<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>DEV Community: Orlando Ezequiel Salazar Cruz</title>
    <description>The latest articles on DEV Community by Orlando Ezequiel Salazar Cruz (@ezequie1sc).</description>
    <link>https://dev.to/ezequie1sc</link>
    <image>
      <url>https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4153245%2F8fc2b649-ba64-47bb-9269-b66ea0e516d8.jpg</url>
      <title>DEV Community: Orlando Ezequiel Salazar Cruz</title>
      <link>https://dev.to/ezequie1sc</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/ezequie1sc"/>
    <language>en</language>
    <item>
      <title>PulseFlow: Bringing Real-Time Physiological Sensing to Your Desktop</title>
      <dc:creator>Orlando Ezequiel Salazar Cruz</dc:creator>
      <pubDate>Sun, 22 Nov 2026 05:47:15 +0000</pubDate>
      <link>https://dev.to/ezequie1sc/pulseflow-bringing-real-time-physiological-sensing-to-your-desktop-2m9g</link>
      <guid>https://dev.to/ezequie1sc/pulseflow-bringing-real-time-physiological-sensing-to-your-desktop-2m9g</guid>
      <description>&lt;p&gt;&lt;em&gt;This is a submission for the &lt;a href="https://dev.to/challenges/presage-2026-11-04"&gt;Presage SmartSpectra SDK Challenge&lt;/a&gt;&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  What I Built
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;PulseFlow — Your body, in real time.&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;I built &lt;strong&gt;PulseFlow&lt;/strong&gt;, a Windows desktop application that transforms an ordinary webcam into a real-time, contactless physiological sensing experience using the &lt;strong&gt;Presage SmartSpectra SDK&lt;/strong&gt;.&lt;/p&gt;

&lt;p&gt;The idea started with a simple question: &lt;em&gt;What if exploring physiological signals didn't require a smartwatch, fitness tracker, or specialized hardware?&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;I wanted to create something that went beyond displaying numbers. My goal was to make human-sensing technology approachable through a minimal desktop interface, live visualizations, and a guided breathing experience.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fmyxsdn1i1nkckbbk3qrj.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fmyxsdn1i1nkckbbk3qrj.png" alt="PulseFlow — Main Interface" width="800" height="455"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;h3&gt;
  
  
  Real-Time Physiological Monitoring
&lt;/h3&gt;

&lt;p&gt;PulseFlow uses a standard webcam and SmartSpectra to estimate two physiological signals:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Heart Rate:&lt;/strong&gt; Pulse rate in beats per minute (BPM).&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Breathing Rate:&lt;/strong&gt; Respiratory frequency in breaths per minute (br/min).&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;The application presents these measurements through live numerical indicators and time-series charts.&lt;/p&gt;

&lt;p&gt;It also provides camera status, signal-validation feedback, and clear waiting states when measurements are unavailable.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F97vmu6c2bau06jekycoo.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F97vmu6c2bau06jekycoo.png" alt="PulseFlow — Real-Time Physiological Monitoring" width="800" height="472"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;During testing with my webcam, SmartSpectra successfully provided live measurements including &lt;strong&gt;83.7 BPM&lt;/strong&gt; and &lt;strong&gt;7.2 breaths/min&lt;/strong&gt;.&lt;/p&gt;

&lt;p&gt;These were genuine measurements from a test session, not hardcoded or simulated values.&lt;/p&gt;

&lt;h3&gt;
  
  
  Guided Breathing Experience
&lt;/h3&gt;

&lt;p&gt;I wanted PulseFlow to offer an interactive experience rather than function only as a physiological monitoring dashboard.&lt;/p&gt;

&lt;p&gt;The application includes a guided breathing session with:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Configurable inhalation and exhalation durations.&lt;/li&gt;
&lt;li&gt;An animated breathing indicator.&lt;/li&gt;
&lt;li&gt;Start, pause, resume, and end controls.&lt;/li&gt;
&lt;li&gt;A synchronized session timer.&lt;/li&gt;
&lt;li&gt;Collection of valid physiological measurements during the exercise.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;The breathing animation follows the configured timer. It operates independently of the measured physiological signals and does not fabricate breathing-rate values.&lt;/p&gt;

&lt;h3&gt;
  
  
  Session Summaries
&lt;/h3&gt;

&lt;p&gt;After a breathing exercise, PulseFlow presents:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Session duration.&lt;/li&gt;
&lt;li&gt;Average observed heart rate.&lt;/li&gt;
&lt;li&gt;Average observed breathing rate.&lt;/li&gt;
&lt;li&gt;Valid pulse sample count.&lt;/li&gt;
&lt;li&gt;Valid breathing sample count.&lt;/li&gt;
&lt;li&gt;Session completion status.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;For example, one completed 60-second session produced:&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Measurement&lt;/th&gt;
&lt;th&gt;Observed Result&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Duration&lt;/td&gt;
&lt;td&gt;60 seconds&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Average heart rate&lt;/td&gt;
&lt;td&gt;84.9 BPM&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Average breathing rate&lt;/td&gt;
&lt;td&gt;8.0 breaths/min&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Valid pulse samples&lt;/td&gt;
&lt;td&gt;69&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Valid breathing samples&lt;/td&gt;
&lt;td&gt;15&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Status&lt;/td&gt;
&lt;td&gt;Completed&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;These are results from one demonstration session, not claims about physiological improvement.&lt;/p&gt;

&lt;h3&gt;
  
  
  Design and Experience
&lt;/h3&gt;

&lt;p&gt;PulseFlow follows a minimal, monochromatic design direction influenced by modern developer tools and Geist-inspired typography.&lt;/p&gt;

&lt;p&gt;I focused on large, readable measurements, subtle borders, smooth animations, contextual notifications, and an uncluttered interface.&lt;/p&gt;

&lt;p&gt;I wanted to make a complex sensing experience feel simple and understandable.&lt;/p&gt;




&lt;h2&gt;
  
  
  Demo
&lt;/h2&gt;

&lt;p&gt;I recorded a demonstration of PulseFlow running on Windows with a real webcam and the Presage SmartSpectra SDK.&lt;/p&gt;

&lt;p&gt;The video showcases the desktop interface, live heart rate and breathing rate estimates, guided breathing, and the completed session summary.&lt;/p&gt;

&lt;p&gt;  &lt;iframe src="https://www.youtube.com/embed/8ztd4aoxduk" width="710" height="399"&gt;
  &lt;/iframe&gt;
&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;&lt;a href="https://youtu.be/8ztd4aoxduk" rel="noopener noreferrer"&gt;Watch the PulseFlow demo on YouTube&lt;/a&gt;&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;For privacy, the camera imagery in the public recording is obscured. The displayed physiological measurements come from the live SmartSpectra integration.&lt;/p&gt;

&lt;p&gt;The application runs locally as an Electron desktop application. Public web hosting is not required to use it.&lt;/p&gt;




&lt;h2&gt;
  
  
  Code
&lt;/h2&gt;

&lt;p&gt;The complete source code is available in my public GitHub repository:&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;&lt;a href="https://github.com/Ezequie1Sc/PulseFlow" rel="noopener noreferrer"&gt;PulseFlow — GitHub Repository&lt;/a&gt;&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;The repository contains the desktop application, SmartSpectra integration, measurement-processing logic, breathing-session functionality, interface assets, tests, and setup instructions.&lt;/p&gt;

&lt;h3&gt;
  
  
  Technology Stack
&lt;/h3&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Technology&lt;/th&gt;
&lt;th&gt;Purpose&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Electron&lt;/td&gt;
&lt;td&gt;Desktop application runtime&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Node.js&lt;/td&gt;
&lt;td&gt;Native application and SDK services&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;TypeScript&lt;/td&gt;
&lt;td&gt;Application logic and type safety&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Presage SmartSpectra SDK&lt;/td&gt;
&lt;td&gt;Contactless physiological sensing&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Protocol Buffers&lt;/td&gt;
&lt;td&gt;Measurement decoding&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;HTML &amp;amp; CSS&lt;/td&gt;
&lt;td&gt;Interface and animations&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Canvas API&lt;/td&gt;
&lt;td&gt;Camera preview and live charts&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;esbuild&lt;/td&gt;
&lt;td&gt;Application bundling&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;h3&gt;
  
  
  How to Run PulseFlow
&lt;/h3&gt;

&lt;p&gt;&lt;strong&gt;1. Clone the repository&lt;/strong&gt;&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;git clone https://github.com/Ezequie1Sc/PulseFlow.git
&lt;span class="nb"&gt;cd&lt;/span&gt; PulseFlow
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;&lt;strong&gt;2. Install the dependencies&lt;/strong&gt;&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;npm ci
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;&lt;strong&gt;3. Configure the Presage API key&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;Create your local environment file from the included example.&lt;/p&gt;

&lt;p&gt;On Windows PowerShell:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight powershell"&gt;&lt;code&gt;&lt;span class="nc"&gt;Copy&lt;/span&gt;&lt;span class="o"&gt;-&lt;/span&gt;&lt;span class="nc"&gt;Item&lt;/span&gt; &lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;env&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;example&lt;/span&gt; &lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;env&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Configure the API key in &lt;code&gt;.env&lt;/code&gt;:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;SMARTSPECTRA_API_KEY=your_presage_api_key
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;&lt;strong&gt;4. Launch the application&lt;/strong&gt;&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;npm run dev
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;You will need a compatible webcam, camera permissions, and a valid Presage SmartSpectra API key.&lt;/p&gt;

&lt;p&gt;The repository's README includes additional documentation and troubleshooting guidance.&lt;/p&gt;

&lt;p&gt;No separate application backend or database is required.&lt;/p&gt;




&lt;h2&gt;
  
  
  How It Responds to Human Signals
&lt;/h2&gt;

&lt;p&gt;SmartSpectra is the foundation of PulseFlow.&lt;/p&gt;

&lt;p&gt;Rather than simulating physiological data, the application uses live measurements generated from webcam input.&lt;/p&gt;

&lt;h3&gt;
  
  
  1. Capturing Live Signals
&lt;/h3&gt;

&lt;p&gt;After the user explicitly consents to camera monitoring, PulseFlow initializes the SmartSpectra Node.js SDK and connects to the selected camera.&lt;/p&gt;

&lt;p&gt;The application requests these signals:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight typescript"&gt;&lt;code&gt;&lt;span class="n"&gt;requestedMetrics&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;[&lt;/span&gt;
  &lt;span class="nc"&gt;MetricType&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="no"&gt;PULSE_RATE&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
  &lt;span class="nc"&gt;MetricType&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="no"&gt;BREATHING_RATE&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
&lt;span class="p"&gt;]&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;SmartSpectra processes camera input and provides physiological measurements together with signal-quality information.&lt;/p&gt;

&lt;h3&gt;
  
  
  2. Handling SmartSpectra Events
&lt;/h3&gt;

&lt;p&gt;The application listens for SDK events including:&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;SDK Event&lt;/th&gt;
&lt;th&gt;Application Behavior&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;&lt;code&gt;processingStatus&lt;/code&gt;&lt;/td&gt;
&lt;td&gt;Updates processing information&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;code&gt;validationStatus&lt;/code&gt;&lt;/td&gt;
&lt;td&gt;Provides signal-quality guidance&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;code&gt;metrics&lt;/code&gt;&lt;/td&gt;
&lt;td&gt;Receives physiological estimates&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;code&gt;videoOutput&lt;/code&gt;&lt;/td&gt;
&lt;td&gt;Provides camera preview frames&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;code&gt;error&lt;/code&gt;&lt;/td&gt;
&lt;td&gt;Handles SDK-related errors&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;Physiological measurement messages are decoded using the SDK's Protocol Buffers definitions.&lt;/p&gt;

&lt;h3&gt;
  
  
  3. Processing and Validating Measurements
&lt;/h3&gt;

&lt;p&gt;One of my main technical priorities was avoiding misleading readings.&lt;/p&gt;

&lt;p&gt;PulseFlow processes the incoming measurements and filters samples before displaying them.&lt;/p&gt;

&lt;p&gt;If a measurement is unavailable or invalid, the interface communicates that state rather than generating an artificial value.&lt;/p&gt;

&lt;p&gt;For example, the dashboard can display a waiting state while SmartSpectra attempts to acquire an acceptable signal.&lt;/p&gt;

&lt;h3&gt;
  
  
  4. Real-Time Responses
&lt;/h3&gt;

&lt;p&gt;When SmartSpectra provides an accepted physiological measurement:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;PulseFlow updates the corresponding heart rate or breathing rate indicator.&lt;/li&gt;
&lt;li&gt;The historical chart incorporates the new reading.&lt;/li&gt;
&lt;li&gt;An active guided breathing session collects the valid sample.&lt;/li&gt;
&lt;li&gt;The accepted measurements contribute to the session summary.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;The animated breathing indicator operates independently of the measured respiration signal.&lt;/p&gt;

&lt;p&gt;This separation ensures that visual breathing guidance is not mistaken for real physiological data.&lt;/p&gt;

&lt;h3&gt;
  
  
  5. Integration Challenges
&lt;/h3&gt;

&lt;p&gt;During development, I encountered challenges involving camera framing, frame rate, signal quality, and SDK validation.&lt;/p&gt;

&lt;p&gt;For example, SmartSpectra provided feedback such as:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;&lt;code&gt;Place more of the chest in view.&lt;/code&gt;&lt;/li&gt;
&lt;li&gt;&lt;code&gt;Use a camera mode that provides at least 25 frames per second.&lt;/code&gt;&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;These messages helped me understand that successfully capturing video does not automatically mean a valid physiological signal is available.&lt;/p&gt;

&lt;p&gt;I adjusted the capture conditions and camera positioning until I could obtain live heart rate and breathing rate measurements.&lt;/p&gt;

&lt;p&gt;I also worked on separating native SDK processing from the Electron renderer, maintaining a modular architecture for camera handling, measurement processing, and user interface updates.&lt;/p&gt;

