to coexist.</p><h3>The Taxonomy Problem</h3><p>We initially thought categorization was simple — just a list of categories. It turned out to be one of the most important and most iterative pieces of the whole system.</p><p>We ended up with <strong>22 tags across 6 domains</strong>:</p><ol><li>Technical Issues — crashes, bugs, performance</li><li>Authentication &amp; Login — OTP, country blocks, sign-in failures</li><li>Stats &amp; Progress — streaks, coins, badges</li><li>Meditation Experience — playback, instructions, audio</li><li>Profile &amp; Account — edits, multiple accounts, data issues</li><li>General — feedback, unclear queries, language, wisdom content</li></ol><p>The interesting discovery was that the real ticket language often doesn’t match what you’d expect. Tickets about streak issues almost never say “streak” explicitly. Users write things like <em>“my number went to zero”</em>, <em>“something happened to my days”</em>, <em>“glitch in my progress”</em>. The taxonomy only works if the keyword lists are built from actual ticket data — not from what you imagine users will write.</p><p>We pulled a dump of ~9,800 historical tickets from December 2025 and did a full pattern analysis before finalising the keyword lists. This changed about a third of the keywords we had assumed. That historical data turned out to be more valuable than any amount of upfront design.</p><p><strong>What we learned:</strong> Build your taxonomy from real ticket language, not assumed language. The two are surprisingly different.</p><h3>The Multi-Tenant Architecture</h3><p>Fairly early we realised: if this is worth building, it should be reusable. Not “copy the code and edit it” reusable — actually designed for multiple teams to run off the same core.</p><p>The solution was a <strong>tenant pack</strong> system. Each team gets a small configuration folder:</p><pre>tenants/<br>  your-app/<br>    manifest.json        ←  Zoho scope (department, team, queue)<br>    taxonomy.json        ←  tag definitions, keywords, routing rules<br>    canned_responses.md  ←  approved reply templates</pre><p>The manifest.json enforces scope at the code level — the system will only read and write tickets within the configured department and team boundary. This is non-negotiable. A misconfigured run can&#39;t accidentally touch another team&#39;s queue.</p><p>Scaffolding a new tenant takes one command:</p><pre>python3 scripts/zoho_desk_cli.py tenant scaffold new-team --display-name &quot;New Team&quot;</pre><p>The CLI generates the folder with sample files pre-populated. The new team edits what’s specific to them — categories, keywords, responses, field mappings — and the shared engine handles the rest.</p><p><strong>What we learned:</strong> The “just edit the code” approach doesn’t scale past one team. The configuration/runtime separation is worth the upfront effort.</p><h3>The Memory System — and Why We Chose Memvid</h3><p>This is the part we’re most excited about, and also the most honest about still being in progress.</p><p>The baseline system — keyword matching + confidence scoring — got us functional. But it had a fundamental limitation: <strong>every ticket was processed in isolation</strong>. No awareness of what happened yesterday. No sense of whether this is a repeat issue from the same user. No ability to say “the last 8 times we saw this phrasing, it was a streak data loss issue.”</p><p>We wanted to add memory. The question was how.</p><h3>Why Memvid</h3><p>We chose <strong>Memvid</strong> (memvid-sdk) as the storage layer for one specific reason: it gives us <strong>hybrid search out of the box in a single local file</strong>.</p><p>A .mv2 file is a self-contained memory store that supports both:</p><ul><li><strong>BM25 lexical search</strong> — keyword overlap, good for exact phrase matches</li><li><strong>Vector/semantic search</strong> — cosine similarity on embeddings, good for meaning-level matches</li></ul><p>For support tickets, you need both. Lexical search catches tickets with the same exact keywords. Semantic search catches tickets that describe the same problem in different words — and users describe problems in wildly different ways.</p><p>The alternative would have been spinning up a proper vector database (Pinecone, Weaviate, etc.) — more infrastructure, more moving parts, more things to manage. Memvid gives us a single file that lives inside the repo, requires no external service, and supports the same hybrid search pattern.</p><p>We pair it with <strong>Google Gemini embeddings</strong> (text-embedding-004) for the semantic side. Gemini&#39;s embedding quality is strong, the API is reliable, and we were already using Gemini elsewhere in the stack.</p><h3>What the Memory Does</h3><p>When a new ticket comes in, the system:</p><ol><li>Queries the memory store for the 5 most similar historical tickets (hybrid search, 50/50 lexical + semantic weight)</li><li>Looks up the requester’s email for prior ticket history</li><li>Checks for recent pattern signals (same issue appearing 3+ times in 24 hours)</li></ol><p>The results feed into the confidence calculation:</p><p>python</p><pre># If 3+ similar tickets agree on the same category<br>if matching_count &gt;= 3:<br>    boost = matching_count * 3  # +3% per agreeing ticket<br><br># If similar tickets were frequently agent-corrected<br>if correction_rate &gt; 0.3:<br>    penalty = int(correction_rate * 10)<br>    memory_boost -= penalty  # System gets humble</pre><p>That last part — the penalty for correction history — is the one we’re most proud of. If agents keep overriding the AI’s categorization on a certain ticket type, the system automatically starts sending those for review instead of auto-responding. The correction signal travels backwards into future processing.</p><h3>Seeding the Memory</h3><p>The memory was seeded with the full historical ticket dataset — ~9,800 tickets from December 2025. Each ticket gets embedded and stored with its category, resolution, response sent, and a was_corrected flag for future corrections.</p><pre>python -m memory.seed_memory \<br>  --csv data-dumps/historical_tickets_dec2025.csv \<br>  --api-key $GEMINI_API_KEY</pre><p>Every ticket processed going forward gets added to the store, so the memory grows continuously.</p><h3>Where We Are Now</h3><p>Phase 1 — the basic memory retrieval and confidence boosting — is live and running. It’s already making a visible difference on borderline tickets where keywords alone weren’t enough.</p><p><strong>Baseline before memory:</strong> 73% of tickets going to manual review, 0% auto-responded (we hadn’t opened auto-respond yet), avg confidence 75–85%.</p><p><strong>Target:</strong> Manual review rate below 40%, auto-response at 10–15%, avg confidence above 90%.</p><p>We’re not there yet. But the trajectory is good.</p><h3>What’s Coming: Phase 2 and 3</h3><h3>Phase 2 — Learning Loop</h3><p>The next phase is closing the feedback loop more formally.</p><p>Right now, when an agent overrides the AI’s categorization, that correction gets stored. But the system doesn’t yet actively mine those corrections to improve itself. Phase 2 adds a <strong>Correction Tracker</strong> that periodically analyses which categories have high override rates, which keywords are showing up in corrected tickets, and whether confidence thresholds need adjusting.</p><p>The goal is <strong>confidence calibration</strong> — if the 90–94% bucket is actually correct 95%+ of the time, we should lower the auto-respond threshold. If the 95–100% bucket is being corrected more than expected, we should raise it. The thresholds should be data-driven, not guesses.</p><p>Phase 2 also adds proper <strong>pattern detection</strong> — if the same issue appears across 5+ tickets in a 7-day window, the system surfaces it as an alert with affected ticket IDs and a suggested action. This is particularly useful for catching app version regressions before they turn into a flood of tickets.</p><h3>Phase 3 — Self-Evolution (The Interesting Part)</h3><p>This is the long-term vision, and the most experimental.</p><p>The idea: the system should be able to propose improvements to itself — new keywords to add to the taxonomy, new canned responses for recurring patterns, threshold adjustments based on accuracy data — and present them for human review before applying them.</p><p>bash</p><pre># List pending proposals<br>python -m memory.proposals list --status pending<br><br># Approve a proposal<br>python -m memory.proposals approve --id prop_20260108_001 --reviewer &quot;your-name&quot;<br><br># Apply it (with automatic backup)<br>python -m memory.proposals apply --id prop_20260108_001</pre><p>Every applied change gets backed up first, and rollback is a single command. The human stays in control — the system proposes, the human approves.</p><p>The vision is that the gap between “what the taxonomy covers” and “what users actually ask” closes automatically over time, rather than requiring someone to manually audit the keyword lists every few months.</p><h3>The Honest Scorecard</h3><p>A few things we got right, a few things we’d do differently:</p><p><strong>Got right:</strong></p><ul><li>Designing around safety-first from day one — it forced good habits everywhere else</li><li>Building the CLI as the primary interface before anything else — made testing and debugging 10x easier</li><li>Using real historical ticket data to build the taxonomy rather than guessing</li></ul><p><strong>Would do differently:</strong></p><ul><li>We started without a proper .gitignore and had to clean up early; set that up on day one next time</li><li>The tenant pack structure took longer to design than it should have — we should have done it earlier instead of retrofitting it</li><li>The Memvid integration plan was detailed but Phase 1 took longer to stabilise than expected; the real-ticket edge cases (encoding issues, empty descriptions, malformed subjects) weren’t fully accounted for in the original design</li></ul><h3>What We’re Watching</h3><p>The broader question this project keeps raising is: <strong>what is the right role for AI in a support queue?</strong></p><p>The naive answer is “replace as much as possible.” The answer we’ve landed on is more nuanced: AI is good at handling volume at the edges — the clearly simple tickets and the clearly critical ones — so that humans can spend their time on the middle. The ambiguous tickets, the upset users, the edge cases. That’s where the real support work happens, and that’s where the judgment call matters.</p><p>Our 10–15% auto-response target isn’t a ceiling we’re trying to push through. It’s a floor we’re trying to maintain confidently. Getting the auto-respond accuracy to near-perfect on that 10–15% frees up more time than trying to push auto-respond to 50% at lower confidence.</p><p>We’ll keep sharing what we learn as the memory system matures and as Phase 2 comes together. If you’re building something similar or have thoughts on the approach — would love to hear it.</p><p><em>Built with Python, Claude, Memvid, and a lot of real support tickets from real people.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f11926c176f3" width="1" height="1" alt=""><hr><p><a href="https://medium.com/swlh/building-an-ai-support-workflow-for-zoho-desk-key-learnings-honest-mistakes-and-whats-next-f11926c176f3">Building an AI Support Workflow for Zoho Desk — Key Learnings, Honest Mistakes, and What’s Next</a> was originally published in <a href="https://medium.com/swlh">The Startup</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[ Beating ChatGPT & TikTok : How our App Broke Records Overnight, almost crashing our servers]]></title>