&lt;p&gt;This experience taught me the importance of testing with real hardware, handling unavailable measurements correctly, and designing interfaces that communicate uncertainty.&lt;/p&gt;




&lt;h2&gt;
  
  
  Prize Categories
&lt;/h2&gt;

&lt;p&gt;I am entering PulseFlow in &lt;strong&gt;both special prize categories&lt;/strong&gt;.&lt;/p&gt;

&lt;h3&gt;
  
  
  Most Creative or Delightful Use
&lt;/h3&gt;

&lt;p&gt;The inspiration behind PulseFlow was to make physiological sensing feel like a natural, interactive desktop experience.&lt;/p&gt;

&lt;p&gt;Instead of building only a traditional measurement dashboard, I combined real-time heart rate and breathing rate monitoring with guided breathing, subtle animations, live charts, and session summaries.&lt;/p&gt;

&lt;p&gt;I wanted the interface to feel calm and approachable while demonstrating the capabilities of contactless human sensing.&lt;/p&gt;

&lt;p&gt;PulseFlow brings together native desktop development, real physiological signals, and interaction design in an experience that does not require specialized wearable hardware.&lt;/p&gt;

&lt;h3&gt;
  
  
  Best Thoughtful Real-World Use
&lt;/h3&gt;

&lt;p&gt;PulseFlow explores a practical use of SmartSpectra: helping individuals observe physiological estimates during voluntary breathing exercises using an ordinary webcam.&lt;/p&gt;

&lt;p&gt;The experience is intended for people curious about physiological sensing and general wellness exploration, particularly those who do not have wearable monitoring devices.&lt;/p&gt;

&lt;p&gt;I designed PulseFlow around several usability principles:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Accessibility:&lt;/strong&gt; A standard webcam is the primary sensing device.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Consent:&lt;/strong&gt; Monitoring requires explicit user authorization.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Transparency:&lt;/strong&gt; The interface communicates when measurements are unavailable.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Control:&lt;/strong&gt; Users can start and stop monitoring.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Clarity:&lt;/strong&gt; Live measurements are presented through readable indicators and charts.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Privacy:&lt;/strong&gt; The application does not require user accounts or a dedicated physiological-history database.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;PulseFlow does not provide medical diagnoses, treatments, or clinical recommendations.&lt;/p&gt;

&lt;p&gt;It is an experimental demonstration of contactless sensing and user-controlled interaction.&lt;/p&gt;




&lt;h2&gt;
  
  
  Consent and Data Handling
&lt;/h2&gt;

&lt;p&gt;Because PulseFlow uses camera-based physiological sensing, consent, transparency, and responsible data handling were important parts of the application design.&lt;/p&gt;

&lt;h3&gt;
  
  
  Camera Consent
&lt;/h3&gt;

&lt;p&gt;PulseFlow requires explicit user consent before physiological monitoring begins.&lt;/p&gt;

&lt;p&gt;The application explains the purpose of camera access and provides controls for starting and stopping monitoring.&lt;/p&gt;

&lt;p&gt;The user remains in control of the camera-based experience.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F672e9mj1gse191uiy8nk.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F672e9mj1gse191uiy8nk.png" alt="PulseFlow — Camera Consent" width="800" height="449"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;h3&gt;
  
  
  Camera Input and Physiological Data
&lt;/h3&gt;

&lt;p&gt;PulseFlow uses webcam input through the Presage SmartSpectra SDK to obtain physiological estimates.&lt;/p&gt;

&lt;p&gt;Camera frames can be displayed in the desktop interface as a local preview.&lt;/p&gt;

&lt;p&gt;Accepted measurements are used to update:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Live physiological indicators.&lt;/li&gt;
&lt;li&gt;Historical charts.&lt;/li&gt;
&lt;li&gt;Guided-session observations.&lt;/li&gt;
&lt;li&gt;Session summaries.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;The public demonstration obscures camera imagery to protect the participant's identity.&lt;/p&gt;

&lt;h3&gt;
  
  
  Storage and Retention
&lt;/h3&gt;

&lt;p&gt;PulseFlow does not implement a dedicated persistent database for physiological measurements, user accounts, or session histories.&lt;/p&gt;

&lt;p&gt;The application's measurement-processing functionality is designed around temporary data used during monitoring and guided breathing sessions.&lt;/p&gt;

&lt;p&gt;This describes PulseFlow's application-level architecture. It does not establish the independent storage or retention behavior of Presage's SDK or services.&lt;/p&gt;

&lt;h3&gt;
  
  
  Data Sharing and SDK Telemetry
&lt;/h3&gt;

&lt;p&gt;PulseFlow configures the SmartSpectra SDK with:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight typescript"&gt;&lt;code&gt;&lt;span class="n"&gt;enableTelemetry&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="kc"&gt;false&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;This disables the configurable SDK telemetry option used by PulseFlow.&lt;/p&gt;

&lt;p&gt;However, this setting does not necessarily disable SDK authentication requests, required network communication, or any provider-side processing.&lt;/p&gt;

&lt;p&gt;The application does not implement an additional backend for uploading physiological session histories.&lt;/p&gt;

&lt;p&gt;For details regarding SDK-level data handling, refer to the official Presage documentation.&lt;/p&gt;

&lt;h3&gt;
  
  
  Optional AI Services
&lt;/h3&gt;

&lt;p&gt;PulseFlow does not require an external generative AI service for physiological measurements or guided breathing.&lt;/p&gt;

&lt;p&gt;Physiological estimates originate from the Presage SmartSpectra SDK.&lt;/p&gt;

&lt;h3&gt;
  
  
  Deletion and User Control
&lt;/h3&gt;

&lt;p&gt;Users can stop monitoring and manage or reset their active breathing session through application controls.&lt;/p&gt;

&lt;p&gt;PulseFlow does not provide persistent cloud-based physiological histories.&lt;/p&gt;

&lt;p&gt;Temporary measurement information is intended for the active application experience rather than long-term storage.&lt;/p&gt;

&lt;h3&gt;
  
  
  Responsible Use
&lt;/h3&gt;

&lt;p&gt;PulseFlow is an experimental application intended for education, technical demonstration, and general wellness exploration.&lt;/p&gt;

&lt;p&gt;It is &lt;strong&gt;not a medical device&lt;/strong&gt; and must not be used for:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Medical diagnosis or treatment.&lt;/li&gt;
&lt;li&gt;Clinical decision-making.&lt;/li&gt;
&lt;li&gt;Emergency monitoring.&lt;/li&gt;
&lt;li&gt;Covert surveillance.&lt;/li&gt;
&lt;li&gt;Non-consensual physiological monitoring.&lt;/li&gt;
&lt;/ul&gt;




&lt;p&gt;&lt;strong&gt;Session Summary Example&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;One of the final features I implemented was the measurement-based session summary.&lt;/p&gt;

&lt;p&gt;After completing a 60-second exercise, PulseFlow displayed the observed heart rate and breathing rate averages, valid sample counts, and completion status.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fxfuq1lietxus5ggvxqoh.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fxfuq1lietxus5ggvxqoh.png" alt="PulseFlow — Session Summary" width="800" height="600"&gt;&lt;/a&gt;&lt;/p&gt;




&lt;p&gt;Building PulseFlow taught me that real-time physiological sensing requires more than successfully connecting an SDK.&lt;/p&gt;

&lt;p&gt;Camera conditions, signal validation, measurement availability, native integration, and user experience must work together.&lt;/p&gt;

&lt;p&gt;The most valuable lesson was that &lt;strong&gt;an honest interface should explain when a measurement is unavailable instead of inventing data&lt;/strong&gt;.&lt;/p&gt;

&lt;p&gt;I am grateful to Presage and DEV Community for the opportunity to explore contactless human sensing through this challenge.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;PulseFlow — Your body, in real time.&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;GitHub:&lt;/strong&gt; &lt;a href="https://github.com/Ezequie1Sc/PulseFlow" rel="noopener noreferrer"&gt;https://github.com/Ezequie1Sc/PulseFlow&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Demo:&lt;/strong&gt; &lt;a href="https://youtu.be/8ztd4aoxduk" rel="noopener noreferrer"&gt;https://youtu.be/8ztd4aoxduk&lt;/a&gt;&lt;/p&gt;

</description>
      <category>devchallenge</category>
      <category>presagechallenge</category>
    </item>
    <item>
      <title>PulseFlow: Turning a Webcam into a Real-Time Human-Sensing Experience</title>
      <dc:creator>Orlando Ezequiel Salazar Cruz</dc:creator>
      <pubDate>Sun, 22 Nov 2026 05:43:52 +0000</pubDate>
      <link>https://dev.to/ezequie1sc/pulseflow-turning-a-webcam-into-a-real-time-human-sensing-experience-2hj6</link>
      <guid>https://dev.to/ezequie1sc/pulseflow-turning-a-webcam-into-a-real-time-human-sensing-experience-2hj6</guid>
      <description>&lt;p&gt;&lt;em&gt;This is a submission for the &lt;a href="https://dev.to/challenges/presage-2026-11-04"&gt;Presage SmartSpectra SDK Challenge&lt;/a&gt;.&lt;/em&gt;&lt;/p&gt;

&lt;h1&gt;
  
  
  PulseFlow: Turning a Webcam into a Real-Time Human-Sensing Experience
&lt;/h1&gt;

&lt;p&gt;&lt;strong&gt;Your body, in real time.&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;What if exploring your body's physiological signals didn't require a smartwatch, fitness tracker, or specialized hardware?&lt;/p&gt;

&lt;p&gt;What if all you needed was a webcam?&lt;/p&gt;

&lt;p&gt;That question inspired me to build &lt;strong&gt;PulseFlow&lt;/strong&gt;, a Windows desktop application powered by the &lt;strong&gt;Presage SmartSpectra SDK&lt;/strong&gt; that turns ordinary camera input into a real-time human-sensing experience.&lt;/p&gt;

&lt;p&gt;PulseFlow combines contactless heart rate and breathing rate estimation with live charts, guided breathing sessions, and meaningful session summaries.&lt;/p&gt;

&lt;p&gt;I wanted to explore how complex physiological sensing technology could become a simple, approachable, and privacy-conscious desktop experience.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fmyxsdn1i1nkckbbk3qrj.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fmyxsdn1i1nkckbbk3qrj.png" alt="PulseFlow — Your body, in real time" width="800" height="455"&gt;&lt;/a&gt;&lt;/p&gt;




&lt;h2&gt;
  
  
  What I Built
&lt;/h2&gt;

&lt;p&gt;I built &lt;strong&gt;PulseFlow&lt;/strong&gt;, a desktop application developed with &lt;strong&gt;Electron, Node.js, TypeScript, and Presage SmartSpectra&lt;/strong&gt;.&lt;/p&gt;

&lt;p&gt;The application uses a standard webcam to estimate physiological signals and presents them through a minimal interface designed for clarity and ease of use.&lt;/p&gt;

&lt;p&gt;Rather than building only a technical monitoring dashboard, I wanted to create an experience that people could interact with.&lt;/p&gt;

&lt;p&gt;PulseFlow has three main components:&lt;/p&gt;

&lt;h3&gt;
  
  
  1. Real-Time Physiological Monitoring
&lt;/h3&gt;

&lt;p&gt;PulseFlow requests two physiological measurements from SmartSpectra:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Heart Rate:&lt;/strong&gt; Estimated pulse in beats per minute (BPM).&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Breathing Rate:&lt;/strong&gt; Estimated respiratory frequency in breaths per minute (br/min).&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;The application displays accepted readings through live numerical indicators and time-series charts.&lt;/p&gt;

&lt;p&gt;Users can also see camera status, signal availability, and feedback when capture conditions need improvement.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F97vmu6c2bau06jekycoo.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F97vmu6c2bau06jekycoo.png" alt="PulseFlow — Live Physiological Monitoring" width="800" height="472"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;h3&gt;
  
  
  2. Guided Breathing
&lt;/h3&gt;

&lt;p&gt;I wanted users to do more than simply observe their measurements.&lt;/p&gt;

&lt;p&gt;PulseFlow includes a guided breathing experience with:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Configurable inhalation and exhalation durations.&lt;/li&gt;
&lt;li&gt;A minimal animated breathing indicator.&lt;/li&gt;
&lt;li&gt;Start, pause, resume, and end controls.&lt;/li&gt;
&lt;li&gt;A synchronized session timer.&lt;/li&gt;
&lt;li&gt;Physiological measurement collection during the exercise.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;The breathing animation is driven by the exercise timer, not by artificially generated physiological measurements.&lt;/p&gt;

&lt;h3&gt;
  
  
  3. Session Summaries
&lt;/h3&gt;

&lt;p&gt;At the end of a session, PulseFlow provides a summary containing:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Total session duration.&lt;/li&gt;
&lt;li&gt;Average observed heart rate.&lt;/li&gt;
&lt;li&gt;Average observed breathing rate.&lt;/li&gt;
&lt;li&gt;Number of valid samples for each signal.&lt;/li&gt;
&lt;li&gt;Session completion status.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;These statistics are calculated from accepted SmartSpectra measurements collected during the exercise.&lt;/p&gt;

&lt;h3&gt;
  
  
  Design Philosophy
&lt;/h3&gt;

&lt;p&gt;PulseFlow follows a minimalist, monochromatic design direction influenced by modern developer tools and the Geist design language.&lt;/p&gt;

&lt;p&gt;The interface emphasizes readable physiological metrics, subtle borders, restrained animations, real-time charts, and contextual notifications.&lt;/p&gt;

&lt;p&gt;My goal was to make complex sensing technology feel understandable without overwhelming the user.&lt;/p&gt;

&lt;h3&gt;
  
  
  Real Measurements, Not Simulations
&lt;/h3&gt;

&lt;p&gt;An important requirement I set for myself was that PulseFlow should not fabricate physiological readings.&lt;/p&gt;

&lt;p&gt;During a live test, SmartSpectra successfully provided measurements including:&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Signal&lt;/th&gt;
&lt;th&gt;Observed Value&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Heart rate&lt;/td&gt;
&lt;td&gt;83.7 BPM&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Breathing rate&lt;/td&gt;
&lt;td&gt;7.2 breaths/min&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;These values came from an actual demonstration rather than hardcoded data.&lt;/p&gt;

&lt;p&gt;When valid measurements are unavailable, the application displays an appropriate waiting state instead of inventing readings.&lt;/p&gt;




&lt;h2&gt;
  
  
  Demo
&lt;/h2&gt;

&lt;p&gt;I recorded a demonstration showing PulseFlow running with a real webcam and the Presage SmartSpectra SDK.&lt;/p&gt;

&lt;p&gt;The video showcases live physiological monitoring, the guided breathing experience, and a completed session summary.&lt;/p&gt;

&lt;p&gt;  &lt;iframe src="https://www.youtube.com/embed/8ztd4aoxduk" width="710" height="399"&gt;
  &lt;/iframe&gt;
&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;&lt;a href="https://youtu.be/8ztd4aoxduk" rel="noopener noreferrer"&gt;Watch the full demonstration on YouTube&lt;/a&gt;&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;For privacy, the public recording uses anonymized camera imagery while preserving the live physiological measurements displayed by the application.&lt;/p&gt;




&lt;h2&gt;
  
  
  Code
&lt;/h2&gt;

&lt;p&gt;The complete source code is publicly available on GitHub:&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;&lt;a href="https://github.com/Ezequie1Sc/PulseFlow" rel="noopener noreferrer"&gt;PulseFlow — GitHub Repository&lt;/a&gt;&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;The repository includes the Electron application, native SmartSpectra integration, measurement-processing logic, guided breathing functionality, tests, screenshots, and setup instructions.&lt;/p&gt;

&lt;h3&gt;
  
  
  Technology Stack
&lt;/h3&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Technology&lt;/th&gt;
&lt;th&gt;Purpose&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Electron&lt;/td&gt;
&lt;td&gt;Desktop application runtime&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Node.js&lt;/td&gt;
&lt;td&gt;Native SDK integration&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;TypeScript&lt;/td&gt;
&lt;td&gt;Type-safe application logic&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Presage SmartSpectra SDK&lt;/td&gt;
&lt;td&gt;Contactless physiological sensing&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Protocol Buffers&lt;/td&gt;
&lt;td&gt;Measurement decoding&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;HTML &amp;amp; CSS&lt;/td&gt;
&lt;td&gt;User interface and animations&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Canvas API&lt;/td&gt;
&lt;td&gt;Camera preview and live charts&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;esbuild&lt;/td&gt;
&lt;td&gt;Application bundling&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;h3&gt;
  
  
  Running PulseFlow Locally
&lt;/h3&gt;

&lt;p&gt;Clone the repository:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;git clone https://github.com/Ezequie1Sc/PulseFlow.git
&lt;span class="nb"&gt;cd&lt;/span&gt; PulseFlow
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Install dependencies:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;npm ci
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Create a local environment file.&lt;/p&gt;

&lt;p&gt;On Windows PowerShell:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight powershell"&gt;&lt;code&gt;&lt;span class="nc"&gt;Copy&lt;/span&gt;&lt;span class="o"&gt;-&lt;/span&gt;&lt;span class="nc"&gt;Item&lt;/span&gt; &lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;env&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;example&lt;/span&gt; &lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;env&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Configure your Presage SmartSpectra API key in &lt;code&gt;.env&lt;/code&gt;:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;SMARTSPECTRA_API_KEY=your_presage_api_key
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Start the desktop application:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;npm run dev
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;A compatible webcam, camera permissions, and a valid API key are required.&lt;/p&gt;

&lt;p&gt;The repository's README contains additional information about installation, application architecture, privacy, and troubleshooting.&lt;/p&gt;




&lt;h2&gt;
  
  
  How It Responds to Human Signals
&lt;/h2&gt;

&lt;p&gt;The &lt;strong&gt;Presage SmartSpectra SDK&lt;/strong&gt; is the foundation of PulseFlow.&lt;/p&gt;

&lt;p&gt;The application uses live camera-based physiological estimates rather than simulated datasets.&lt;/p&gt;

&lt;h3&gt;
  
  
  1. Camera-Based Signal Acquisition
&lt;/h3&gt;

&lt;p&gt;After the user explicitly consents to monitoring, PulseFlow initializes the SmartSpectra Node.js SDK and connects to a selected webcam.&lt;/p&gt;

&lt;p&gt;The application requests these metrics:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight typescript"&gt;&lt;code&gt;&lt;span class="n"&gt;requestedMetrics&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;[&lt;/span&gt;
  &lt;span class="nc"&gt;MetricType&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="no"&gt;PULSE_RATE&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
  &lt;span class="nc"&gt;MetricType&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="no"&gt;BREATHING_RATE&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
&lt;span class="p"&gt;]&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;SmartSpectra processes the camera input and generates physiological estimates along with signal-validation information.&lt;/p&gt;

&lt;h3&gt;
  
  
  2. Processing SDK Events
&lt;/h3&gt;

&lt;p&gt;PulseFlow listens for several SmartSpectra events:&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Event&lt;/th&gt;
&lt;th&gt;Purpose&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;&lt;code&gt;processingStatus&lt;/code&gt;&lt;/td&gt;
&lt;td&gt;Reports SDK processing states&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;code&gt;validationStatus&lt;/code&gt;&lt;/td&gt;
&lt;td&gt;Provides signal-quality feedback&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;code&gt;metrics&lt;/code&gt;&lt;/td&gt;
&lt;td&gt;Delivers physiological measurements&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;code&gt;videoOutput&lt;/code&gt;&lt;/td&gt;
&lt;td&gt;Supplies camera preview frames&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;code&gt;error&lt;/code&gt;&lt;/td&gt;
&lt;td&gt;Reports SDK errors&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;Incoming measurement messages are decoded using Protocol Buffers.&lt;/p&gt;

&lt;p&gt;The application processes the measurements before forwarding accepted values to the user interface.&lt;/p&gt;

&lt;h3&gt;
  
  
  3. Validating Physiological Measurements
&lt;/h3&gt;

&lt;p&gt;One of my priorities was avoiding misleading readings.&lt;/p&gt;

&lt;p&gt;PulseFlow checks incoming measurements using its validation and filtering logic.&lt;/p&gt;

&lt;p&gt;If SmartSpectra cannot provide a suitable signal, the interface communicates that state rather than presenting fabricated numbers.&lt;/p&gt;

&lt;p&gt;This makes the experience responsive not only to the physiological readings themselves, but also to whether valid measurements are available.&lt;/p&gt;

&lt;h3&gt;
  
  
  4. Updating the Interface in Real Time
&lt;/h3&gt;

&lt;p&gt;When an accepted measurement arrives:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;The current physiological value updates.&lt;/li&gt;
&lt;li&gt;The corresponding historical chart receives a new reading.&lt;/li&gt;
&lt;li&gt;An active breathing session collects the accepted sample.&lt;/li&gt;
&lt;li&gt;Collected samples contribute to the final session summary.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;The guided breathing animation operates independently of the measurement pipeline.&lt;/p&gt;

&lt;p&gt;This separation is intentional: an animated breathing guide should never be mistaken for an actual measured respiration signal.&lt;/p&gt;

&lt;h3&gt;
  
  
  5. Challenges During Development
&lt;/h3&gt;

&lt;p&gt;Integrating a native physiological sensing SDK introduced challenges beyond typical frontend development.&lt;/p&gt;

&lt;p&gt;I encountered issues involving camera capture conditions, frame rate, signal validation, and measurement availability.&lt;/p&gt;

&lt;p&gt;During debugging, SmartSpectra returned feedback including:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;&lt;code&gt;Place more of the chest in view.&lt;/code&gt;&lt;/li&gt;
&lt;li&gt;&lt;code&gt;Use a camera mode that provides at least 25 frames per second.&lt;/code&gt;&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;These messages taught me that successfully displaying a camera preview does not necessarily mean physiological sensing can produce usable measurements.&lt;/p&gt;

&lt;p&gt;I adjusted the camera framing and capture conditions and eventually obtained valid live readings.&lt;/p&gt;

&lt;p&gt;Another important challenge was integrating native capture operations into Electron while keeping the application's main process, renderer, and measurement-processing responsibilities separated.&lt;/p&gt;

&lt;p&gt;This project reinforced the importance of testing with actual hardware rather than relying exclusively on successful builds or automated tests.&lt;/p&gt;




&lt;h2&gt;
  
  
  Prize Categories
&lt;/h2&gt;

&lt;p&gt;I'm entering PulseFlow in both special prize categories.&lt;/p&gt;

&lt;h3&gt;
  
  
  Most Creative or Delightful Use
&lt;/h3&gt;

&lt;p&gt;What inspired PulseFlow was the idea of transforming physiological sensing into a simple, interactive desktop experience.&lt;/p&gt;

&lt;p&gt;Instead of displaying measurements on a conventional monitoring dashboard, I combined real-time sensing with guided breathing, restrained motion, live charts, and session summaries.&lt;/p&gt;

&lt;p&gt;The experience is deliberately minimal.&lt;/p&gt;

&lt;p&gt;I wanted to demonstrate that advanced sensing technology does not need a complicated interface to be engaging.&lt;/p&gt;

&lt;p&gt;The combination of real physiological measurements and a calm, interactive design is what makes PulseFlow distinctive.&lt;/p&gt;

&lt;h3&gt;
  
  
  Best Thoughtful Real-World Use
&lt;/h3&gt;

&lt;p&gt;PulseFlow explores a practical use of contactless sensing: allowing people to observe physiological estimates while participating in a voluntary breathing exercise.&lt;/p&gt;

&lt;p&gt;It is designed for users curious about general wellness and physiological sensing who may not own a wearable device.&lt;/p&gt;

&lt;p&gt;The experience emphasizes:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Using an ordinary webcam.&lt;/li&gt;
&lt;li&gt;Requiring explicit camera consent.&lt;/li&gt;
&lt;li&gt;Providing clear measurement feedback.&lt;/li&gt;
&lt;li&gt;Allowing users to stop monitoring.&lt;/li&gt;
&lt;li&gt;Explaining when readings are unavailable.&lt;/li&gt;
&lt;li&gt;Presenting understandable charts and session summaries.&lt;/li&gt;
&lt;li&gt;Avoiding unnecessary persistent physiological histories.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;PulseFlow is intended for educational and general wellness exploration.&lt;/p&gt;

&lt;p&gt;It is not intended to diagnose medical conditions, recommend treatments, or support clinical decision-making.&lt;/p&gt;




&lt;h2&gt;
  
  
  Consent and Data Handling
&lt;/h2&gt;

&lt;p&gt;Privacy and user control were important considerations during development.&lt;/p&gt;

&lt;p&gt;Because PulseFlow uses camera-based physiological sensing, I wanted the application to make its purpose clear and require intentional participation.&lt;/p&gt;

&lt;h3&gt;
  
  
  Camera Consent
&lt;/h3&gt;

&lt;p&gt;PulseFlow requires explicit user consent before beginning camera-based monitoring.&lt;/p&gt;

&lt;p&gt;The application presents information about camera access and allows users to control when monitoring starts and stops.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F672e9mj1gse191uiy8nk.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F672e9mj1gse191uiy8nk.png" alt="PulseFlow — Camera Consent" width="800" height="449"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;h3&gt;
  
  
  Camera and Physiological Data
&lt;/h3&gt;

&lt;p&gt;PulseFlow uses a webcam through SmartSpectra to obtain physiological estimates.&lt;/p&gt;

&lt;p&gt;Camera frames can be displayed as a preview in the desktop application.&lt;/p&gt;

&lt;p&gt;Accepted measurements are used to update live indicators, historical charts, and temporary session summaries.&lt;/p&gt;

&lt;h3&gt;
  
  
  Storage and Retention
&lt;/h3&gt;

&lt;p&gt;PulseFlow does not implement a dedicated persistent database for physiological measurements, user profiles, or session histories.&lt;/p&gt;

&lt;p&gt;The application is designed around temporary measurement data used during active monitoring and guided breathing sessions.&lt;/p&gt;

&lt;p&gt;This describes the application's own architecture, not the independent storage or retention behavior of Presage's SDK or services.&lt;/p&gt;

&lt;h3&gt;
  
  
  Data Sharing and SDK Telemetry
&lt;/h3&gt;

&lt;p&gt;PulseFlow configures SmartSpectra with:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight typescript"&gt;&lt;code&gt;&lt;span class="n"&gt;enableTelemetry&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="kc"&gt;false&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;This disables the configurable telemetry option used by the application.&lt;/p&gt;

&lt;p&gt;However, this setting should not be interpreted as proof that the SDK performs no network communication.&lt;/p&gt;

&lt;p&gt;SDK authentication, operational communication, and any provider-side processing are governed by Presage's implementation and documentation.&lt;/p&gt;

&lt;p&gt;PulseFlow does not implement an additional application backend for uploading physiological session histories.&lt;/p&gt;

&lt;h3&gt;
  
  
  Optional AI Services
&lt;/h3&gt;

&lt;p&gt;PulseFlow does not require a separate generative AI service for physiological measurement processing or guided breathing.&lt;/p&gt;

&lt;p&gt;Physiological estimates originate from the Presage SmartSpectra SDK.&lt;/p&gt;

&lt;h3&gt;
  
  
  Deletion and User Control
&lt;/h3&gt;

&lt;p&gt;Users can stop monitoring and manage or reset the active session through the application's controls.&lt;/p&gt;

&lt;p&gt;PulseFlow does not offer a permanent cloud-based physiological history or user account system.&lt;/p&gt;

&lt;p&gt;Temporary session data is intended for the active application experience.&lt;/p&gt;

&lt;h3&gt;
  
  
  Responsible Use
&lt;/h3&gt;

&lt;p&gt;PulseFlow is an experimental application for general wellness exploration and technical demonstration.&lt;/p&gt;