            <link>https://nateshmbhat.medium.com/beating-chatgpt-tiktok-how-our-app-broke-records-overnight-almost-crashing-our-servers-4b21c39c68c3?source=rss-c2c2a0e489e7------2</link>
            <guid isPermaLink="false">https://medium.com/p/4b21c39c68c3</guid>
            <category><![CDATA[chatgpt]]></category>
            <category><![CDATA[application-development]]></category>
            <category><![CDATA[meditation]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[apps]]></category>
            <dc:creator><![CDATA[Natesh Bhat]]></dc:creator>
            <pubDate>Sat, 01 Mar 2025 11:00:49 GMT</pubDate>
            <atom:updated>2025-03-01T11:00:49.679Z</atom:updated>
            <content:encoded><![CDATA[<h3>🚀 Beating ChatGPT &amp; TikTok? How our Meditation App Broke Records Overnight, almost crashing our servers</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/948/0*iDO4JkcZKousUkMi" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/590/1*tg80DKO50q04dgEFN9FdEA.jpeg" /><figcaption>#1 Spot in App Store</figcaption></figure><blockquote><strong>When Sadhguru’s </strong><a href="https://miracleofmind.sng.link/Aoy32/bsig6?_dl=sgmiracle%3A%2F%2F&amp;_smtype=3"><strong>Miracle of Mind</strong></a><strong> meditation app launched during Mahashivaratri, none of us expected to break ChatGPT’s initial download record. But that’s exactly what happened — 1 million downloads in just 15 hours! 🤯</strong></blockquote><p>As part of the development team, I’m still processing the wild ride we experienced. Let me take you behind the scenes of what became one of the most successful wellness app launches in recent history.</p><h3>📈 A Milestone That Defies Silicon Valley Logic</h3><p>Let’s be real about what just happened: While ChatGPT needed 5 full days to hit the 1M user mark and TikTok took a whopping 8 months, our meditation app crossed this threshold in just 15 hours! 🔥 We’ve claimed the #1 Free App spot in India’s App Store, reached #3 on Android, and broken into Healthcare Top 10 charts across the US, UK, Canada, Australia and numerous other countries worldwide. 🌏</p><p>The most beautiful part? This wasn’t accomplished by a tech giant with unlimited venture capital. We’re a nonprofit team based in Coimbatore, working with modest resources but extraordinary commitment. ❤️ No billion-dollar budgets or Silicon Valley headquarters — just passionate volunteers united by one powerful mission: bringing daily meditation to everyone on the planet. 🧘‍♀️✨</p><h4>What made it resonate so powerfully? The app offers:</h4><ul><li>A 7-minute daily meditation for stress reduction ⏱️ with streaks and rewards for consistency.</li><li>Sadhguru’s wisdom and insights. 🧘‍♂️</li><li>“Ask Sadhguru” — an AI-powered guide for life’s questions. 🤖</li></ul><p>But the journey wasn’t without its dramatic moments, particularly from an engineering perspective. 🧑🏼‍💻</p><h3>💣 The Midnight Time Bomb</h3><p>Picture this: It’s the stroke of midnight during Mahashivaratri, and Sadhguru is initiating thousands into the mahamantra. Simultaneously, our monitoring systems light up like a Christmas tree. 🎄</p><p>Our server was suddenly hammered with <strong>24,000 requests per second</strong>. For context, our normal load during the announcement was around 160,000 requests per <em>minute</em>. This midnight spike was pushing our infrastructure dangerously close to its breaking point. ⚠️</p><p>Our DevOps engineer jokingly accused me of planting a “midnight time bomb” in the app. As it turns out, he wasn’t entirely wrong. 😅</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*XLiu3zySweHRxIBIk5kbjg.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*2ZZ2A-mLqF2teyUNWeNdUA.png" /><figcaption>Server Spikes</figcaption></figure><h3>🕵️‍♀️ The Culprit: A Well-Intentioned Design Decision</h3><p>After diving into the logs, we identified the cause: a seemingly innocent design decision in our date-handling logic.</p><p>We built the app using Riverpod for state management on the client side, with all meditation stats (streaks, minutes meditated, etc.) dependent on the current date. To ensure users would always see fresh stats without loading delays, we programmed the app to invalidate cached data and fetch new information precisely at midnight when the date changes.</p><p>The intention was good — users would have updated UI states ready the moment they opened the app the next day. But we didn’t consider the consequences of hundreds of thousands of users’ apps all making simultaneous API calls at exactly 12:00 AM. 🕛</p><h3>🚑 The Emergency Fix</h3><p>Response times jumped from milliseconds to 1.5 seconds during the spike. While that might not sound catastrophic, it represented a serious problem at our scale. We quickly implemented a three-part solution:</p><ol><li>Invalidate date based cache when the app comes from background to foreground 📱 and when the date has changed.</li><li>Keep the midnight invalidation only when app is active in the foreground 🕒</li><li>When app is in background, implement a randomized delay (few seconds to minutes) before fetching new data ⏰</li></ol><p>This simple change distributed the load across a wider timeframe, dramatically reducing server pressure. 📉</p><h3>💡 Lessons for Scaling Meditation (and Other Apps)</h3><p>What began as a stressful engineering crisis evolved into a valuable lesson about building for scale. Even the most wellness-focused apps need to consider the technical implications of success.</p><p>Our experience highlights the importance of:</p><ul><li>Anticipating concurrent user actions, especially time-based triggers 📊</li><li>Implementing staggered or randomized scheduling for background processes 🔄</li><li>Thoroughly stress-testing time-sensitive features 🧪</li><li>Never underestimating how a spiritual event can drive digital engagement 🙏</li></ul><blockquote>The irony isn’t lost — our app was designed to reduce stress, yet the servers had a massive stress spike on launch night :D</blockquote><h3>🌍 The Bigger Picture</h3><p>Beyond our technical challenges, the app’s success reveals something profound about our current moment: people worldwide are hungry for tools that support mental well-being and inner growth. The fact that a meditation app outperformed entertainment and commerce platforms speaks volumes about shifting priorities.</p><p>As we continue refining the app, this launch day learning will inform how we build features going forward. Sometimes the most valuable engineering insights come when systems are pushed to their limits. 💪</p><p>For now, we’re celebrating this milestone while keeping our servers (and ourselves) in a more meditative state. 🧘‍♀️✨</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=4b21c39c68c3" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Shh! Mastering Do Not Disturb in Flutter: A Quirky Guide to Digital Zen]]></title>
            <link>https://nateshmbhat.medium.com/shh-mastering-do-not-disturb-in-flutter-a-quirky-guide-to-digital-zen-704359b67e6c?source=rss-c2c2a0e489e7------2</link>
            <guid isPermaLink="false">https://medium.com/p/704359b67e6c</guid>
            <category><![CDATA[flutter-app-development]]></category>
            <category><![CDATA[android]]></category>
            <category><![CDATA[android-notification]]></category>
            <category><![CDATA[notifications]]></category>
            <category><![CDATA[flutter]]></category>
            <dc:creator><![CDATA[Natesh Bhat]]></dc:creator>
            <pubDate>Thu, 03 Oct 2024 16:21:27 GMT</pubDate>