&lt;p&gt;It is not a medical device and must not be used for:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Medical diagnosis or treatment.&lt;/li&gt;
&lt;li&gt;Clinical decision-making.&lt;/li&gt;
&lt;li&gt;Emergency monitoring.&lt;/li&gt;
&lt;li&gt;Covert surveillance.&lt;/li&gt;
&lt;li&gt;Non-consensual physiological monitoring.&lt;/li&gt;
&lt;/ul&gt;




&lt;h2&gt;
  
  
  Session Results
&lt;/h2&gt;

&lt;p&gt;One of the features I enjoyed building was the measurement-based session summary.&lt;/p&gt;

&lt;p&gt;Instead of ending the breathing exercise without feedback, PulseFlow presents an overview of the measurements accepted during the session.&lt;/p&gt;

&lt;p&gt;In one completed 60-second demonstration, the application reported:&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Measurement&lt;/th&gt;
&lt;th&gt;Result&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Session duration&lt;/td&gt;
&lt;td&gt;60.0 seconds&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Average heart rate&lt;/td&gt;
&lt;td&gt;84.9 BPM&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Average breathing rate&lt;/td&gt;
&lt;td&gt;8.0 breaths/min&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Valid pulse samples&lt;/td&gt;
&lt;td&gt;69&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Valid breathing samples&lt;/td&gt;
&lt;td&gt;15&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Session status&lt;/td&gt;
&lt;td&gt;Completed&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;These results are observations from a particular demonstration session, not evidence of therapeutic or medical benefits.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fxfuq1lietxus5ggvxqoh.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fxfuq1lietxus5ggvxqoh.png" alt="PulseFlow — Session Summary" width="800" height="600"&gt;&lt;/a&gt;&lt;/p&gt;




&lt;h2&gt;
  
  
  What I Learned
&lt;/h2&gt;

&lt;p&gt;Building PulseFlow taught me several lessons about developing software that interacts with real hardware and physiological signals.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Native SDK integration is more than connecting an API.&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;Camera processing, SDK events, measurement decoding, signal validation, and visual presentation all need to work together.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Hardware conditions directly affect the experience.&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;Lighting, movement, camera positioning, and frame rate influence whether contactless sensing can produce usable measurements.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Missing data should remain missing.&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;An honest interface communicates when a measurement is unavailable rather than fabricating values to make the dashboard appear functional.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Thoughtful design can make complex technology approachable.&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;Clear typography, restrained animations, and descriptive feedback help users understand an unfamiliar sensing experience.&lt;/p&gt;




&lt;h2&gt;
  
  
  Final Thoughts
&lt;/h2&gt;

&lt;p&gt;PulseFlow began with a simple question: what could I build using an ordinary webcam and real-time human signals?&lt;/p&gt;

&lt;p&gt;It became an opportunity to explore native desktop development, physiological sensing, data validation, real-time visualization, and responsible interaction design.&lt;/p&gt;

&lt;p&gt;There is still room for improvement, particularly in camera compatibility, accessibility, and additional privacy controls.&lt;/p&gt;

&lt;p&gt;However, successfully receiving and displaying live SmartSpectra measurements was an important milestone for this project.&lt;/p&gt;

&lt;p&gt;Thank you to &lt;strong&gt;Presage&lt;/strong&gt; and &lt;strong&gt;DEV Community&lt;/strong&gt; for organizing the challenge and encouraging developers to explore new forms of human-computer interaction.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;PulseFlow — Your body, in real time.&lt;/strong&gt;&lt;/p&gt;




&lt;p&gt;&lt;strong&gt;GitHub:&lt;/strong&gt; &lt;a href="https://github.com/Ezequie1Sc/PulseFlow" rel="noopener noreferrer"&gt;https://github.com/Ezequie1Sc/PulseFlow&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Video Demonstration:&lt;/strong&gt; &lt;a href="https://youtu.be/8ztd4aoxduk" rel="noopener noreferrer"&gt;https://youtu.be/8ztd4aoxduk&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;SmartSpectra SDK:&lt;/strong&gt; &lt;a href="https://smartspectra.presagetech.com/docs/nodejs/" rel="noopener noreferrer"&gt;https://smartspectra.presagetech.com/docs/nodejs/&lt;/a&gt;&lt;/p&gt;

</description>
      <category>devchallenge</category>
      <category>presagechallenge</category>
    </item>
    <item>
      <title>Listify AI — Create Better Product Listings with Open-Weight AI</title>
      <dc:creator>Orlando Ezequiel Salazar Cruz</dc:creator>
      <pubDate>Mon, 19 Oct 2026 03:31:55 +0000</pubDate>
      <link>https://dev.to/ezequie1sc/listify-ai-create-better-product-listings-with-open-weight-ai-2n63</link>
      <guid>https://dev.to/ezequie1sc/listify-ai-create-better-product-listings-with-open-weight-ai-2n63</guid>
      <description>&lt;p&gt;&lt;em&gt;This is a submission for the &lt;a href="https://dev.to/challenges/hacktoberfest-week2-2026-10-12"&gt;Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle&lt;/a&gt;&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;What I Built&lt;/p&gt;

&lt;p&gt;Meet Listify AI, an AI-powered web application designed to help small businesses, entrepreneurs, and independent sellers create professional product listings in seconds.&lt;/p&gt;

&lt;p&gt;When selling products online, having a great product is only part of the challenge. Writing engaging descriptions, choosing the right words, creating hashtags, and adapting content for different platforms can take time.&lt;/p&gt;

&lt;p&gt;For small business owners who manage everything themselves, that time matters.&lt;/p&gt;

&lt;p&gt;Listify AI solves this problem by transforming simple product details into ready-to-use marketing content using open-weight artificial intelligence.&lt;/p&gt;

&lt;p&gt;Users provide information about their products, choose their preferred writing style, and let AI generate the content.&lt;/p&gt;

&lt;p&gt;What can Listify AI do?&lt;/p&gt;

&lt;p&gt;Generate product titles that highlight important features.&lt;/p&gt;

&lt;p&gt;Write compelling descriptions based on the information provided.&lt;/p&gt;

&lt;p&gt;Create relevant hashtags for social media.&lt;/p&gt;

&lt;p&gt;Generate calls to action to encourage customer engagement.&lt;/p&gt;

&lt;p&gt;Customize writing tones, including professional, friendly, persuasive, creative, and minimalist.&lt;/p&gt;

&lt;p&gt;Adapt content for different platforms, including Instagram, Facebook Marketplace, and online stores.&lt;/p&gt;

&lt;p&gt;Support English and Spanish for generated content.&lt;/p&gt;

&lt;p&gt;Copy generated content with a single click.&lt;/p&gt;

&lt;p&gt;The application features a clean, responsive interface designed to make AI-powered content creation accessible to everyone.&lt;/p&gt;

&lt;p&gt;As a potential side hustle, Listify AI could evolve into an affordable content creation service for small businesses and independent online sellers.&lt;/p&gt;

&lt;p&gt;The mission is simple:&lt;/p&gt;

&lt;p&gt;Less time writing. More time selling.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F1xthp5sydg9h933w94q8.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F1xthp5sydg9h933w94q8.png" alt=" " width="799" height="391"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Demo&lt;/p&gt;

&lt;p&gt;Listify AI is deployed and available to try directly in your browser.&lt;/p&gt;

&lt;p&gt;Live Demo: &lt;a href="https://listify-ai-mbrq.vercel.app/" rel="noopener noreferrer"&gt;https://listify-ai-mbrq.vercel.app/&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;No registration is required.&lt;/p&gt;

&lt;p&gt;How It Works&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;Describe your product&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;Enter the product name, category, price, and key features.&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;Customize your content&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;Select your preferred selling platform, writing tone, and output language.&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;Generate with AI&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;Click Generate Listing and let the AI transform your product details into marketing content.&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;Copy and share&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;Review the generated title, description, hashtags, and call to action. Copy the results and use them on your preferred platform.&lt;/p&gt;

&lt;p&gt;Example: Wireless Bluetooth Headphones&lt;/p&gt;

&lt;p&gt;For testing, I used a sample product with the following information:&lt;/p&gt;

&lt;p&gt;Product: Wireless Bluetooth Headphones&lt;/p&gt;

&lt;p&gt;Category: Electronics&lt;/p&gt;

&lt;p&gt;Features: Bluetooth 5.3, up to 30 hours of battery life, USB-C charging, built-in microphone, and comfortable cushioned ear cups.&lt;/p&gt;

&lt;p&gt;Price: $49.99 USD&lt;/p&gt;

&lt;p&gt;Platform: Instagram&lt;/p&gt;

&lt;p&gt;Tone: Persuasive&lt;/p&gt;

&lt;p&gt;Language: English&lt;/p&gt;

&lt;p&gt;Listify AI uses this information to generate a product title, description, hashtags, and a call to action tailored to the selected settings.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F49csf2ber9ihc6rnps55.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F49csf2ber9ihc6rnps55.png" alt=" " width="800" height="533"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;This demonstrates how an open-weight language model can turn a few product specifications into content that sellers can review and publish.&lt;/p&gt;

&lt;p&gt;Code&lt;/p&gt;

&lt;p&gt;Listify AI was developed using a lightweight web architecture with a JavaScript frontend and a serverless backend.&lt;/p&gt;

&lt;p&gt;GitHub Repository: &lt;a href="https://github.com/Ezequie1Sc/listify-ai" rel="noopener noreferrer"&gt;https://github.com/Ezequie1Sc/listify-ai&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Tech Stack&lt;/p&gt;

&lt;p&gt;Technology&lt;/p&gt;

&lt;p&gt;Purpose&lt;/p&gt;

&lt;p&gt;HTML5&lt;/p&gt;

&lt;p&gt;Application structure&lt;/p&gt;

&lt;p&gt;CSS3&lt;/p&gt;

&lt;p&gt;Responsive design and animations&lt;/p&gt;

&lt;p&gt;JavaScript&lt;/p&gt;

&lt;p&gt;Frontend functionality&lt;/p&gt;

&lt;p&gt;Node.js&lt;/p&gt;

&lt;p&gt;Serverless backend runtime&lt;/p&gt;

&lt;p&gt;Google Gemma 3 12B&lt;/p&gt;

&lt;p&gt;Open-weight AI model&lt;/p&gt;

&lt;p&gt;OpenRouter&lt;/p&gt;

&lt;p&gt;Model inference API&lt;/p&gt;

&lt;p&gt;Vercel&lt;/p&gt;

&lt;p&gt;Hosting and serverless deployment&lt;/p&gt;

&lt;p&gt;Git &amp;amp; GitHub&lt;/p&gt;

&lt;p&gt;Version control&lt;/p&gt;

&lt;p&gt;The main application components include:&lt;/p&gt;

&lt;p&gt;index.html — User interface and application structure.&lt;/p&gt;

&lt;p&gt;css/styles.css — Visual design, responsive layouts, and animations.&lt;/p&gt;

&lt;p&gt;js/app.js — User interactions, form validation, API requests, and generated content rendering.&lt;/p&gt;

&lt;p&gt;api/generate.js — Serverless endpoint that communicates with the AI model.&lt;/p&gt;

&lt;p&gt;The project keeps sensitive API credentials in server-side environment variables rather than exposing them in the browser.&lt;/p&gt;

&lt;p&gt;How I Built It&lt;/p&gt;

&lt;p&gt;I wanted Listify AI to be simple, fast, and easy to use.&lt;/p&gt;

&lt;p&gt;Instead of introducing a large frontend framework, I built the interface with HTML, CSS, and vanilla JavaScript.&lt;/p&gt;

&lt;p&gt;For the AI functionality, I integrated Google Gemma 3 12B, an open-weight language model accessed through the OpenRouter API.&lt;/p&gt;

&lt;p&gt;How the AI Works&lt;/p&gt;

&lt;p&gt;The application follows a straightforward generation process:&lt;/p&gt;

&lt;p&gt;Product Information → JavaScript Frontend → Vercel Serverless Function → OpenRouter → Google Gemma 3 → Generated Listing&lt;/p&gt;

&lt;p&gt;When users submit the form, the frontend collects their product information and sends it to the /api/generate endpoint.&lt;/p&gt;

&lt;p&gt;The backend validates the request and constructs a structured prompt containing the product name, features, category, price, platform, tone, and language.&lt;/p&gt;

&lt;p&gt;The AI model then generates a response containing four main elements:&lt;/p&gt;

&lt;p&gt;Product title&lt;/p&gt;

&lt;p&gt;Product description&lt;/p&gt;

&lt;p&gt;Relevant hashtags&lt;/p&gt;

&lt;p&gt;Call to action&lt;/p&gt;

&lt;p&gt;The backend processes the response and returns the generated content to the frontend, where users can review and copy it.&lt;/p&gt;

&lt;p&gt;Why AI Is Essential to the Project&lt;/p&gt;

&lt;p&gt;The open-weight model is not an optional addition to Listify AI.&lt;/p&gt;

&lt;p&gt;It is the core of the application.&lt;/p&gt;

&lt;p&gt;Without the language model, the application would only collect product information. Gemma 3 transforms that information into useful, customized marketing copy.&lt;/p&gt;

&lt;p&gt;The model is instructed to work with the information provided by the user and avoid inventing unsupported product specifications.&lt;/p&gt;

&lt;p&gt;Technical Challenges&lt;/p&gt;

&lt;p&gt;During development, I encountered several challenges involving AI integration and deployment.&lt;/p&gt;

&lt;p&gt;Some models had unavailable inference endpoints, while others encountered rate limits.&lt;/p&gt;

&lt;p&gt;I also faced a production configuration issue where the serverless backend could not detect the configured model because an environment variable had been named incorrectly.&lt;/p&gt;

&lt;p&gt;By examining Vercel function logs, I identified the problem, corrected the configuration, and successfully restored the AI generation functionality.&lt;/p&gt;

&lt;p&gt;These challenges helped me better understand API integration, serverless architecture, environment variables, structured responses, and production debugging.&lt;/p&gt;