            <atom:updated>2024-10-03T16:21:27.333Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*v-zmyl--H-TAZcxIyEtW6A.png" /></figure><p>Picture this: You’re deep in the zone, coding the next big app that’s going to revolutionize… I don’t know, maybe cat videos? Suddenly, your phone erupts with notifications about your aunt’s latest farmville achievements. Poof! There goes your concentration, and your cat video empire crumbles before it even begins.</p><p>Fear not, fellow Flutter enthusiasts! Today, we’re diving into the world of Do Not Disturb (DND) mode, or as Android likes to call it, “Zen mode” (because nothing says inner peace like silencing your mother-in-law’s calls, right?).</p><h3>Enter the do_not_disturb Plugin :</h3><p>Allow me to introduce you to this new package authored by yours truly (that would be me 🥳) : The <a href="https://pub.dev/packages/do_not_disturb">do_not_disturb</a> Flutter plugin. This nifty little package allows you to check, and even control, the DND status on Android devices. iOS users, I’m afraid you’ll have to find inner peace the old-fashioned way — maybe try some actual yoga?</p><h4>Getting Started: Installation and Setup</h4><p>First things first, let’s add this bad boy to your `pubspec.yaml`:</p><pre>dependencies:<br> do_not_disturb: &lt;latest_version&gt;</pre><h4>Reading the DND Status: Are We Zen Yet?</h4><p>Let’s start with the basics — checking if DND is enabled:</p><pre>import &#39;package:do_not_disturb/do_not_disturb.dart&#39;;<br>final dndPlugin = DoNotDisturbPlugin();<br>Future&lt;void&gt; checkDndStatus() async {<br> bool isDndEnabled = await dndPlugin.isDndEnabled();<br> print(isDndEnabled ? &quot;Shh, we&#39;re in zen mode!&quot; : &quot;LOUD NOISES!&quot;);<br>}</pre><p>But wait, there’s more! What if you want to know exactly how zen we are? The plugin’s got you covered with the `getDNDStatus` method:</p><pre>Future&lt;void&gt; getDetailedDndStatus() async {<br> InterruptionFilter status = await dndPlugin.getDNDStatus();<br> switch (status) {<br> case InterruptionFilter.all:<br> print(&quot;All interruptions welcome. Chaos reigns!&quot;);<br> break;<br> case InterruptionFilter.priority:<br> print(&quot;VIPs only. Sorry, telemarketers!&quot;);<br> break;<br> case InterruptionFilter.none:<br> print(&quot;Maximum zen achieved. Not even your thoughts are allowed.&quot;);<br> break;<br> case InterruptionFilter.alarms:<br> print(&quot;Alarms only. Because oversleeping is the real disturbance.&quot;);<br> break;<br> default:<br> print(&quot;Unknown state. Did we enter the quantum realm?&quot;);<br> }<br>}</pre><h3>Taking Control: Becoming the DND Master</h3><p>Now, for the power move — actually setting the DND mode. But remember, with great power comes great responsibility (and user permissions):</p><pre>Future&lt;void&gt; embraceTheZen() async {<br> if (await dndPlugin.hasNotificationPolicyAccess()) {<br> await dndPlugin.setInterruptionFilter(InterruptionFilter.none);<br> print(&quot;Zen mode activated. You are now one with the void.&quot;);<br> } else {<br> print(&quot;We need the sacred permission to control the zen!&quot;);<br> await dndPlugin.openNotificationPolicyAccessSettings();<br> // Politely ask the user to grant permission and return to the app<br> }<br>}</pre><p>But before you go off silencing the world, make sure your `AndroidManifest.xml` is properly dressed for the occasion:</p><pre>&lt;uses-permission android:name=&quot;android.permission.ACCESS_NOTIFICATION_POLICY&quot;/&gt;</pre><h3>The User Experience: A Journey to Digital Nirvana</h3><p>When building your app, remember that controlling DND is a sacred power. Guide your users gently:</p><ol><li>Check if your app has the mystical notification policy access.</li><li>If not, open the settings screen and ask nicely.</li><li>Once granted, use your powers wisely.</li></ol><p>Here’s a little function to lead your users on this spiritual journey:</p><pre>Future&lt;void&gt; guideThroughDNDPermission() async {<br> bool hasAccess = await dndPlugin.hasNotificationPolicyAccess();<br>if (!hasAccess) {<br>   // Show a dialog explaining why you need this power<br>   bool userWantsToGrantAccess = await showExplanationDialog();<br>   <br>   if (userWantsToGrantAccess) {<br>   await dndPlugin.openNotificationPolicyAccessSettings();<br>   // Wait for the user to return and check again<br>   // You might want to use a stream or state management for this<br>   } else {<br>   print(&quot;User chose to remain in the noisy world. Their loss!&quot;);<br>   }<br> } else {<br> print(&quot;We have the power! Use it wisely, young padawan.&quot;);<br> }<br>}</pre><h3>Conclusion: Embrace the Silence</h3><p>And there you have it, folks! With the `do_not_disturb` plugin, you now have the power to check, control, and conquer the world of notifications. Use this power wisely — remember, with great silence comes great productivity (and possibly missed calls from your boss, but that’s a problem for future you).</p><p>So go forth, build amazing Flutter apps, and may the zen be with you! 🧘‍♂️📱✨</p><p>P.S. If you’re feeling particularly zen-like, why not [star the repository](<a href="https://github.com/nateshmbhat/flutter_do_not_disturb">flutter_do_not_disturb</a>) or [buy the developer a coffee](<a href="https://www.buymeacoffee.com/nateshmbhat">https://www.buymeacoffee.com/nateshmbhat</a>)? After all, good karma is the best notification of all!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=704359b67e6c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Flutter Vs React Native : Performance Benchmarks you can’t miss ! ⚡️]]></title>
            <link>https://nateshmbhat.medium.com/flutter-vs-react-native-performance-benchmarks-you-cant-miss-%EF%B8%8F-2e31905df9b4?source=rss-c2c2a0e489e7------2</link>
            <guid isPermaLink="false">https://medium.com/p/2e31905df9b4</guid>
            <category><![CDATA[react-native]]></category>
            <category><![CDATA[flutter]]></category>
            <category><![CDATA[apps]]></category>
            <category><![CDATA[app-development]]></category>
            <category><![CDATA[mobile-app-development]]></category>
            <dc:creator><![CDATA[Natesh Bhat]]></dc:creator>
            <pubDate>Tue, 13 Aug 2024 11:51:27 GMT</pubDate>
            <atom:updated>2024-08-31T13:47:17.033Z</atom:updated>
            <content:encoded><![CDATA[<h3>Flutter Vs React Native : Performance Benchmarks you can’t miss ! 🔥⚡️ [Part -1]</h3><p>When it comes to mobile app development, the choice between Flutter and React Native often hinges on performance considerations. While both frameworks have their advocates, actual benchmark results are what many of us care for, especially for apps that can have complex UI/UX.</p><p>We were surprised to see very few benchmark comparisons for flutter and React Native and <a href="https://medium.com/swlh/flutter-vs-native-vs-react-native-examining-performance-31338f081980">even those were quite old</a>. So with current versions of Flutter &amp; React Native, here’s a thorough benchmarking comparison in terms of <strong>App size, Memory &amp; CPU</strong> <strong>usage</strong>. The results are interesting and surprising to say the least.</p><h3>Benchmarked Apps</h3><p>Three types of apps were benchmarked:</p><ol><li><strong>Large Listview of 1000 items</strong>: Each item includes a static image and an infinitely rotating image.</li><li><strong>Bulk Images Animation</strong>: 200 images are simultaneously rotated, faded, and scaled.</li><li><strong>Bulk Lottie Animations</strong>: 36 Lottie animations displayed on a single screen.</li></ol><h3>Benchmarking Methodology</h3><ul><li>Each benchmark was conducted on an <strong>M1 Mac</strong>.</li><li>The benchmark apps were run on Oneplus 7 device for Android and iPhone 8 and iPhone 15 for IOS.</li><li>In Android,<strong> Memory &amp; CPU</strong> usage were measured using <a href="https://developer.android.com/studio/profile"><strong>Android Profiler</strong></a><strong> </strong>on the release apk for both flutter and React Native. <br>In IOS, each benchmark app was run in release mode. <strong>Memory &amp; CPU usages </strong>were measured using Allocations &amp; Time Profiler instruments respectively.</li><li>FPS was measured on <strong>“profile mode” for Flutter app</strong> and a <strong>JS minified &amp; dev mode disabled </strong>build for React Native but with metro bundler connection. (Since <a href="https://reactnative.dev/docs/performance#what-you-need-to-know-about-frames">RN performance stats</a> are supported only when connected to metro bundler)</li><li>Here’s the benchmarking source code : <a href="https://github.com/nateshmbhat/flutter-rn-performance-benchmarks">https://github.com/nateshmbhat/flutter-rn-performance-benchmarks</a></li><li>For animation , built in Animation api is used in Flutter and industry standard Reanimated v3 is used in React Native.</li></ul><h4>Framework Versions Used :</h4><p>React Native : <strong>0.74.1</strong> | Flutter: <strong>3.19.5</strong> | Dart: <strong>3.3.3</strong></p><h3>Listview Performance</h3><p>This app uses built in ListView and Animated api in flutter. Uses the latest Reanimated v3 library in React native and built in FlatList. Both the lists are being used as is without explicit optimisation or fixed height extents.</p><p><a href="https://github.com/nateshmbhat/flutter-rn-performance-benchmarks/tree/main/list_view_with_images/results">Click to see : Benchmark Results</a></p><h4>Android :</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*1-_PKdiWz_JP9tq-" /><figcaption>Android list items benchmark</figcaption></figure><p><strong>Flutter :</strong></p><ul><li>FPS: 60 (no jank)</li><li>Dart heap memory usage: 7–8 MB</li><li>APK size: 16.8 MB (7.6 seconds to build)</li><li>Process memory: 120–130 MB (remained same when scrolling)</li><li>CPU: 5–8% (remained same when scrolling)</li><li>No jank or framedrops even on rapid scrolling</li></ul><p><strong>React Native :</strong></p><ul><li>FPS: 50–55 (noticeable jank)</li><li>APK size: 21.9 MB (23 seconds to build)</li><li>Memory: 160 MB (when not scrolling), 180–190 MB (scrolling)</li><li>CPU: 11–13% (when not scrolling), spiking to 25–30% (scrolling)</li><li>Occasional framedrops and blank items while scrolling</li></ul><h4>IOS :</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*iiqizLgczmia9zFS2DRDFw.png" /><figcaption>IOS list items benchmark</figcaption></figure><p><strong>Flutter :</strong></p><ul><li>Memory: <br>iPhone8 : <strong>48 MB</strong> (remained same when scrolling)<br>iPhone15: <strong>93 MB</strong> (remained same when scrolling)</li><li>CPU: <strong>40–60%</strong> (remained same when scrolling)</li><li>No jank or framedrops on rapid scrolling</li><li>Archive size: 71.5 MB (Build Time : 30 seconds)</li></ul><p><strong>React Native :</strong></p><ul><li>Memory: <br>iPhone8 : <strong>1.05 GB and drops to ~453 MB on scrolling.<br></strong>iPhone15 : <strong>1.12 GB and drops to ~430 MB on scrolling.</strong></li><li>CPU: <strong>120–140%</strong> for first 5 seconds, then drops to ~50% when not scrolling. Spikes to 140–150% on scrolling.</li><li>No framedrops but sometimes abruptly stops the scrolling when scrolled rapidly (user has to scroll again)</li><li>Archive size: <strong>112.3 MB</strong> (Build Time : <strong>318 seconds</strong>)</li></ul><p><strong>Android :<em> Flutter scrolled smoothly with no memory or CPU spikes during scrolling, whereas React Native exhibited frame drops and significant CPU &amp; Memory spikes.