&lt;p&gt;Why Does Open Innovation Matter?&lt;/p&gt;

&lt;p&gt;Open innovation matters to me because it makes powerful technology more accessible to students, independent developers, and small teams.&lt;/p&gt;

&lt;p&gt;When building Listify AI, I wanted to explore how an open-weight language model could power a practical application without relying entirely on a proprietary model.&lt;/p&gt;

&lt;p&gt;Freedom to Experiment&lt;/p&gt;

&lt;p&gt;Using Gemma 3 gives me the flexibility to explore different inference configurations and potentially replace the underlying model with another compatible open-weight model.&lt;/p&gt;

&lt;p&gt;This makes the architecture more adaptable as the project evolves.&lt;/p&gt;

&lt;p&gt;More Deployment Possibilities&lt;/p&gt;

&lt;p&gt;The current version of Listify AI uses OpenRouter for hosted inference.&lt;/p&gt;

&lt;p&gt;However, because Gemma 3 is an open-weight model, its weights can be used through other compatible inference solutions, subject to its license and hardware requirements.&lt;/p&gt;

&lt;p&gt;This creates opportunities to explore self-hosting and alternative deployment environments in the future.&lt;/p&gt;

&lt;p&gt;Lower Barriers to Entry&lt;/p&gt;

&lt;p&gt;As a student developer, being able to experiment with accessible AI models and development tools is extremely valuable.&lt;/p&gt;

&lt;p&gt;Open-weight AI makes it easier to learn, prototype, and build applications without needing a large development budget.&lt;/p&gt;

&lt;p&gt;Building Useful Tools&lt;/p&gt;

&lt;p&gt;What excites me most about open innovation is its potential to help solve everyday problems.&lt;/p&gt;

&lt;p&gt;Listify AI is a small example of how AI can support independent sellers who may not have the time or resources to create professional marketing content for every product.&lt;/p&gt;

&lt;p&gt;I believe the future of AI should include opportunities for developers of all backgrounds to experiment, contribute, and create useful applications.&lt;/p&gt;

&lt;p&gt;Open innovation turns powerful technology into opportunities for more people to build.&lt;/p&gt;

&lt;p&gt;What I Learned&lt;/p&gt;

&lt;p&gt;Building Listify AI helped me strengthen several technical skills:&lt;/p&gt;

&lt;p&gt;Integrating open-weight language models into web applications.&lt;/p&gt;

&lt;p&gt;Designing structured prompts for practical use cases.&lt;/p&gt;

&lt;p&gt;Developing and deploying serverless API endpoints.&lt;/p&gt;

&lt;p&gt;Handling AI service errors and rate limits.&lt;/p&gt;

&lt;p&gt;Protecting API credentials through environment variables.&lt;/p&gt;

&lt;p&gt;Building responsive interfaces with vanilla JavaScript and CSS.&lt;/p&gt;

&lt;p&gt;Debugging production deployments through serverless logs.&lt;/p&gt;

&lt;p&gt;Most importantly, I learned that creating a useful AI application requires more than choosing a model.&lt;/p&gt;

&lt;p&gt;It requires thinking about the user experience, reliability, security, and how the technology solves a real problem.&lt;/p&gt;

&lt;p&gt;What's Next?&lt;/p&gt;

&lt;p&gt;Listify AI is an early version, and I would like to continue improving it.&lt;/p&gt;

&lt;p&gt;Some features I plan to explore include:&lt;/p&gt;

&lt;p&gt;Multiple AI-generated listing variations.&lt;/p&gt;

&lt;p&gt;Additional languages and selling platforms.&lt;/p&gt;

&lt;p&gt;Product listing history and reusable templates.&lt;/p&gt;

&lt;p&gt;Support for product images.&lt;/p&gt;

&lt;p&gt;More open-weight AI model options.&lt;/p&gt;

&lt;p&gt;Improved rate limiting and API protection.&lt;/p&gt;

&lt;p&gt;Alternative inference providers and self-hosting experiments.&lt;/p&gt;

&lt;p&gt;These improvements could help transform Listify AI into a more complete tool for entrepreneurs and online sellers.&lt;/p&gt;

&lt;p&gt;Final Thoughts&lt;/p&gt;

&lt;p&gt;Listify AI started with a simple question:&lt;/p&gt;

&lt;p&gt;What if creating a professional product listing could be as easy as describing the product itself?&lt;/p&gt;

&lt;p&gt;This project allowed me to explore how open-weight artificial intelligence can turn that idea into a working application.&lt;/p&gt;

&lt;p&gt;It was a valuable experience combining frontend development, serverless architecture, AI integration, and deployment.&lt;/p&gt;

&lt;p&gt;I'm excited to continue learning and building with open technologies.&lt;/p&gt;

&lt;p&gt;Thanks to the DEV Community and Hacktoberfest for encouraging developers to experiment, share their projects, and explore the possibilities of open AI.&lt;/p&gt;

&lt;p&gt;Live Demo: &lt;a href="https://listify-ai-mbrq.vercel.app/" rel="noopener noreferrer"&gt;https://listify-ai-mbrq.vercel.app/&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Source Code: &lt;a href="https://github.com/Ezequie1Sc/listify-ai" rel="noopener noreferrer"&gt;https://github.com/Ezequie1Sc/listify-ai&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Thanks for reading!&lt;/p&gt;

&lt;h1&gt;
  
  
  hacktoberfest #opensource #ai #webdev
&lt;/h1&gt;

</description>
      <category>devchallenge</category>
      <category>hf26challenge</category>
    </item>
    <item>
      <title>Listify AI — Turning Simple Product Details into Better Listings with Open-Weight A</title>
      <dc:creator>Orlando Ezequiel Salazar Cruz</dc:creator>
      <pubDate>Mon, 19 Oct 2026 03:25:04 +0000</pubDate>
      <link>https://dev.to/ezequie1sc/listify-ai-turning-simple-product-details-into-better-listings-with-open-weight-a-47ad</link>
      <guid>https://dev.to/ezequie1sc/listify-ai-turning-simple-product-details-into-better-listings-with-open-weight-a-47ad</guid>
      <description>&lt;p&gt;&lt;em&gt;This is a submission for the &lt;a href="https://dev.to/challenges/hacktoberfest-week2-2026-10-12"&gt;Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle&lt;/a&gt;&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;What I Built&lt;/p&gt;

&lt;p&gt;Meet Listify AI, a web application that helps small businesses, entrepreneurs, and independent sellers create professional product listings using open-weight artificial intelligence.&lt;/p&gt;

&lt;p&gt;Selling products online involves more than uploading a picture and adding a price. A good listing needs an attractive title, a clear description, relevant hashtags, and a compelling call to action.&lt;/p&gt;

&lt;p&gt;However, creating this content repeatedly can take valuable time, especially for people managing their businesses on their own.&lt;/p&gt;

&lt;p&gt;That's where Listify AI comes in.&lt;/p&gt;

&lt;p&gt;Listify AI transforms basic product information into ready-to-use marketing content in seconds.&lt;/p&gt;

&lt;p&gt;Users simply enter their product details, choose a platform and writing style, and let AI handle the creative writing.&lt;/p&gt;

&lt;p&gt;Key Features&lt;/p&gt;

&lt;p&gt;AI-generated titles: Create engaging titles that highlight the product.&lt;/p&gt;

&lt;p&gt;Product descriptions: Turn simple specifications into clear, attractive descriptions.&lt;/p&gt;

&lt;p&gt;Relevant hashtags: Generate hashtags related to the product.&lt;/p&gt;

&lt;p&gt;Calls to action: Encourage potential customers to take the next step.&lt;/p&gt;

&lt;p&gt;Multiple writing tones: Choose between professional, friendly, persuasive, creative, and minimalist styles.&lt;/p&gt;

&lt;p&gt;Platform customization: Adapt listings for Facebook Marketplace, Instagram, online stores, or general use.&lt;/p&gt;

&lt;p&gt;Multilingual generation: Generate content in English or Spanish.&lt;/p&gt;

&lt;p&gt;Copy-to-clipboard: Easily reuse generated content on selling platforms.&lt;/p&gt;

&lt;p&gt;Responsive interface: Use the application on desktop or mobile devices.&lt;/p&gt;

&lt;p&gt;The idea behind Listify AI is simple: help sellers spend less time writing listings and more time growing their businesses.&lt;/p&gt;

&lt;p&gt;As a potential side hustle, the application could evolve into an affordable content creation service for online sellers who regularly publish new products.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F4k67pypqh6zjdozu9u1l.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F4k67pypqh6zjdozu9u1l.png" alt=" " width="799" height="391"&gt;&lt;/a&gt;&lt;br&gt;
Demo&lt;/p&gt;

&lt;p&gt;Listify AI is deployed on Vercel and available to try directly in your browser.&lt;/p&gt;

&lt;p&gt;Live Demo: &lt;a href="https://listify-ai-mbrq.vercel.app/" rel="noopener noreferrer"&gt;https://listify-ai-mbrq.vercel.app/&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;How It Works&lt;/p&gt;

&lt;p&gt;Step 1 — Enter your product information&lt;/p&gt;

&lt;p&gt;Provide the product name, category, price, and main features.&lt;/p&gt;

&lt;p&gt;Step 2 — Customize your listing&lt;/p&gt;

&lt;p&gt;Choose your preferred selling platform, writing tone, and language.&lt;/p&gt;

&lt;p&gt;Step 3 — Generate with AI&lt;/p&gt;

&lt;p&gt;Click the generation button and let the AI create a product listing based on your information.&lt;/p&gt;

&lt;p&gt;Step 4 — Copy and publish&lt;/p&gt;

&lt;p&gt;Review the generated title, description, hashtags, and call to action. Copy the content and use it on your preferred selling platform.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fol3b050uwttatvaniru7.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fol3b050uwttatvaniru7.png" alt=" " width="800" height="533"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;The goal was to make the experience intuitive, even for users who have never worked with AI tools before.&lt;/p&gt;

&lt;p&gt;Code&lt;/p&gt;

&lt;p&gt;Listify AI is built with a lightweight architecture that separates the frontend from the AI integration.&lt;/p&gt;

&lt;p&gt;GitHub Repository: &lt;a href="https://github.com/Ezequie1Sc/listify-ai" rel="noopener noreferrer"&gt;https://github.com/Ezequie1Sc/listify-ai&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Project Structure&lt;/p&gt;

&lt;p&gt;listify-ai/&lt;br&gt;
├── api/&lt;br&gt;
│   └── generate.js&lt;br&gt;
├── css/&lt;br&gt;
│   └── styles.css&lt;br&gt;
├── js/&lt;br&gt;
│   └── app.js&lt;br&gt;
├── index.html&lt;br&gt;
├── package.json&lt;br&gt;
├── vercel.json&lt;br&gt;
├── .env.example&lt;br&gt;
├── .gitignore&lt;br&gt;
└── README.md&lt;/p&gt;

&lt;p&gt;The application uses vanilla JavaScript for user interactions and a Vercel serverless function for communicating with the AI provider.&lt;/p&gt;

&lt;p&gt;API credentials are stored in server-side environment variables rather than exposed in the frontend.&lt;/p&gt;

&lt;p&gt;The repository contains the application code, styling, AI integration, and deployment configuration.&lt;/p&gt;

&lt;p&gt;How I Built It&lt;/p&gt;

&lt;p&gt;I wanted Listify AI to be fast, accessible, and easy to maintain.&lt;/p&gt;

&lt;p&gt;Instead of using a complex frontend framework, I chose HTML, CSS, and vanilla JavaScript.&lt;/p&gt;

&lt;p&gt;This allowed me to focus on the user experience and the AI-powered functionality.&lt;/p&gt;

&lt;p&gt;Technology Stack&lt;/p&gt;

&lt;p&gt;Technology&lt;/p&gt;

&lt;p&gt;Purpose&lt;/p&gt;

&lt;p&gt;HTML5&lt;/p&gt;

&lt;p&gt;Application structure&lt;/p&gt;

&lt;p&gt;CSS3&lt;/p&gt;

&lt;p&gt;Responsive design and animations&lt;/p&gt;

&lt;p&gt;JavaScript&lt;/p&gt;

&lt;p&gt;Frontend logic and API integration&lt;/p&gt;

&lt;p&gt;Node.js&lt;/p&gt;

&lt;p&gt;Serverless function runtime&lt;/p&gt;

&lt;p&gt;Google Gemma 3 12B&lt;/p&gt;

&lt;p&gt;Open-weight language model&lt;/p&gt;

&lt;p&gt;OpenRouter&lt;/p&gt;

&lt;p&gt;AI inference provider&lt;/p&gt;

&lt;p&gt;Vercel&lt;/p&gt;

&lt;p&gt;Hosting and serverless deployment&lt;/p&gt;

&lt;p&gt;Git &amp;amp; GitHub&lt;/p&gt;

&lt;p&gt;Version control and source code&lt;/p&gt;

&lt;p&gt;The AI Model&lt;/p&gt;

&lt;p&gt;For the AI functionality, I integrated Google Gemma 3 12B, an open-weight language model accessed through OpenRouter.&lt;/p&gt;

&lt;p&gt;The model is responsible for transforming structured product information into useful marketing content.&lt;/p&gt;

&lt;p&gt;Instead of returning generic text, the application builds a prompt using the user's product details, selected platform, preferred writing tone, and language.&lt;/p&gt;

&lt;p&gt;The AI is instructed to generate a structured response containing:&lt;/p&gt;

&lt;p&gt;Product title&lt;/p&gt;

&lt;p&gt;Product description&lt;/p&gt;

&lt;p&gt;Relevant hashtags&lt;/p&gt;

&lt;p&gt;Call to action&lt;/p&gt;

&lt;p&gt;This is the core functionality of Listify AI. Without the language model, the application would not be able to transform raw product details into complete marketing copy.&lt;/p&gt;

&lt;p&gt;Application Architecture&lt;/p&gt;