</em></strong></p><p><strong>IOS<em> : Very high memory usage in React Native. Flutter uses least amount of memory and CPU. Interestingly on ios, unlike android, memory usage drops to half in React Native on scrolling while CPU usage spikes.</em></strong></p><h3>Bulk Images Animation</h3><p><a href="https://github.com/nateshmbhat/flutter-rn-performance-benchmarks/tree/main/image_anims_in_one_screen/results">Click to see : Benchmark Results</a></p><h4>Android :</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*RJk_jXw6Zbd30bRZ" /><figcaption>Android : Bulk Images Animation benchmark results</figcaption></figure><p><strong>Flutter :</strong></p><ul><li>FPS: 58–60</li><li>Dart Heap: 13.4 MB</li><li>APK size: 11.6 MB (19.6 seconds to build)</li><li>Memory: 128–135 MB</li><li>CPU: 8%</li></ul><p><strong>React Native :</strong></p><ul><li>FPS: 58–60 (with occasional drops)</li><li>APK size: 21 MB (20 seconds to build)</li><li>Memory: 380–396 MB</li><li>CPU: 12–16%</li></ul><h4>IOS :</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xJjZJQ8Pq8nZVdGd9BzWHg.png" /></figure><p><strong>Flutter:</strong></p><ul><li>Memory: <br>iPhone8 : 49 MB<br>iPhone15 : 94 MB</li><li>CPU: <br>iPhone 8 : 50–60% <br>iPhone 15 : 23%</li><li>No visible janks or framedrops</li><li>Archive size: 71.2 MB (Build Time : 33 seconds)</li></ul><p><strong>React Native:</strong></p><ul><li>Memory: <br>iPhone 8 : 1.2 GB (Low Memory warning and eventual crash)<br>iPhone 15 : 1.38 GB (No crash)</li><li>CPU : <br>iPhone 8 : 135–140% (Eventual crash after about few seconds)<br>iPhone 15 : 200–300% for first 3.5 seconds then reduced to 50–60%</li><li>Severe janks, framedrops and eventual crash in iPhone 8. Only initial janks in iPhone 15</li><li>Archive size: 112.3 MB (Build Time : 317 seconds)</li></ul><p><strong>Android :<em> Both frameworks maintained a similar FPS, but React Native experienced FPS drops at animation restarts and significantly higher memory usage.</em></strong></p><p><strong>IOS<em> : High CPU and very high memory spike in React Native and eventual crash in older iPhones. Flutter seems to run fine with lower resource consumption.</em></strong></p><h3>Bulk Lottie Animations</h3><p><a href="https://github.com/nateshmbhat/flutter-rn-performance-benchmarks/tree/main/lottie_anims/results">Click to see : Benchmark Results</a></p><h4><strong>Android :</strong></h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*YRNnRJPdBYu5Kol2" /><figcaption>Bulk Lottie Animation benchmark results</figcaption></figure><p><strong>Flutter :</strong></p><ul><li>FPS: 36</li><li>APK size: 7.6 MB (20.8 seconds to build)</li><li>Memory: 220 MB</li><li>CPU: 11%</li></ul><p><strong>Android :</strong></p><ul><li>FPS: 43</li><li>APK size: 18.5 MB (23 seconds to build)</li><li>Memory: 240 MB</li><li>CPU: 22%</li></ul><h4><strong>IOS :</strong></h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xYcM4sPlHZNlB6wiidzlIg.png" /></figure><p><strong>Flutter :</strong></p><ul><li>Memory: <br>iPhone8 : 150–165 MB<br>iPhone15 : Increased up to 255 MB and remained at 255 MB</li><li>CPU: <br>iPhone8 : 120–140% <br>iPhone15 : 50–70%</li><li>Archive size: 72.9 MB (Build Time : 29 seconds)</li></ul><p><strong>React Native:</strong></p><ul><li>Memory: Increased slowly upto 125 MB and remained at 125 MB (both iPhones)</li><li>CPU: <br>iPhone8 : 100% for 6 seconds then reduces to 50–60% <br>iPhone15 : 60–80%</li><li>Archive size: 133.5 MB (Build Time : 304 seconds)</li></ul><p><strong><em>Android : React Native had higher FPS than Flutter but had higher CPU and memory usage.</em></strong></p><p><strong>IOS: <em>React Native consumed less memory and CPU than Flutter.</em></strong></p><blockquote><strong>New Update</strong> : As per flutter’s recent <strong>3.22</strong> release article, <a href="https://medium.com/flutter/whats-new-in-flutter-3-22-fbde6c164fe3"><strong>lottie rendering performance has significantly improved</strong> for it’s impeller rendering engine.</a> This benchmark uses Flutter 3.19.5. I’m yet to run this benchmark with impeller for this new version on android.</blockquote><h3>Starter Project Comparisons</h3><p>When we first create a project in react native or flutter, an unexpected finding was that the <strong>Apk Size </strong>for react native is significantly higher than that of flutter app for the starter project.</p><p>Although this particular case is not a perfectly fair comparison (given the fact both starter apps are different) and can be excluded as a performance benchmark result, it does present some interesting observations.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*Vxw7uZQnug6Kqu7B" /><figcaption>Starter Project Build time and App Size comparisons in Android</figcaption></figure><p>Note that app creation time is more in React Native since the entire react-native library gets downloaded when setting up a new project while flutter sdk already needs to be downloaded before creating new app.</p><h3>Final Verdict</h3><blockquote>Android : All benchmarks consistently showed Flutter’s APK sizes were smaller, and its CPU &amp; memory usage was marginally better than that of React Native.</blockquote><blockquote>IOS : Very high resource consumption by React Native compared to flutter except in Bulk Lottie animations app where flutter consumed more resources.</blockquote><h4>Next Steps :</h4><ul><li>Do performance benchmark with React Native’s <strong>New Architecture enabled</strong>.</li><li>Improve visualization of the metrics for ios.</li></ul><h4>Upcoming posts</h4><p>Will soon be posting more detailed benchmark comparisons along with snapshot data for memory and cpu usages.<br>So, stay tuned for that !</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/500/0*dYbJK5NYUGUlm5Xk.gif" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2e31905df9b4" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Complete your GRPC development and testing workflow with Tap-Rpc]]></title>
            <link>https://medium.com/geekculture/complete-your-grpc-development-and-testing-workflow-with-tap-rpc-bc185a1adce5?source=rss-c2c2a0e489e7------2</link>
            <guid isPermaLink="false">https://medium.com/p/bc185a1adce5</guid>
            <category><![CDATA[web]]></category>
            <category><![CDATA[testing]]></category>
            <category><![CDATA[api]]></category>
            <category><![CDATA[development]]></category>
            <category><![CDATA[technology]]></category>
            <dc:creator><![CDATA[Natesh Bhat]]></dc:creator>
            <pubDate>Mon, 28 Jun 2021 15:58:45 GMT</pubDate>
            <atom:updated>2021-06-30T07:07:22.230Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/820/1*P-MdnTZeKk05oaiqeR2Ymw.png" /></figure><p>We at <a href="http://gonuclei.com/">Nuclei</a> use GRPC for our services, mainly because of the rich benefits that GRPC provides such as simple service definition through proto files, offers scalability, works across languages and platforms, supports bi-directional streaming with http2 support.</p><p>Despite these benefits, we address some of the inevitable limitations and nuances of the development and testing workflow which we often face.</p><h3>Problems that we are trying to solve</h3><p>The usual development and testing workflow of any project involves frontend and backend development along with QA who will be testing the project.</p><p><strong>Frontend Concerns</strong> : Frontend developers always have the dependency on backend services to be ready, in order to build the UI and many times services would still be work in progress. Without services ready, front-end would have to manually mock the responses, which is extremely time-consuming and tedious.</p><p><strong>Backend Concerns</strong> : Testing of backend service solely depends on the test-case that they write, and backend developers cannot modify requests in real-time coming from the application client which makes it harder to find issues.</p><p><strong>QA Concerns</strong> : Mocking certain edge cases and scenarios at backend or client side is very difficult and time consuming which often requires changes in codebase of both client and backend. QA has no easy way to see or modify the request and responses exchanged between client and backend in real time which reveal bugs and unhandled cases many times.</p><p>To tackle all the above concerns, we needed a tool that’s sits in between frontend ( client-app ) and backend ( service )</p><ul><li>A tool that lets developers see the request and response in a better way in real time, than the usual logs.</li><li>Allows you to modify the request and responses exchanged between client app and server in real time.</li><li>Allows you to generate mock responses with relevant information needed for the client-app and provide ways to send mock requests to services.</li></ul><p>And that is where <strong>Tap-Rpc</strong> (inspired by Bloom Rpc grpc client) comes into picture.</p><h3>TAP-RPC :</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*QhyqcxM8x-U1jhSfHf2NdA.png" /></figure><p>Tap-Rpc solves the above problems in an effective way.</p><ul><li>Eases frontend development and unblocks frontend even when backend apis are not ready.</li><li>Significantly reduces dependency among developers and QA</li><li>Helps in effective app and backend testing</li></ul><h3>Installation and Usage :</h3><ul><li>Download the latest release from here : <a href="https://github.com/nateshmbhat/tap-rpc/releases">https://github.com/nateshmbhat/tap-rpc/releases</a></li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*WVVDVDfWqAZZwc-k" /></figure><ul><li>In your client app, set the host and port for the grpc service to point to the Tap-Rpc’s ip address and port shown at the middle of the Tap-Rpc window title. In above picture, the address as shown in the window title is “<strong>192.168.6.54:50051</strong>”.</li><li>Set the default target server URL, which is needed for the monitor and client modes. In above picture, the target server is set as “dev-01.gonuclei.com:80” which is the address of the real grpc service.</li><li>User selects the proto files containing the services. Once you select proto files, all the services and Rpcs in those services appear in a tree structure and you can select a particular rpc.</li><li>At this point any requests coming from the client will be forwarded to the default target server in the background.</li><li>Now your client app is ready to access all the features of Tap-Rpc.</li></ul><p>Tap-Rpc provides 3 modes of operation, each addressing a different set of problems :</p><ol><li>Mock Rpc mode</li><li>Monitor mode</li><li>Client mode</li></ol><h3>Mock Rpc mode :</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*-RfN5gO0wa_vMLWy" /><figcaption>Mock Rpc mode in action</figcaption></figure><p>Mock rpc mode serves mock responses to clients which can be customized. This is the mode which tries to remove the dependency on backend services. The client just needs to point to the address given in the window title to start using Tap-Rpc as a mock server.</p><ul><li>The generated mock response is dependent on the proto property names and the type of the properties which helps in generating realistic responses. Eg : image urls, address text, phone, File paths, email, names etc.</li><li>The mock response can be <strong>edited</strong> on the fly.</li><li>You can also choose to <strong>send</strong> <strong>mock grpc error codes</strong> with custom error messages to test error scenarios.</li><li>In essence, every user of the tap-rpc tool will have the ability to have his own server that serves requests with realistic responses without the client needing to manually hardcode mock responses.</li></ul><h3>Monitor mode :</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*clVH7veYlZK3hHwu" /><figcaption>Monitor mode in action</figcaption></figure><p>Monitor mode is the most interesting feature of Tap-Rpc which acts as a bridge between the application client and the real backend server (target server) and allows you to monitor and modify the requests and responses in real time.</p><ul><li>Users can see the requests and responses in real time as they are exchanged between client app and the server.</li><li>Users can modify requests coming from clients before it reaches the backend.</li><li>Users can modify responses coming back from the server before it reaches the client.</li><li>User can also access and modify metadata, toggle TLS usage, forward errors, change target server to forward the requests on the fly etc.</li></ul><h3>Client mode :</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*35cJlZGTYC2KNMAz" /><figcaption>Client mode in action</figcaption></figure><p>In client mode, you can independently test your backend services by sending realistic mock requests to the target server.</p><ul><li>Use the client mode to send realistic mock requests to the backend services.</li><li>Helps in testing each backend service and rpc for all various scenarios.</li><li>Generates mock requests with realistic data.</li></ul><h3>Using REST and not GRPC ?</h3><p>If you’re using a REST service instead of grpc, we still got you covered with another tool built specifically for handling REST services and requests which you can get here : <a href="https://github.com/nateshmbhat/tap-rest/releases">https://github.com/nateshmbhat/tap-rest/releases</a></p><blockquote>This post is co-authored by <a href="https://medium.com/u/94450a387c10">manu kj</a></blockquote><blockquote><em>Don’t forget to Smash that </em><em>CLAP button 👏🏻 before you go :)</em></blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/299/0*JSklfL8Hku31_ttG.gif" /><figcaption>Clapping is Caring :)</figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=bc185a1adce5" width="1" height="1" alt=""><hr><p><a href="https://medium.com/geekculture/complete-your-grpc-development-and-testing-workflow-with-tap-rpc-bc185a1adce5">Complete your GRPC development and testing workflow with Tap-Rpc</a> was originally published in <a href="https://medium.com/geekculture">Geek Culture</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Custom Shaking Error widget to fully control form field errors in Flutter]]></title>
            <link>https://medium.com/flutter-community/creating-the-animated-shaking-error-widget-for-extensive-control-on-form-field-errors-5713d3cd44e1?source=rss-c2c2a0e489e7------2</link>
            <guid isPermaLink="false">https://medium.com/p/5713d3cd44e1</guid>
            <category><![CDATA[ui]]></category>
            <category><![CDATA[app-development]]></category>
            <category><![CDATA[flutter-ui]]></category>
            <category><![CDATA[flutter-widget]]></category>
            <category><![CDATA[flutter]]></category>
            <dc:creator><![CDATA[Natesh Bhat]]></dc:creator>
            <pubDate>Sat, 03 Oct 2020 16:38:24 GMT</pubDate>
            <atom:updated>2020-10-11T23:39:36.568Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*gnv5L1LYwQctXi4hTm74nQ.gif" /></figure><p>More often than not, forms and form fields are coupled with form field validations that validate the text inside the field and show an error if the validation failed.</p><p>Before we move into looking at creating a custom animated error widget, let’s see the <strong>Good</strong>, the <strong>Meh</strong> and the <strong>Bad</strong> of default form field provided by flutter.</p><h4>The Good :</h4><ul><li>Text Form field comes with a <strong>validator </strong>callback function that allows us to validate the given text on form submission.</li><li>There is an <strong>autovalidate</strong> parameter that will keep validating text on every change. (No need for the user to wait until form submission)</li></ul><pre>TextFormField(<br>  validator : (value){<br>      if(value.isEmpty) return &#39;Please enter some text&#39;;<br>       return null;<br>    },<br>   autoValidate : true<br> )</pre><h4>The Meh :</h4><ul><li>Since autovalidate validates the text on every keypress, it’s not a good experience for the user.</li><li>We might end up putting a lot of our business logic inside the validator of each field which harms testability.</li></ul><h4>The Bad :</h4><ul><li>There is no option to independently control the error shown below the form field based on the logic.</li><li>No options to control when the error animation (shake/jiggle) happens.</li><li>Many times we may need to show error text for the whole form or for a completely different widget that doesn’t have a form element but is purely based on business logic or API response.</li></ul><p>In this article, we’ll see how to create a fully controllable error widget that you can use anywhere on your screen with or without a form, control its <strong>visibility</strong>, <strong>mounting</strong>, <strong>animation</strong> and <strong>text</strong> all externally from your <strong>business logic components</strong>.</p><p>Here’s a sample of how its gonna look once we build it.<br>[ Full code snippet at the end :D ]</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/640/1*JRep8MgkoUqcyzqFaK0hmQ.gif" /><figcaption>Shaking Error Text in action</figcaption></figure><h3>Creating the ErrorController :</h3><p>Create a Controller or (ChangeNotifier) that allows us to control the <strong>visibility</strong>, <strong>error text and animation state </strong>independently.</p><p>We also make revealing with animation and initial visibility optional when the widget is first built. For animation, we use <a href="https://pub.dev/packages/simple_animations">simple_animations</a> to avoid having to manage animation controllers manually.</p><pre>class ShakingErrorController extends ChangeNotifier {<br>  String _errorText;<br>  bool _isVisible = true;<br>  bool _isMounted = true;</pre><pre>CustomAnimationControl _controlSignal = CustomAnimationControl.PLAY;</pre><pre>ShakingErrorController(<br>      {String initialErrorText = &#39;Error&#39;, bool revealWithAnimation = true, bool hiddenInitially = true})<br>      : _errorText = initialErrorText ?? &#39;&#39;,<br>        _isVisible = !hiddenInitially,<br>        _controlSignal = (revealWithAnimation ?? true) ? CustomAnimationControl.PLAY : CustomAnimationControl.STOP;<br>}</pre><h4>Expose methods to change the controller properties :</h4><p>We add the methods to hide, show, reveal, mount and unmount the error.</p><pre>class ShakingErrorController extends ChangeNotifier {<br>  ...<br>    <br>  bool get isMounted =&gt; _isMounted;<br>  <br>  bool get isVisible =&gt; _isVisible;</pre><pre>String get errorText =&gt; _errorText;</pre><pre>CustomAnimationControl get controlSignal =&gt; _controlSignal;</pre><pre>set errorText(String errorText) { <br>    _errorText = errorText;<br>    notifyListeners();<br>  }</pre><pre>void onAnimationStarted() {<br>    _controlSignal = CustomAnimationControl.PLAY;<br>  }</pre><pre>void shakeErrorText() {<br>    _controlSignal = CustomAnimationControl.PLAY_FROM_START;<br>    notifyListeners();<br>  }</pre><pre>///fully [unmount] and remove the error text<br>  void unMountError() {<br>    _isMounted = false;<br>    notifyListeners();<br>  }</pre><pre>///[remount] error text. will not be effective if its already mounted<br>  void mountError() {<br>    _isMounted = true;<br>    notifyListeners();<br>  }</pre><pre>///hide the error. but it will still be taking its space.<br>  void hideError() {<br>    _isVisible = false;<br>    notifyListeners();<br>  }</pre><pre>///just shows error without any animation<br>  void showError() {<br>    _isVisible = true;<br>    notifyListeners();<br>  }</pre><pre>///shows error with the reveal [animation]<br>  void revealError() {<br>    showError();<br>    shakeErrorText();<br>  }<br>}</pre><h3>Making ShakingErrorText widget :</h3><p>ShakingErrorText widget contains the text and the animation builder to move the error text back and forth to perform the shaking animation.</p><ul><li>Create the shaking error text widget that takes the controller and the number of times to shake the text as parameters.</li><li>Using the MultiTween class from <a href="https://pub.dev/packages/simple_animations">simple_animations</a>, we add tweens , each tween corresponding to movement of text along one direction.