&lt;p&gt;The generation process follows this workflow:&lt;/p&gt;

&lt;p&gt;User enters product details&lt;br&gt;
            |&lt;br&gt;
            v&lt;br&gt;
     JavaScript Frontend&lt;br&gt;
            |&lt;br&gt;
            v&lt;br&gt;
      /api/generate&lt;br&gt;
     Vercel Serverless&lt;br&gt;
            |&lt;br&gt;
            v&lt;br&gt;
       OpenRouter API&lt;br&gt;
            |&lt;br&gt;
            v&lt;br&gt;
      Google Gemma 3&lt;br&gt;
            |&lt;br&gt;
            v&lt;br&gt;
    Structured AI Response&lt;br&gt;
            |&lt;br&gt;
            v&lt;br&gt;
      Generated Listing&lt;/p&gt;

&lt;p&gt;When a user submits the form, JavaScript collects and validates the input before sending a request to the backend.&lt;/p&gt;

&lt;p&gt;The serverless function processes the request, builds the AI prompt, and communicates with OpenRouter.&lt;/p&gt;

&lt;p&gt;Once the model returns a response, the backend validates the generated content and sends it back to the frontend.&lt;/p&gt;

&lt;p&gt;The application then displays the result in an organized interface.&lt;/p&gt;

&lt;p&gt;Building the User Interface&lt;/p&gt;

&lt;p&gt;I wanted the application to feel modern, minimal, and approachable.&lt;/p&gt;

&lt;p&gt;The interface uses clean typography, balanced spacing, subtle animations, and responsive layouts.&lt;/p&gt;

&lt;p&gt;I also focused on making the generation process easy to understand through clear input fields, loading indicators, error messages, and organized results.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fhx385gb4i7ewfblxgsyu.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fhx385gb4i7ewfblxgsyu.png" alt=" " width="800" height="381"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fz6z12fz29yf6w6qrpdiy.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fz6z12fz29yf6w6qrpdiy.png" alt=" " width="799" height="348"&gt;&lt;/a&gt;&lt;br&gt;
Challenges I Faced&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fltlj8ahkzqrghq0hmvdx.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fltlj8ahkzqrghq0hmvdx.png" alt=" " width="800" height="392"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;One of the most interesting parts of building Listify AI was connecting the deployed application to the AI service.&lt;/p&gt;

&lt;p&gt;During development, I encountered several issues:&lt;/p&gt;

&lt;p&gt;AI models without available inference endpoints.&lt;/p&gt;

&lt;p&gt;API rate limits when testing free models.&lt;/p&gt;

&lt;p&gt;Serverless configuration errors.&lt;/p&gt;

&lt;p&gt;Environment variables not being detected correctly.&lt;/p&gt;

&lt;p&gt;Handling unsuccessful requests and invalid AI responses.&lt;/p&gt;

&lt;p&gt;One particularly useful debugging experience involved a deployment error caused by an incorrectly named environment variable.&lt;/p&gt;

&lt;p&gt;The frontend was working, but the backend could not find the model configuration.&lt;/p&gt;

&lt;p&gt;By reviewing the Vercel function logs, I identified the issue and corrected the environment variable.&lt;/p&gt;

&lt;p&gt;This experience reminded me that developing an AI-powered application is not just about writing prompts. It also requires reliable backend integration, debugging, deployment, and error handling.&lt;/p&gt;

&lt;p&gt;Why Does Open Innovation Matter?&lt;/p&gt;

&lt;p&gt;I believe open innovation is important because it makes advanced technology more accessible to independent developers, students, and small teams.&lt;/p&gt;

&lt;p&gt;For Listify AI, using an open-weight model was a deliberate architectural choice.&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;Freedom to Experiment&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;Open-weight models give developers opportunities to experiment with different models and inference configurations.&lt;/p&gt;

&lt;p&gt;Rather than permanently depending on a single proprietary language model, Listify AI can be adapted to use other compatible models.&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;More Deployment Options&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;Because Gemma 3 provides openly available model weights under its license, developers can explore alternative inference environments.&lt;/p&gt;

&lt;p&gt;Although the current version of Listify AI uses OpenRouter for hosted inference, the underlying model could potentially be deployed through a different compatible provider or a self-hosted setup.&lt;/p&gt;

&lt;p&gt;This flexibility is valuable for projects that may grow beyond their initial prototype.&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;Lower Barriers to Building&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;As a student developer, access to AI models and affordable development tools makes a meaningful difference.&lt;/p&gt;

&lt;p&gt;Open-weight models and free inference options create opportunities to experiment with AI applications without requiring a large initial budget.&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;Learning and Transparency&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;Working with an open-weight model encouraged me to learn more about model selection, prompt design, inference APIs, structured outputs, and deployment.&lt;/p&gt;

&lt;p&gt;It also made me think about the importance of building applications where the AI component can be understood, evaluated, and replaced.&lt;/p&gt;

&lt;p&gt;Open innovation matters because it gives more people the opportunity to turn their ideas into working applications.&lt;/p&gt;

&lt;p&gt;For a project like Listify AI, that means exploring how accessible AI can help small sellers solve everyday problems.&lt;/p&gt;

&lt;p&gt;What I Learned&lt;/p&gt;

&lt;p&gt;Building Listify AI taught me valuable lessons about creating practical AI-powered web applications.&lt;/p&gt;

&lt;p&gt;Some of my biggest takeaways were:&lt;/p&gt;

&lt;p&gt;AI integration: How to connect an open-weight language model to a real application.&lt;/p&gt;

&lt;p&gt;Prompt engineering: How to structure requests to generate consistent, useful results.&lt;/p&gt;

&lt;p&gt;Serverless architecture: How to build and deploy backend functionality without managing a traditional server.&lt;/p&gt;

&lt;p&gt;Security: Why API keys must remain on the server.&lt;/p&gt;

&lt;p&gt;Error handling: How to respond to model unavailability, rate limits, and failed requests.&lt;/p&gt;

&lt;p&gt;UI/UX design: How to make AI functionality accessible through a simple interface.&lt;/p&gt;

&lt;p&gt;Deployment: How to debug production issues using Vercel logs and environment variables.&lt;/p&gt;

&lt;p&gt;This project also reinforced something I enjoy about software development: taking a simple idea, experimenting with different approaches, solving unexpected problems, and gradually turning it into something functional.&lt;/p&gt;

&lt;p&gt;What's Next?&lt;/p&gt;

&lt;p&gt;Listify AI is still an early version, and I would like to continue improving it.&lt;/p&gt;

&lt;p&gt;Some features I would like to explore include:&lt;/p&gt;

&lt;p&gt;Generating multiple listing variations.&lt;/p&gt;

&lt;p&gt;Adding more languages and selling platforms.&lt;/p&gt;

&lt;p&gt;Supporting additional open-weight AI models.&lt;/p&gt;

&lt;p&gt;Saving previously generated listings.&lt;/p&gt;

&lt;p&gt;Creating reusable product listing templates.&lt;/p&gt;

&lt;p&gt;Generating content from uploaded product images.&lt;/p&gt;

&lt;p&gt;Adding stronger API rate limiting.&lt;/p&gt;

&lt;p&gt;Exploring alternative inference providers and self-hosted models.&lt;/p&gt;

&lt;p&gt;These improvements could help Listify AI become a more complete content creation tool for independent sellers.&lt;/p&gt;

&lt;p&gt;Final Thoughts&lt;/p&gt;

&lt;p&gt;Listify AI started with a simple idea:&lt;/p&gt;

&lt;p&gt;What if creating a professional product listing could be as easy as describing the product itself?&lt;/p&gt;

&lt;p&gt;Through this project, I explored how open-weight artificial intelligence can help solve a practical problem for small businesses and online sellers.&lt;/p&gt;

&lt;p&gt;I'm excited to keep learning, improving the application, and experimenting with open AI technologies.&lt;/p&gt;

&lt;p&gt;Thanks to the DEV Community and Hacktoberfest for creating opportunities to build, share, and learn.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fveqvu76l58tqo8y4o1no.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fveqvu76l58tqo8y4o1no.png" alt=" " width="800" height="533"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Try Listify AI: &lt;a href="https://listify-ai-mbrq.vercel.app/" rel="noopener noreferrer"&gt;https://listify-ai-mbrq.vercel.app/&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Explore the source code: &lt;a href="https://github.com/Ezequie1Sc/listify-ai" rel="noopener noreferrer"&gt;https://github.com/Ezequie1Sc/listify-ai&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Thanks for reading!&lt;/p&gt;

&lt;h1&gt;
  
  
  hacktoberfest #opensource #ai #webdev*This is a submission for the &lt;a href="https://dev.to/challenges/hacktoberfest-week2-2026-10-12"&gt;Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle&lt;/a&gt;*
&lt;/h1&gt;

</description>
      <category>devchallenge</category>
      <category>hf26challenge</category>
    </item>
    <item>
      <title>Benchmarking Free AI Models for Outdoor Adventure Generation</title>
      <dc:creator>Orlando Ezequiel Salazar Cruz</dc:creator>
      <pubDate>Sat, 10 Oct 2026 06:54:33 +0000</pubDate>
      <link>https://dev.to/ezequie1sc/benchmarking-free-ai-models-for-outdoor-adventure-generation-3pik</link>
      <guid>https://dev.to/ezequie1sc/benchmarking-free-ai-models-for-outdoor-adventure-generation-3pik</guid>
      <description>&lt;p&gt;&lt;em&gt;This is a submission for the &lt;a href="https://dev.to/challenges/kaggle-2026-09-23"&gt;Kaggle Benchmarking Challenge&lt;/a&gt;&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  What I Benchmarked
&lt;/h2&gt;

&lt;p&gt;For &lt;strong&gt;Ruta Viva AI&lt;/strong&gt;, an outdoor exploration app, I wanted to evaluate how reliably free AI models can generate structured outdoor adventures.&lt;/p&gt;

&lt;p&gt;The benchmark measures more than whether a model responds successfully. Each adventure must contain valid JSON, all required fields, exactly three missions, and consistent XP totals.&lt;/p&gt;

&lt;p&gt;I designed 30 scenarios across 10 environments, with three prompt variations for each environment. For this initial pilot, I tested three scenarios per model, resulting in nine requests.&lt;/p&gt;

&lt;h2&gt;
  
  
  Models Tested
&lt;/h2&gt;

&lt;p&gt;I evaluated three free models through OpenRouter:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Liquid LFM 2.5 2.6B:&lt;/strong&gt; &lt;code&gt;liquid/lfm-2.5-2.6b:free&lt;/code&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;NVIDIA Nemotron 3.5 Lightning:&lt;/strong&gt; &lt;code&gt;nvidia/nemotron-3.5-lightning:free&lt;/code&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Apodex 1.1 Mini:&lt;/strong&gt; &lt;code&gt;apodex/apodex-1.1-mini:free&lt;/code&gt;
&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;I chose these models to compare structured-output reliability and response speed using a common set of prompts.&lt;/p&gt;

&lt;h2&gt;
  
  
  Findings
&lt;/h2&gt;

&lt;p&gt;The pilot revealed significant differences in structured-output reliability.&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Model&lt;/th&gt;
&lt;th&gt;Valid JSON&lt;/th&gt;
&lt;th&gt;Median latency&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Liquid LFM 2.5 2.6B&lt;/td&gt;
&lt;td&gt;3/3 (100%)&lt;/td&gt;
&lt;td&gt;13.57 s&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;NVIDIA Nemotron 3.5 Lightning&lt;/td&gt;
&lt;td&gt;1/3 (33.3%)&lt;/td&gt;
&lt;td&gt;41.37 s&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Apodex 1.1 Mini&lt;/td&gt;
&lt;td&gt;0/3 (0%)&lt;/td&gt;
&lt;td&gt;6.36 s&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;&lt;strong&gt;Liquid was the most reliable model in this pilot.&lt;/strong&gt; All three responses contained valid JSON, the required fields, exactly three missions, and consistent XP totals.&lt;/p&gt;

&lt;p&gt;Nemotron produced one valid response; two were truncated at the output limit. Apodex was the fastest by median latency, but none of its responses produced valid JSON, and one returned no usable answer content.&lt;/p&gt;

&lt;p&gt;My main takeaway is that &lt;strong&gt;speed alone does not determine usefulness&lt;/strong&gt;. When an application depends on structured data, a fast response that cannot be parsed may be less valuable than a slower, correctly formatted response.&lt;/p&gt;

&lt;p&gt;This benchmark has limitations: three requests per model are not enough to establish general performance. The safety check is a basic keyword heuristic, not proof of safety, and adventure quality still requires human evaluation.&lt;/p&gt;

&lt;p&gt;Next, I would test more scenarios, repeat requests to measure consistency, and manually evaluate relevance, accessibility, variety, practicality, and safety.&lt;/p&gt;

&lt;h2&gt;
  
  
  My Benchmark
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;Public Kaggle notebook:&lt;/strong&gt; &lt;a href="https://www.kaggle.com/code/ezequielsalazar1/outdoor-adventure-ai-benchmark" rel="noopener noreferrer"&gt;https://www.kaggle.com/code/ezequielsalazar1/outdoor-adventure-ai-benchmark&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;GitHub repository:&lt;/strong&gt; &lt;a href="https://github.com/Ezequie1Sc/outdoor-adventure-ai-benchmark" rel="noopener noreferrer"&gt;https://github.com/Ezequie1Sc/outdoor-adventure-ai-benchmark&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;The notebook documents the benchmark methodology and validation logic. The results above come from the pilot I ran locally.&lt;/p&gt;

&lt;h1&gt;
  
  
  kagglechallenge
&lt;/h1&gt;

</description>
      <category>devchallenge</category>
      <category>kagglechallenge</category>
      <category>ai</category>
      <category>machinelearning</category>
    </item>
    <item>
      <title>🌿 Ruta Viva AI: Turning Everyday Walks into AI-Powered Adventures</title>
      <dc:creator>Orlando Ezequiel Salazar Cruz</dc:creator>
      <pubDate>Fri, 09 Oct 2026 22:43:26 +0000</pubDate>
      <link>https://dev.to/ezequie1sc/ruta-viva-ai-turning-everyday-walks-into-ai-powered-adventures-1apa</link>
      <guid>https://dev.to/ezequie1sc/ruta-viva-ai-turning-everyday-walks-into-ai-powered-adventures-1apa</guid>
      <description>&lt;p&gt;What if AI could help us spend less time staring at screens and more time exploring the world?&lt;/p&gt;

&lt;p&gt;That's the idea behind Ruta Viva AI, an AI-powered outdoor adventure platform that transforms ordinary walks into interactive missions. Instead of spending more time scrolling, users can discover their surroundings, complete real-world challenges, and earn XP while exploring nature and local culture.&lt;/p&gt;

&lt;p&gt;The goal is simple: make the screen the shortest part of the adventure.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F75tovh37q4cmparxflor.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F75tovh37q4cmparxflor.png" alt=" " width="800" height="376"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;What I Built&lt;/p&gt;

&lt;p&gt;Ruta Viva AI turns going outside into a game.&lt;/p&gt;

&lt;p&gt;Users choose their preferred adventure type, difficulty, and duration. The application then selects an adventure containing missions designed to encourage real-world exploration.&lt;/p&gt;

&lt;p&gt;Adventure categories include:&lt;/p&gt;

&lt;p&gt;🌳 Nature: Observe plants, discover patterns, identify colors, and listen to the sounds around you.&lt;/p&gt;

&lt;p&gt;🏛️ Culture: Explore architecture, public art, and interesting details in your community.&lt;/p&gt;

&lt;p&gt;🚶 Walking: Turn an ordinary walk into a sequence of small challenges.&lt;/p&gt;

&lt;p&gt;🎲 Surprise: Let the adventure system choose an unexpected exploration experience.&lt;/p&gt;

&lt;p&gt;Each adventure includes missions, descriptions, and experience points (XP). The idea is to give people a reason to step outside, pay attention to their surroundings, and turn everyday activities into memorable experiences.&lt;/p&gt;

&lt;p&gt;The project is designed for curious people who want to explore their communities, enjoy the outdoors, and make their walks more engaging.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fjynmqtb5ey8d08q1hlua.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fjynmqtb5ey8d08q1hlua.png" alt=" " width="800" height="391"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F2e3i119ktbc2op5o6d2p.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F2e3i119ktbc2op5o6d2p.png" alt=" " width="800" height="382"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Demo&lt;/p&gt;

&lt;p&gt;Try the deployed application:&lt;/p&gt;

&lt;p&gt;Live application: &lt;a href="https://ruta-viva-ai-hbdn.vercel.app/" rel="noopener noreferrer"&gt;https://ruta-viva-ai-hbdn.vercel.app/&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Backend API: &lt;a href="https://ruta-viva-ai-backend.onrender.com/" rel="noopener noreferrer"&gt;https://ruta-viva-ai-backend.onrender.com/&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Interactive API documentation: &lt;a href="https://ruta-viva-ai-backend.onrender.com/docs" rel="noopener noreferrer"&gt;https://ruta-viva-ai-backend.onrender.com/docs&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;The frontend is deployed on Vercel, and the FastAPI backend is deployed on Render.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fbd28jsqod55vkcaynz1d.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fbd28jsqod55vkcaynz1d.png" alt=" " width="800" height="662"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Adventure Results&lt;/p&gt;

&lt;p&gt;After each adventure, Ruta Viva AI displays the distance traveled, time spent exploring, and XP earned, giving users a clear overview of their progress and achievements.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fl3ncz0u5ikpnt5yx3cuf.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fl3ncz0u5ikpnt5yx3cuf.png" alt=" " width="800" height="783"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Badges and Rewards&lt;/p&gt;

&lt;p&gt;Based on their adventure achievements, users can earn badges to celebrate their progress and build their personal collection. Each badge encourages them to keep exploring, complete new missions, and reach new milestones.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fyoa8uv1q3e0wx1sbuhs7.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fyoa8uv1q3e0wx1sbuhs7.png" alt=" " width="800" height="511"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Code&lt;/p&gt;

&lt;p&gt;GitHub repository: &lt;a href="https://github.com/Ezequie1Sc/ruta-viva-ai" rel="noopener noreferrer"&gt;https://github.com/Ezequie1Sc/ruta-viva-ai&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;The project is organized into a frontend and a backend:&lt;/p&gt;

&lt;p&gt;Frontend: Angular and TypeScript.&lt;/p&gt;

&lt;p&gt;Backend: Python and FastAPI.&lt;/p&gt;

&lt;p&gt;Adventure catalog: A structured JSON file containing predefined adventures and missions.&lt;/p&gt;

&lt;p&gt;AI integration: OpenRouter with a configurable model.&lt;/p&gt;

&lt;p&gt;Deployment: Vercel and Render.&lt;/p&gt;

&lt;p&gt;The code is structured so that the adventure catalog, selection logic, and AI provider integration can evolve independently.&lt;/p&gt;

&lt;p&gt;How I Built It&lt;/p&gt;

&lt;p&gt;I built Ruta Viva AI using Angular for the frontend and FastAPI for the backend.&lt;/p&gt;

&lt;p&gt;Initially, the AI was responsible for generating an entire adventure, including its title, description, missions, and XP rewards. However, waiting several minutes for a generation was not a good experience for an application designed to encourage people to go outside.&lt;/p&gt;

&lt;p&gt;I rethought the architecture and introduced a hybrid approach.&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;A structured adventure catalog&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;Instead of generating every mission from scratch, the application maintains a JSON catalog of predefined adventures.&lt;/p&gt;

&lt;p&gt;Each entry includes an identifier, adventure type, difficulty, title, description, missions, and XP rewards.&lt;/p&gt;

&lt;p&gt;This gives the application a consistent source of content and makes it easier to validate adventure data.&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;AI as the decision-making layer&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;The AI receives the user's preferences and a filtered list of candidate adventures. Its job is to select the most appropriate option instead of writing every mission from scratch.&lt;/p&gt;

&lt;p&gt;Python filters the catalog by adventure type and difficulty before sending relevant candidates to the model. This reduces unnecessary context and the amount of generation required.&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;A Python fallback&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;An external AI provider can be slow, unavailable, or rate-limited. To make the experience more resilient, the backend can select an adventure locally when the AI call fails or times out.&lt;/p&gt;

&lt;p&gt;This means the application can still return a valid adventure even when the external model cannot provide a selection.&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;Open-source AI and Ollama&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;One of my goals is to keep the AI layer flexible rather than tying the project to a single proprietary model or provider.&lt;/p&gt;

&lt;p&gt;The current deployed implementation uses OpenRouter for hosted inference and supports a configurable model identifier. This makes it easier to experiment with different compatible models.&lt;/p&gt;

&lt;p&gt;I am also interested in Ollama, which provides a practical way to run supported open-weight models locally. A local Ollama integration is a natural next step for Ruta Viva AI, allowing the project to explore local inference without requiring every AI request to go through a hosted provider.&lt;/p&gt;

&lt;p&gt;Local inference is not currently part of the deployed version, but the separation between the catalog, selection logic, and AI service provides a foundation for adding it.&lt;/p&gt;

&lt;p&gt;Why Does Open Innovation Matter?&lt;/p&gt;

&lt;p&gt;Open innovation matters because AI-powered applications should be adaptable, inspectable, and open to experimentation.&lt;/p&gt;

&lt;p&gt;For Ruta Viva AI, separating the AI provider from the adventure system makes it easier to change models, compare their behavior, and explore alternative deployment strategies.&lt;/p&gt;

&lt;p&gt;Using a structured catalog also means that the application does not depend entirely on a model to create every piece of content. The missions can be reviewed, improved, and expanded independently of the AI provider.&lt;/p&gt;

&lt;p&gt;Exploring Ollama and open-weight models could also make local inference possible in a future version. That would provide more control over where inference runs and could reduce dependence on an external service, although hardware requirements and model licenses would still matter.&lt;/p&gt;

&lt;p&gt;There is an important distinction in the current implementation: the application uses hosted inference through OpenRouter, so it requires an internet connection for AI selection. The local Python fallback can still select from the catalog if the provider fails, but that does not mean the entire application currently runs offline.&lt;/p&gt;

&lt;p&gt;For me, open innovation is about preserving the freedom to experiment and improve the system rather than building the entire experience around a single AI provider.&lt;/p&gt;

&lt;p&gt;My Agent Session&lt;/p&gt;

&lt;p&gt;I used AI-assisted development to explore the architecture, improve the generation workflow, and implement the JSON catalog and fallback selection.&lt;/p&gt;

&lt;p&gt;Prize Categories&lt;/p&gt;

&lt;p&gt;This project is submitted to the Hacktoberfest Open-Source AI Challenge — Week 1: Touch Grass.&lt;/p&gt;

&lt;p&gt;What's Next?&lt;/p&gt;

&lt;p&gt;There is still plenty I want to improve:&lt;/p&gt;

&lt;p&gt;Integrate an optional local inference provider using Ollama.&lt;/p&gt;

&lt;p&gt;Improve response times and measure AI selection latency.&lt;/p&gt;

&lt;p&gt;Expand the adventure catalog with more varied missions.&lt;/p&gt;

&lt;p&gt;Improve duration-aware selection so adventures better match the time users have available.&lt;/p&gt;

&lt;p&gt;Develop GPS-based progress tracking and outdoor discovery features.&lt;/p&gt;

&lt;p&gt;Continue improving the mobile experience.&lt;/p&gt;

&lt;p&gt;Final Thoughts&lt;/p&gt;

&lt;p&gt;Building Ruta Viva AI reminded me that AI does not always need to create more screen time. It can also help people step away from their devices and engage with the physical world.&lt;/p&gt;

&lt;p&gt;Sometimes, the best technology is the kind that gives you a reason to put your phone away.&lt;/p&gt;

&lt;p&gt;Build with AI. Explore the real world. Touch grass. 🌿&lt;/p&gt;

&lt;p&gt;Thanks for checking out Ruta Viva AI!&lt;/p&gt;

</description>
      <category>devchallenge</category>
      <category>hf26challenge</category>
      <category>hacktoberfest</category>
      <category>programming</category>
    </item>
    <item>
      <title>Collectify: AI-Powered Inventory Management for Smarter Business Decisions</title>
      <dc:creator>Orlando Ezequiel Salazar Cruz</dc:creator>
      <pubDate>Mon, 05 Oct 2026 05:01:35 +0000</pubDate>
      <link>https://dev.to/ezequie1sc/collectify-ai-powered-inventory-management-for-smarter-business-decisions-4j6n</link>
      <guid>https://dev.to/ezequie1sc/collectify-ai-powered-inventory-management-for-smarter-business-decisions-4j6n</guid>
      <description>&lt;p&gt;What if AI wasn't a chatbot added to a business application, but part of the business itself?&lt;/p&gt;

&lt;p&gt;For Hacktoberfest 2026, I decided to build something around a problem that is actually happening in my community.&lt;/p&gt;

&lt;p&gt;I built Collectify, an AI-first inventory and business management platform designed for a real local store that sells anime and pop-culture collectibles such as Funko, Pokémon products, plushies, pins, posters, and other merchandise.&lt;/p&gt;

&lt;p&gt;This is not a hypothetical business scenario.&lt;br&gt;
The project is being developed around the needs of a real local store with multiple partners, where keeping track of products, ownership, sales, inventory, earnings, and purchasing decisions can become difficult when information is handled manually or spread across different places.&lt;br&gt;
And that is where the idea for Collectify started.&lt;br&gt;
The Real Problem&lt;br&gt;
A traditional inventory system can tell you:&lt;br&gt;
"You have 5 units of this product."&lt;/p&gt;

&lt;p&gt;But a real business needs to ask much more:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Who owns those products?&lt;/li&gt;
&lt;li&gt;Which partner's products are selling?&lt;/li&gt;
&lt;li&gt;Which products are performing better?&lt;/li&gt;
&lt;li&gt;What should be restocked?&lt;/li&gt;
&lt;li&gt;What products are becoming slow-moving?&lt;/li&gt;
&lt;li&gt;How much is each partner generating?&lt;/li&gt;
&lt;li&gt;Where could we find suppliers?&lt;/li&gt;
&lt;li&gt;What should we consider purchasing next?&lt;/li&gt;
&lt;li&gt;What does the business data actually tell us?
For a store with multiple partners, inventory is not just about counting products.
It is also about understanding the business behind the inventory.
That is the problem Collectify is designed to address.
Built for a Real Business
Collectify is being developed around a real local store, not an imaginary customer created specifically for the hackathon.
The store has multiple partners who contribute and sell products through the same physical location.
That creates a very practical challenge:
How do you maintain control over everyone's products and sales while still having a unified view of the business?
Collectify addresses this by connecting:
Products → Partners → Sales → Earnings → Inventory → AI
The objective is to give the people running the store a centralized place where they can understand what is happening and make better decisions.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fnozt3smsifeok12l3anw.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fnozt3smsifeok12l3anw.jpg" alt=" " width="800" height="1067"&gt;&lt;/a&gt;&lt;br&gt;
&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F5txx6ay99qok8cy8dk4x.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F5txx6ay99qok8cy8dk4x.jpg" alt=" " width="800" height="1067"&gt;&lt;/a&gt;&lt;br&gt;
The Idea Behind Collectify&lt;br&gt;
I wanted to avoid a very common pattern in AI applications:&lt;br&gt;
Inventory System&lt;br&gt;
       +&lt;br&gt;
Generic Chatbot&lt;br&gt;
       =&lt;br&gt;
"AI-powered" application&lt;/p&gt;