</li></ul><pre>enum ErrorAnimationProp { offset }</pre><pre>class ShakingErrorText extends StatelessWidget {<br>  final ShakingErrorController controller;<br>  final int timesToShake;<br>  final MultiTween&lt;ErrorAnimationProp&gt; _tween;</pre><pre>ShakingErrorText({<br>    this.controller,<br>    this.timesToShake = 4,<br>  }) : _tween = MultiTween&lt;ErrorAnimationProp&gt;() {<br>    List.generate(<br>        timesToShake,<br>        (_) =&gt; _tween<br>          ..add(ErrorAnimationProp.offset, Tween&lt;double&gt;(begin: 0, end: 10), Duration(milliseconds: 100))<br>          ..add(ErrorAnimationProp.offset, Tween&lt;double&gt;(begin: 10, end: -10), Duration(milliseconds: 100))<br>          ..add(ErrorAnimationProp.offset, Tween&lt;double&gt;(begin: -10, end: 0), Duration(milliseconds: 100)));<br>  }</pre><pre>...</pre><pre>}</pre><ul><li>Here, we want to move the text from original position to 10 pixels right and then back again to original position and repeat it with left side movement.</li><li>We repeat adding the above tweens for timesToShake number of times to get the continuous shaking effect.</li></ul><p><strong>Building the full widget : <br></strong>With the tweens in place, we can now build the final widget that consumes the ShakingErrorController and uses the above created tween for translation animation of text.</p><ul><li>For controlling the animation, we use the CustomAnimation widget from simple_animations.</li><li>In the animation builder, we translate the error text on x-axis and keep y-axis offset at 0.</li></ul><pre><a href="http://twitter.com/override">@override</a><br>  Widget build(BuildContext context) {<br>    return ChangeNotifierProvider&lt;ShakingErrorController&gt;.value(<br>      value: controller,<br>      child: Consumer&lt;ShakingErrorController&gt;(<br>        builder: (context, errorController, child) {<br>          return CustomAnimation&lt;MultiTweenValues&lt;ErrorAnimationProp&gt;&gt;(<br>            control: errorController.controlSignal,<br>            curve: Curves.easeOut,<br>            duration: _tween.duration,<br>            tween: _tween,<br>            animationStatusListener: (status) {<br>              if (status == AnimationStatus.forward) {<br>                controller._onAnimationStarted();<br>              }<br>            },<br>            builder: (BuildContext context, Widget child, tweenValues) {<br>              return Transform.translate(<br>                offset: Offset(tweenValues.get(ErrorAnimationProp.offset), 0),<br>                child: child,<br>              );<br>            },<br>            child: Visibility(<br>              visible: controller.isVisible &amp;&amp; controller.isMounted,<br>              maintainSize: controller.isMounted,<br>              maintainAnimation: controller.isMounted,<br>              maintainState: controller.isMounted,<br>              child: Text(<br>                errorController.errorText,<br>                textAlign: TextAlign.start,<br>                style: TextStyle(color: Colors.red)<br>              ),<br>            ),<br>          );<br>        },<br>      ),<br>    );<br>  }</pre><h3>Testing time !!!</h3><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fgiphy.com%2Fembed%2Ff6z5TkrTIBZILYOd1t%2Ftwitter%2Fiframe&amp;display_name=Giphy&amp;url=https%3A%2F%2Fmedia.giphy.com%2Fmedia%2Ff6z5TkrTIBZILYOd1t%2Fgiphy.gif&amp;image=https%3A%2F%2Fmedia0.giphy.com%2Fmedia%2Fv1.Y2lkPTc5MGI3NjExaml2MzEyY2NhZ3J4M3BpZnFrZmRnOTc1a2w0MzB3Mm5wMWwzM2lpNyZlcD12MV9naWZzX2dpZklkJmN0PWc%2Ff6z5TkrTIBZILYOd1t%2Fgiphy.gif&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=giphy" width="435" height="435" frameborder="0" scrolling="no"><a href="https://medium.com/media/bfbbcd4a5d3731ea8947a92c56b41650/href">https://medium.com/media/bfbbcd4a5d3731ea8947a92c56b41650/href</a></iframe><p>Now that we are done creating our widget, we can test it out with some buttons using the below widget.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/640/1*JRep8MgkoUqcyzqFaK0hmQ.gif" /><figcaption>Shaking Error Text in action</figcaption></figure><pre><em>class </em>TestShakingErrorText <em>extends </em>StatefulWidget {<br>  @override<br>  _TestShakingErrorTextState createState() =&gt; _TestShakingErrorTextState();<br>}<br><br><em>class </em>_TestShakingErrorTextState <em>extends </em>State&lt;TestShakingErrorText&gt; {<br>  <em>final </em>ShakingErrorController controller =<br>      ShakingErrorController(initialErrorText: &#39;Hello its me. Hi ....&#39;, hiddenInitially: <em>false</em>);<br><br>  @override<br>  Widget build(BuildContext context) {<br>    <em>return </em>Scaffold(<br>      body: Center(<br>        child: Column(<br>          children: [<br>            ShakingErrorText(<br>              controller: controller,<br>            ),<br>            RaisedButton(<br>              onPressed: () <em>async </em>{<br>                controller.isMounted ? controller.unMountError() : controller.mountError();<br>              },<br>              child: Text(&#39;toggle mounting&#39;),<br>            ),<br>            RaisedButton(<br>              onPressed: () <em>async </em>{<br>                controller.isVisible ? controller.hideError() : controller.showError();<br>              },<br>              child: Text(&#39;toggle visibility&#39;),<br>            ),<br>            RaisedButton(<br>              onPressed: () <em>async </em>{<br>                controller.shakeErrorText();<br>              },<br>              child: Text(&#39;shake test&#39;),<br>            )<br>          ],<br>        ),<br>      ),<br>    );<br>  }<br><br>  @override<br>  <em>void </em>dispose() {<br>    controller.dispose();<br>    <em>super</em>.dispose();<br>  }<br>}</pre><p>Here’s the full code snippet if you’re lazy ( like me 😉 ) :</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/fa2d1a163d88f4bf64420a03806f8dfc/href">https://medium.com/media/fa2d1a163d88f4bf64420a03806f8dfc/href</a></iframe><blockquote>Don’t forget to Smash that CLAP button 👏🏻 before you go :)</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/299/0*jJCFyS9u8jyc7rxo.gif" /><figcaption>Clapping is Caring :)</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*AuT0GXEd7j21RI6uXlxYTQ.png" /><figcaption><a href="https://www.twitter.com/FlutterComm">https://www.twitter.com/FlutterComm</a></figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5713d3cd44e1" width="1" height="1" alt=""><hr><p><a href="https://medium.com/flutter-community/creating-the-animated-shaking-error-widget-for-extensive-control-on-form-field-errors-5713d3cd44e1">Custom Shaking Error widget to fully control form field errors in Flutter</a> was originally published in <a href="https://medium.com/flutter-community">Flutter Community</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Bring your CustomPainter paintings to life in Flutter using Touchable]]></title>
            <link>https://medium.com/flutter-community/bring-your-custompainter-paintings-to-life-in-flutter-using-touchable-c2413cd1897?source=rss-c2c2a0e489e7------2</link>
            <guid isPermaLink="false">https://medium.com/p/c2413cd1897</guid>
            <category><![CDATA[flutter-ui]]></category>
            <category><![CDATA[flutter-widget]]></category>
            <category><![CDATA[custompainter]]></category>
            <category><![CDATA[touch]]></category>
            <category><![CDATA[flutter]]></category>
            <dc:creator><![CDATA[Natesh Bhat]]></dc:creator>
            <pubDate>Tue, 25 Aug 2020 15:18:39 GMT</pubDate>
            <atom:updated>2024-09-02T12:31:24.744Z</atom:updated>
            <content:encoded><![CDATA[<h3>How to animate CustomPainter shapes in Flutter ?</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*CCroaxiDdmSbRVAK.png" /></figure><p>Despite the humongous number of builtin widgets provided by the flutter library, we often find ourselves delving into :</p><ul><li>Creating complex widgets filled with domain-specific user interface elements.</li><li>Applying and orchestrating animations to multiple widgets.</li><li>Customising the look and feel of widgets which expands our creativity to a new level.</li></ul><p>To create such art of beauty and wonder, flutter offers us the CustomPainter Widget that gives us the raw power of “painting directly onto the canvas”.<br>With Flutter’s CustomPainter widget, you can draw pretty much anything on the canvas that you can imagine.</p><p><strong>&gt; But that’s NOT what we are here to talk about today… Nope.</strong></p><blockquote>Today, we’re here to learn how to <strong>make your paintings come alive</strong> ! ✨🍾</blockquote><blockquote>What we truly want, is not just the ability to create eye catching paintings, we want to transform the static paintings into user interface !</blockquote><blockquote>Though “CustomPainter&quot; provides limitless drawing possibilities, it comes with an expensive limitation.</blockquote><blockquote>That is, Interaction !</blockquote><p>Well, let me unbox the above words in simple terms, by stating out the <br><strong>limitations of </strong><strong>CustomPainter</strong>:</p><ul><li>The Custom Painter lets you only <strong>draw</strong> shapes on the canvas but many would want to let the <strong>user interact</strong> with the drawings.</li><li>You might wanna add <strong>touch</strong> and <strong>drag gestures</strong> to each shape or object you draw.</li><li>You may even wish to <strong>animate </strong>a specific <strong>shape</strong> when the user clicks on it.</li></ul><p>With CustomPainter all this was extremely difficult to manage…… <br>Until now.</p><h3>Presenting Touchable :</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/480/1*VsBxBDiDT5bHtPZixApOCw.gif" /></figure><p>To solve the above limitations of CustomPainter, I createdTouchable, a flutter library that allows you to add various types of gestures to every shape you draw on the canvas, thus making them touchable 😉 !</p><p>Let’s start by looking at an example of a widget that uses CustomPainter .</p><p>For simplicity, here’s a minimal page with 2 circles drawn on the canvas (blue and green) with a white background.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fcodepen.io%2Fnateshmbhat%2Fembed%2Fpreview%2FMWyjwmX%3Fheight%3D600%26slug-hash%3DMWyjwmX%26default-tabs%3Djs%2Cresult%26host%3Dhttps%3A%2F%2Fcodepen.io&amp;display_name=CodePen&amp;url=https%3A%2F%2Fcodepen.io%2Fnateshmbhat%2Fpen%2FMWyjwmX&amp;image=https%3A%2F%2Fassets.codepen.io%2F3181638%2Finternal%2Fscreenshots%2Fpens%2FMWyjwmX.default.png%3Ffit%3Dcover%26amp%3Bformat%3Dauto%26amp%3Bha%3Dfalse%26amp%3Bheight%3D360%26amp%3Bquality%3D75%26amp%3Bv%3D2%26amp%3Bversion%3D1597795003%26amp%3Bwidth%3D640&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=codepen" width="800" height="600" frameborder="0" scrolling="no"><a href="https://medium.com/media/1fdfc46be226500d6c0af96d2893566f/href">https://medium.com/media/1fdfc46be226500d6c0af96d2893566f/href</a></iframe><p>Those poor circles are just staying there, unmoving :( <br>So let’s add some interactivity to the shapes :)</p><h4>Adding Interactivity :</h4><ul><li>Add touchable as a dependency in your pubspec.yaml file.</li><li>Wrap your CustomPaint widget inside a CanvasTouchDetector .</li></ul><pre>CanvasTouchDetector(<br>   builder: (context) =&gt; CustomPaint(<br>        painter: MyPainter(context),<br>   ),        <br>)</pre><ul><li>Use the TouchyCanvas to draw your shapes and get access to gesture callbacks.</li></ul><pre>void paint(Canvas canvas, Size size) {<br>    TouchyCanvas touchyCanvas = TouchyCanvas(context, canvas);<br>    <br>    touchyCanvas.drawCircle( ... , onTapDown: (_) {<br>      print(&#39;You clicked BLUE circle&#39;);<br>    });<br>    <br>    touchyCanvas.drawCircle( ... , onLongPressStart: (_) {<br>      print(&#39;long pressed on GREEN circle&#39;);<br>    });<br>  }</pre><blockquote>Here’s the full code snippet given below :</blockquote><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/a5c793557a56003b89f73aa06888fbf2/href">https://medium.com/media/a5c793557a56003b89f73aa06888fbf2/href</a></iframe><p>As you might have guessed, pressing on the blue circle will print You clicked BLUE circle and <strong>long pressing </strong>on the green circle will run its corresponding callback function.</p><p>Alright, we just handled the gestures for the shapes but that’s not particularly exciting, is it? <strong>Time to add some excitement!</strong></p><h3>Animation Time!</h3><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fgiphy.com%2Fembed%2FWUq1cg9K7uzHa%2Ftwitter%2Fiframe&amp;display_name=Giphy&amp;url=https%3A%2F%2Fmedia.giphy.com%2Fmedia%2FWUq1cg9K7uzHa%2Fgiphy.gif&amp;image=https%3A%2F%2Fi.giphy.com%2Fmedia%2FWUq1cg9K7uzHa%2Fgiphy-downsized-large.gif&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=giphy" width="435" height="326" frameborder="0" scrolling="no"><a href="https://medium.com/media/4f29ebaeb21ef584de0924ab3f1b1fd3/href">https://medium.com/media/4f29ebaeb21ef584de0924ab3f1b1fd3/href</a></iframe><blockquote>Now, we add some magic to the painter : )</blockquote><h4>Setting the objectives :</h4><p>We start by setting an objective to create something like this below where we draw two circles :</p><ul><li>One circle can be dragged around. Let’s call it the earth!</li><li>The other circle toggles the background colour. This is our Sun!</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/352/1*9-dtwMpDdmO0AhJjkavuXA.gif" /><figcaption>The Sun and the Earth!</figcaption></figure><p>Since we now have a relatively complex set of changing values, we use a state management solution like the <strong><em>provider</em></strong> to handle it in a simple way.</p><h4>Create the ChangeNotifier (PaintingController) :</h4><p>First, we create a <em>ChangeNotifier that contains all the painter related properties like the position of the circle and the background colour and methods that change these state properties.</em></p><pre><em>class </em>PaintingController <em>extends </em>ChangeNotifier {<br>  Offset circlePosition;<br>  Color backgroundColor = Colors.<em>white</em>;<br><br>  <em>void </em>moveCircle(Offset newPosition) {<br>    circlePosition = newPosition;<br>    notifyListeners();<br>  }<br><br>  <em>void </em>toggleBackGroundColor() {<br>    backgroundColor = backgroundColor == Colors.<em>white </em>? Colors.<em>blueGrey </em>: Colors.<em>white</em>;<br>    notifyListeners();<br>  }<br>}</pre><h4>Create the Painter :</h4><p>We create the <em>painter </em>class keeping the following points in mind :</p><ul><li>Return true from the shouldRepaint method to enable repaints and animations.</li></ul><pre>@override<br>bool shouldRepaint(CustomPainter oldDelegate) {<br>  <em>return true</em>;<br>}</pre><ul><li>Get ahold of the PaintingController inside the painter by using the context we send to the painter constructor.</li></ul><pre><em>class </em>MyPainter <em>extends </em>CustomPainter {<br>  <em>final </em>BuildContext context;<br>  <em>final </em>PaintingController controller;<br><br>  MyPainter(<em>this</em>.context) : controller = Provider.<em>of</em>&lt;PaintingController&gt;(context, listen: <em>false</em>);</pre><pre>...</pre><ul><li>In the paint method we draw the earth and the sun :D</li></ul><pre>@override<br><em>void </em>paint(Canvas canvas, Size size) {<br>  TouchyCanvas touchyCanvas = TouchyCanvas(context, canvas);<br>  drawEarth(touchyCanvas, size);<br>  drawSun(touchyCanvas, size);<br>}</pre><ul><li>We make the earth draggle by adding a onPanUpdate callback and call the moveCircle method of the controller passing the new position.</li></ul><pre>canvas.drawCircle(..., onPanUpdate: (details) {<br>  controller.moveCircle(Offset(blueCircleCenter.dx + details.delta.dx * 2, blueCircleCenter.dy + details.delta.dy * 2));<br>});</pre><ul><li>Similarly, we make the sun toggle the background colour by adding onTapDown callback.</li></ul><pre>canvas.drawCircle(... , onTapDown: (_) {<br>  controller.toggleBackGroundColor();<br>});</pre><ul><li>To make it look cooler 😎 , we draw some more colourful circles to put some life into it.</li></ul><blockquote>Here’s the full code snippet for the Sun and the Earth , below :</blockquote><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/7a61fc92e18c984e00b1dcca88d0d5a4/href">https://medium.com/media/7a61fc92e18c984e00b1dcca88d0d5a4/href</a></iframe><p><strong>Touchable also handles :</strong></p><ul><li>The painting style (filled ▮ , stroke ▯) of the shape</li><li>Painting <strong>stroke width</strong>. So if your shapes are painted thick, it still got it covered.</li><li><strong>clipping</strong> and different <strong>clipping modes</strong>.</li><li><strong>HitTestBehavior</strong> for each shape, particularly useful when you want gestures on overlapping shapes.</li></ul><h4>Links :</h4><ul><li><a href="https://github.com/nateshmbhat/touchable">nateshmbhat/touchable</a></li><li><a href="https://pub.dev/packages/touchable">touchable | Flutter Package</a></li></ul><h4>Have you learnt the art of SMASHing ?</h4><h4>Time to test it on the Clap button below 👏🏻 !!!</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/299/0*XoXLstG69axCheXk.gif" /><figcaption>Clapping is Caring :)</figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c2413cd1897" width="1" height="1" alt=""><hr><p><a href="https://medium.com/flutter-community/bring-your-custompainter-paintings-to-life-in-flutter-using-touchable-c2413cd1897">Bring your CustomPainter paintings to life in Flutter using Touchable</a> was originally published in <a href="https://medium.com/flutter-community">Flutter Community</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[If I share MY ideas and resources with others then THEY could surpass ME.]]></title>
            <link>https://nateshmbhat.medium.com/if-i-share-my-ideas-and-resources-with-others-then-they-could-achieve-more-than-me-b1b9227f325c?source=rss-c2c2a0e489e7------2</link>
            <guid isPermaLink="false">https://medium.com/p/b1b9227f325c</guid>
            <category><![CDATA[self-improvement]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[motivation]]></category>
            <category><![CDATA[tech]]></category>
            <category><![CDATA[engineering]]></category>
            <dc:creator><![CDATA[Natesh Bhat]]></dc:creator>
            <pubDate>Fri, 05 Jun 2020 06:15:05 GMT</pubDate>
            <atom:updated>2020-06-05T14:37:11.171Z</atom:updated>
            <content:encoded><![CDATA[<h3>If I share MY ideas and resources with others then THEY could surpass ME. Isn’t it better to just do it ALONE ? 🤔</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/450/0*TcqkLQ6EbcqYjS47" /></figure><p>I have been recently asked this question by an engineer</p><blockquote>If I share MY ideas and resources with others then THEY could surpass ME. Isn’t it better to just do it ALONE ?</blockquote><p>This was a valid question and has a lot of perspectives to it. Let us look at this question in a way that does <strong>not reflect </strong>as to how one has the <strong>moral </strong>responsibility to do good to others , how one should be <strong>selfless </strong>and other mumbo jumbo ✌️ that floats around.</p><h4>What’s in it for you here ?</h4><p>Let’s take a <strong>pure business approach </strong>here and see how sharing your ideas , resources , collective learning , working for and with others will <strong>benefit you</strong> the most. <strong>Yes ! You !</strong></p><h4>But how …. ? You may ask🤔</h4><p>Let’s look at this from a utilitarian perspective of an engineer.</p><p>This very question comes from a <strong>wrong notion</strong> that when you don’t share your ideas or that when you hide useful things or resources from your peers and friends, you allow yourself to succeed and that you become more competent and better than them.</p><blockquote><strong><em>First of all , you should decide whether you want to be “an engineer” who is just better than the one sitting next to you</em></strong></blockquote><blockquote><strong><em>Or</em></strong></blockquote><blockquote><strong><em>you wanna be a great engineer no matter who’s sitting next to you, No matter the kind of problem thrown at you, no matter the situation and the place you are in, You are ready !</em></strong></blockquote><h4>Solutions are meaningless without the Right Problem !</h4><p>One tremendous advantage for you here is that your peers will learn and ask you questions or suggestions which will improve your understanding of the problem and solution at hand. Engineering is not just reading some stuff and just learning something in detail. It’s an art where you will continuously face a whole lot of issues and problems and have to come up with solutions that would work for you the best.</p><p>You will <strong>NEVER grow </strong>as an engineer if</p><ul><li>You don’t get the opportunity to see new problems and perspectives.</li><li>You don’t get the kind of questions that challenge you and your ideas however small or big they are.</li><li>Sharing and discussing is a great way to get such opportunity many times.</li></ul><blockquote>Solutions and answers are meaningless without questions and the right questions could be more important than some of the most important solutions.</blockquote><p>So what’s the issue if the guy <strong>next to you becomes great </strong>? Do you care more about <strong>your greatness </strong>or about <strong>other’s demise </strong>?</p><p>If you care about <strong>your greatness</strong>, which you should first, you should share. You must care. You have to. There’s no other way around it.</p><blockquote>You would help others <strong>not because it helps them</strong>, but because <strong>it helps you</strong>. It’s a <strong>win win situation</strong> for everyone you see ✌️</blockquote><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fgiphy.com%2Fembed%2Fh5cZakvQ2hq0Vw4G22%2Ftwitter%2Fiframe&amp;display_name=Giphy&amp;url=https%3A%2F%2Fmedia.giphy.com%2Fmedia%2Fh5cZakvQ2hq0Vw4G22%2Fgiphy.gif&amp;image=https%3A%2F%2Fi.giphy.com%2Fmedia%2Fh5cZakvQ2hq0Vw4G22%2Fgiphy.gif&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=giphy" width="435" height="435" frameborder="0" scrolling="no"><a href="https://medium.com/media/a2a823b01f2e965ef484a3c67673d4ff/href">https://medium.com/media/a2a823b01f2e965ef484a3c67673d4ff/href</a></iframe><p>The issue as to why few people may not be getting this simple logic could be because of various dogma fed by people about what is success and what is defined as greatness. Unfortunately some people (even parents many times) have glorified those who are competent to such an extent that you don’t get satisfied anymore <strong>if there comes a solution </strong>to a real problem.</p><p>You would only get satisfied if <strong>you are the one</strong> finding the solution. If <strong>you are the one </strong>who gets patted on the back.</p><p>This unfortunately fuels a society where “<strong>you doing something” , </strong>matters more than “<strong>what it is that you are doing”</strong>.</p><p>We often hear that :</p><blockquote>The heart of engineering lies in problem solving.</blockquote><p>When the focus is shifted from <strong>problem solving </strong>to <strong>problem solver , </strong>where is the question of good engineering anymore ?</p><p>Coming back to the question of whether to help others grow with us or not , when you help them grow along with you , when you nurture those who aren’t as fortunate or as lucky as you , you make them empowered and capable enough to compete with you that actually benefits you the most. If you just remain focused on yourself , you are in essence competing with incompetents and feeling satisfied with your success when it’s not even real.</p><blockquote>What’s the use if you are competing with “Losers” ? What’s the point of winning among the losers anyway?</blockquote><h4>So the question is again just this :</h4><p>Do you wanna be the engineer who’s just <strong>better among </strong>the <strong>losers club </strong>you are with or do you wanna be the engineer who’s <strong>great no matter</strong> who’s with whom.</p><blockquote>Do you wish to celebrate reaching the top of the Rock just above the ground being unaware of the whole mountain in front of you?</blockquote><blockquote>Or</blockquote><blockquote>Do you wanna conquer the mountain without minding a gang trekking along with you ?</blockquote><blockquote>You be the judge :)</blockquote><p>PS : Any disagreements and differences of opinion ? Looking forward to hearing them out 🙏 😅</p><h3>Before you go</h3><p>Thank you for taking the time to read this article. If you found it helpful, please hit the 👏 button 👏 (up to 50x) and share it to help others with similar interest find it! + Feedback is always welcome! 🙏</p><h4>Now , show me how you can “Smash” that clap button 💪</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/299/0*bvvzlpZziCQFHQHU.gif" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b1b9227f325c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Migrate your Flutter App to Flutter Web   [Updated ]]]></title>
            <link>https://medium.com/nuclei-technology-blog/migrate-your-flutter-app-to-flutter-web-updated-31f5be4db7b8?source=rss-c2c2a0e489e7------2</link>
            <guid isPermaLink="false">https://medium.com/p/31f5be4db7b8</guid>
            <category><![CDATA[dart]]></category>
            <category><![CDATA[flutter]]></category>
            <category><![CDATA[flutter-web]]></category>
            <category><![CDATA[mobile]]></category>
            <category><![CDATA[hummingbird]]></category>
            <dc:creator><![CDATA[Natesh Bhat]]></dc:creator>
            <pubDate>Fri, 27 Mar 2020 10:10:17 GMT</pubDate>
            <atom:updated>2020-05-07T09:04:19.495Z</atom:updated>
            <content:encoded><![CDATA[<h3>Migrate your Flutter App to Flutter Web 🥳 [Updated 🚀]</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*sXpR6elYk0iPlMB2XKZkSA.png" /><figcaption>Flutter Web</figcaption></figure><p>If you’re like me and wondering how you could make use of new flutter’s web app capability to deploy your cool flutter application out there as a web app , you’re absolutely at the right place :) At Nuclei , we are stepping up and upgrading our fin-tech products by venturing into the web domain.</p><h4>A note to the reader :</h4><p>Earlier versions of flutter required importing a separate package flutter_web to make it work for web which gave rise to many issues but now , on the newer versions , flutter web project is merged in the main flutter project and all the flutter magic you did to make your cool app now works with web with minimal changes. This article aims to help you migrate your flutter app to make it work on web too 😊 and setting the port and hostname for your web app.</p><p><strong>Let’s start our venture 😃 :</strong></p><ul><li>First , we gotta make sure that we are either on the flutter `beta` channel or the master channel.</li><li>Lets open a terminal and switch to the beta channel :</li></ul><pre>flutter channel beta<br>flutter upgrade</pre><ul><li>Then we enable the support for web by adding into flutter’s configuration :</li></ul><pre>flutter config --enable-web</pre><ul><li>To get a list of all the available config options , just enter<br> flutter config -h command.</li></ul><h4>Testing time 🍾 :</h4><ul><li>If the web support was successful , you should see 2 devices listed when you run flutter devices one is the Chrome device and another is Web Server device.</li></ul><pre>flutter devices<br>2 connected device:<br>Chrome • chrome • web-javascript • Google Chrome 78.0.3904.108<br>Web Server • web-server • web-javascript • Flutter Tools</pre><ul><li>You can run your app directly on chrome by selecting the chrome device in your IDE before you run the app or use the following command :</li></ul><pre>flutter run -d chrome</pre><ul><li>If you want to run it from some external web browser client or something else , it makes sense to just run the web server in which case you would do :</li></ul><pre>flutter run -d web-server</pre><h4>Any Errors 🛑 ?</h4><p>Well , if you get errors when building your app which complains that it’s not a web app, you just need to run the flutter create command on your project folder. Open the root of your project folder in a terminal and run :</p><pre>flutter create .</pre><p>[ There might also be a need to run it with the --org option sometimes : flutter create . --org com.exampleapp . Also make sure that your project folder name is made of only lower case letters and underscore]</p><h4>Specify custom port and hostname :</h4><p>Your app by default runs on localhost ( 127.0.0.1 ) which means you can’t access it from any other device on your network. If also uses a random port every time you run your app which is not what you want most of the time. <br>So , let’s fix this.</p><ul><li>Set the port using the --web-port option and set a globally accessible hostname ( 0.0.0.0 ) using the --web-hostname option like so when you run your app :</li></ul><pre>flutter run -d chrome --web-port 55555 --web-hostname 0.0.0.0</pre><ul><li>Now you can access your web app by opening any web browser in any of your devices in your local network by entering the following address :</li></ul><pre>http://&lt;your-ip-address&gt;:55555</pre><pre>http://192.168.0.100:55555</pre><ul><li>If you are running the web app from your IDE (Android studio or VScode) , you can specify the port and hostname by specifying the additional arguments option on Android Studio like so inside the Run Configurations</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*DR9AXKduqmCVwIB7esWLFQ.png" /></figure><ul><li>If you are using VSCode 🎊 , then you can set it in the launch.json options.</li></ul><pre>{<br>	&quot;name&quot;: &quot;Flutter&quot;,<br>	&quot;request&quot;: &quot;launch&quot;,<br>	&quot;type&quot;: &quot;dart&quot;,<br>	&quot;args&quot;: [&quot;--web-port&quot;, &quot;55555&quot; , &quot;--web-hostname&quot;,&quot;0.0.0.0&quot;]<br>}</pre><h3>Build and DEPLOY ! ! !</h3><p>Now that you have converted your app into a web app , its time to build it and deploy it on free static web hosting services like netlify.com or surge.sh . To build your web app , run the following command in your project.</p><pre> flutter build web</pre><p>This will create the build/web folder which you can deploy later.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=31f5be4db7b8" width="1" height="1" alt=""><hr><p><a href="https://medium.com/nuclei-technology-blog/migrate-your-flutter-app-to-flutter-web-updated-31f5be4db7b8">Migrate your Flutter App to Flutter Web  🥳 [Updated 🚀]</a> was originally published in <a href="https://medium.com/nuclei-technology-blog">Nuclei Technology Blog</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>