&lt;p&gt;That was not the direction I wanted to take.&lt;br&gt;
Instead, Collectify follows this model:&lt;br&gt;
Business Data&lt;br&gt;
      ↓&lt;br&gt;
      AI&lt;br&gt;
      ↓&lt;br&gt;
Understanding&lt;br&gt;
      ↓&lt;br&gt;
Recommendations&lt;br&gt;
      ↓&lt;br&gt;
Business Action&lt;/p&gt;

&lt;p&gt;The AI is not there simply to answer questions.&lt;br&gt;
It is designed to interact with the information generated by the business and turn that information into something useful.&lt;br&gt;
Why AI Is Actually Useful Here&lt;br&gt;
Collectify uses AI in workflows that have a direct relationship with the business.&lt;br&gt;
For example:&lt;br&gt;
Inventory Data&lt;br&gt;
      ↓&lt;br&gt;
AI Analysis&lt;br&gt;
      ↓&lt;br&gt;
Restocking Insights&lt;/p&gt;

&lt;p&gt;Sales Data&lt;br&gt;
      ↓&lt;br&gt;
AI Analysis&lt;br&gt;
      ↓&lt;br&gt;
Product Performance&lt;/p&gt;

&lt;p&gt;Partner Data&lt;br&gt;
      ↓&lt;br&gt;
AI Analysis&lt;br&gt;
      ↓&lt;br&gt;
Individual Business Insights&lt;/p&gt;

&lt;p&gt;Product Search&lt;br&gt;
      ↓&lt;br&gt;
AI Supplier Discovery&lt;br&gt;
      ↓&lt;br&gt;
Potential Suppliers&lt;br&gt;
      ↓&lt;br&gt;
Purchasing Decision&lt;/p&gt;

&lt;p&gt;The important part is that the AI is connected to the application's actual domain.&lt;br&gt;
It is not simply sitting beside the application as a chat window.&lt;br&gt;
AI Business Analysis&lt;br&gt;
Collectify includes an AI analysis workflow where users can ask questions about their business.&lt;br&gt;
For example:&lt;br&gt;
"Which products are performing better and which ones should I consider restocking?"&lt;/p&gt;

&lt;p&gt;The goal is to transform operational information into business insight.&lt;br&gt;
Instead of manually going through multiple records, the user can interact with the business data through the AI layer.&lt;br&gt;
The long-term idea is to make Collectify capable of continuously helping the business understand:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Product performance&lt;/li&gt;
&lt;li&gt;Sales behavior&lt;/li&gt;
&lt;li&gt;Inventory conditions&lt;/li&gt;
&lt;li&gt;Partner performance&lt;/li&gt;
&lt;li&gt;Potential restocking needs&lt;/li&gt;
&lt;li&gt;Business opportunities&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F6pdna2y5s7tu2kusnp4e.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F6pdna2y5s7tu2kusnp4e.png" alt=" " width="800" height="432"&gt;&lt;/a&gt;&lt;br&gt;
&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fui8kus5i3ekn0igsv013.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fui8kus5i3ekn0igsv013.png" alt=" " width="800" height="417"&gt;&lt;/a&gt;&lt;br&gt;
&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fz0eukj3r9e07lqemog50.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fz0eukj3r9e07lqemog50.png" alt=" " width="800" height="400"&gt;&lt;/a&gt;&lt;br&gt;
&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F1wtgsu5tm68rdtnkzvyd.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F1wtgsu5tm68rdtnkzvyd.png" alt=" " width="800" height="391"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;AI-Powered Supplier Discovery&lt;br&gt;
One of the most important AI workflows in Collectify is supplier discovery.&lt;br&gt;
A user can search for something like:&lt;br&gt;
"Funko Pokémon"&lt;/p&gt;

&lt;p&gt;Collectify sends the request through the backend AI workflow and returns structured supplier information.&lt;br&gt;
Instead of returning only a paragraph of text, the system can produce application-ready information such as:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Supplier name&lt;/li&gt;
&lt;li&gt;Description&lt;/li&gt;
&lt;li&gt;Category&lt;/li&gt;
&lt;li&gt;Location&lt;/li&gt;
&lt;li&gt;Website&lt;/li&gt;
&lt;li&gt;Relevance&lt;/li&gt;
&lt;li&gt;Rating when available&lt;/li&gt;
&lt;li&gt;Verification information when available&lt;/li&gt;
&lt;li&gt;Supplier tags
The frontend then consumes those results and renders them as supplier cards.
This distinction is important.
The AI is not only generating language.
It is generating structured information that the application can use.
User
↓
Supplier Search
↓
FastAPI
↓
AI / Supplier Discovery
↓
Structured Results
↓
Angular
↓
Supplier Cards&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Flw4i5h4ub5shbuusptsd.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Flw4i5h4ub5shbuusptsd.png" alt=" " width="799" height="382"&gt;&lt;/a&gt;&lt;br&gt;
&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fjp200ojlp1ae6skvdq9x.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fjp200ojlp1ae6skvdq9x.png" alt=" " width="800" height="380"&gt;&lt;/a&gt;&lt;br&gt;
&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Foefdu4vqnh63av2wula2.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Foefdu4vqnh63av2wula2.png" alt=" " width="800" height="371"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;AI Is a Business Tool, Not a Decoration&lt;br&gt;
This is probably the most important design decision behind Collectify.&lt;br&gt;
I did not want to be able to say:&lt;br&gt;
"My inventory application has AI."&lt;/p&gt;

&lt;p&gt;I wanted to be able to say:&lt;br&gt;
"The AI changes what the application can do."&lt;/p&gt;

&lt;p&gt;Without the AI layer, Collectify can store business information.&lt;br&gt;
With the AI layer, the system can begin to interpret that information and help transform it into decisions.&lt;br&gt;
That creates a much more interesting relationship:&lt;br&gt;
Store&lt;br&gt;
 ↓&lt;br&gt;
Data&lt;br&gt;
 ↓&lt;br&gt;
Understanding&lt;br&gt;
 ↓&lt;br&gt;
Decision&lt;br&gt;
 ↓&lt;br&gt;
Action&lt;/p&gt;

&lt;p&gt;This is the direction I want to continue developing.&lt;br&gt;
Multi-Partner Management&lt;br&gt;
Another important part of Collectify is the multi-partner model.&lt;br&gt;
The store is shared by multiple partners, which means the system needs to understand that not every product belongs to the same person.&lt;br&gt;
Collectify allows the business to organize information around individual partners and their products.&lt;br&gt;
This helps provide visibility into:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Product ownership&lt;/li&gt;
&lt;li&gt;Individual inventory&lt;/li&gt;
&lt;li&gt;Sales associated with products&lt;/li&gt;
&lt;li&gt;Partner information&lt;/li&gt;
&lt;li&gt;Earnings&lt;/li&gt;
&lt;li&gt;Business performance
Instead of having one large inventory with no context, Collectify creates a more structured representation of the business.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fcidjdc72y6keexyyzy4m.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fcidjdc72y6keexyyzy4m.png" alt=" " width="800" height="381"&gt;&lt;/a&gt;&lt;br&gt;
&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fqvmgibnwdgbfewnz9due.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fqvmgibnwdgbfewnz9due.png" alt=" " width="799" height="377"&gt;&lt;/a&gt;&lt;br&gt;
&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Frlg1u7dax4z0hnwfk3on.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Frlg1u7dax4z0hnwfk3on.png" alt=" " width="800" height="428"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Inventory Management&lt;br&gt;
Collectify provides the operational foundation required by the AI layer.&lt;br&gt;
Products can contain information such as:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Name&lt;/li&gt;
&lt;li&gt;Description&lt;/li&gt;
&lt;li&gt;Category&lt;/li&gt;
&lt;li&gt;SKU&lt;/li&gt;
&lt;li&gt;Cost&lt;/li&gt;
&lt;li&gt;Sale price&lt;/li&gt;
&lt;li&gt;Stock&lt;/li&gt;
&lt;li&gt;Status&lt;/li&gt;
&lt;li&gt;Ownership
The inventory is not intended to be the final product.
It is the data foundation that allows the rest of the platform to work.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fm6zs505osys3hpvjvf41.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fm6zs505osys3hpvjvf41.png" alt=" " width="800" height="379"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Sales and Earnings&lt;br&gt;
Sales are connected to products and business information so that the platform can later analyze what is happening.&lt;br&gt;
The earnings section provides another layer of visibility into the operation.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fl14o11v5657u5ql647wf.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fl14o11v5657u5ql647wf.png" alt=" " width="800" height="386"&gt;&lt;/a&gt;&lt;br&gt;
&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fvui5trpyqgq3vbiqbes2.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fvui5trpyqgq3vbiqbes2.png" alt=" " width="800" height="386"&gt;&lt;/a&gt;&lt;br&gt;
&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Ft8rljn9zneecvx1k1ruh.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Ft8rljn9zneecvx1k1ruh.png" alt=" " width="800" height="407"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fo5tbalbrzxpsrmddrs48.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fo5tbalbrzxpsrmddrs48.png" alt=" " width="799" height="384"&gt;&lt;/a&gt;&lt;br&gt;
&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F66i7io60g0adbldq9sh9.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F66i7io60g0adbldq9sh9.png" alt=" " width="800" height="386"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;These sections are important because they provide the data that future AI workflows can use for deeper business analysis.&lt;br&gt;
Open AI and Local Development&lt;br&gt;
One of the requirements of this Hacktoberfest challenge is to build something with open-source AI at its core.&lt;br&gt;
Collectify was designed around an AI layer that can work with open/open-weight models and different providers, rather than making the entire application permanently dependent on a single proprietary model.&lt;br&gt;
The application was also developed and tested locally, including both the Angular frontend and the FastAPI backend.&lt;br&gt;
This is important to me because I want the AI layer to remain replaceable.&lt;br&gt;
Collectify&lt;br&gt;
     ↓&lt;br&gt;
AI Service&lt;br&gt;
     ↓&lt;br&gt;
Model / Provider&lt;/p&gt;

&lt;p&gt;The application should be able to evolve as models evolve.&lt;br&gt;
That means being able to:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Experiment with different models.&lt;/li&gt;
&lt;li&gt;Replace providers.&lt;/li&gt;
&lt;li&gt;Test different AI capabilities.&lt;/li&gt;
&lt;li&gt;Adapt the AI layer without redesigning the entire application.&lt;/li&gt;
&lt;li&gt;Run the core application locally.
An AI-first application should not necessarily mean being locked into one model forever.
Evidence of the AI Workflow
The screenshots below document the actual development and functionality of the AI features.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F1a8jbcgd2m7a4wjyxh9r.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F1a8jbcgd2m7a4wjyxh9r.png" alt=" " width="800" height="407"&gt;&lt;/a&gt;&lt;br&gt;
&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F8k3vplbwruk28q22ffbe.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F8k3vplbwruk28q22ffbe.png" alt=" " width="800" height="400"&gt;&lt;/a&gt;&lt;br&gt;
&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fwch6pyucfbvkfxjrh4da.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fwch6pyucfbvkfxjrh4da.png" alt=" " width="799" height="407"&gt;&lt;/a&gt;&lt;br&gt;
&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fra64j58qswgtku90n8au.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fra64j58qswgtku90n8au.png" alt=" " width="799" height="407"&gt;&lt;/a&gt;&lt;br&gt;
Technology Stack&lt;br&gt;
Frontend&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Angular&lt;/li&gt;
&lt;li&gt;TypeScript&lt;/li&gt;
&lt;li&gt;SCSS&lt;/li&gt;
&lt;li&gt;Angular HttpClient&lt;/li&gt;
&lt;li&gt;Responsive UI
Backend&lt;/li&gt;
&lt;li&gt;Python&lt;/li&gt;
&lt;li&gt;FastAPI&lt;/li&gt;
&lt;li&gt;Pydantic&lt;/li&gt;
&lt;li&gt;Uvicorn&lt;/li&gt;
&lt;li&gt;REST API
Database&lt;/li&gt;
&lt;li&gt;Supabase&lt;/li&gt;
&lt;li&gt;PostgreSQL
AI&lt;/li&gt;
&lt;li&gt;OpenRouter&lt;/li&gt;
&lt;li&gt;Open/open-weight AI models&lt;/li&gt;
&lt;li&gt;AI-powered business analysis&lt;/li&gt;
&lt;li&gt;Structured AI supplier discovery
Deployment&lt;/li&gt;
&lt;li&gt;Vercel&lt;/li&gt;
&lt;li&gt;Render&lt;/li&gt;
&lt;li&gt;Supabase
Architecture
              ┌─────────────────────────┐
              │       Collectify        │
              │     Angular Frontend    │
              └────────────┬────────────┘
                           │
                        HTTP/REST
                           │
                           ▼
              ┌─────────────────────────┐
              │     FastAPI Backend     │
              │                         │
              │ Products                │
              │ Sales                   │
              │ Partners                │
              │ Earnings                │
              │ AI Analysis             │
              │ Supplier Discovery      │
              └────────────┬────────────┘
                           │
                ┌──────────┴──────────┐
                │                     │
                ▼                     ▼
      ┌──────────────────┐   ┌──────────────────┐
      │    Supabase      │   │    OpenRouter    │
      │    PostgreSQL    │   │    AI Models     │
      └──────────────────┘   └──────────────────┘&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Running Locally&lt;br&gt;
Clone the repository:&lt;br&gt;
git clone &lt;a href="https://github.com/Ezequie1Sc/collectify.git" rel="noopener noreferrer"&gt;https://github.com/Ezequie1Sc/collectify.git&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Install the frontend dependencies:&lt;br&gt;
cd frontend/collectify&lt;br&gt;
pnpm install&lt;/p&gt;

&lt;p&gt;Run the frontend:&lt;br&gt;
pnpm start&lt;/p&gt;

&lt;p&gt;The Angular application will be available at:&lt;br&gt;
&lt;a href="http://localhost:4200" rel="noopener noreferrer"&gt;http://localhost:4200&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;For the backend:&lt;br&gt;
uvicorn app.main:app --reload&lt;/p&gt;

&lt;p&gt;The API will run at:&lt;br&gt;
&lt;a href="http://127.0.0.1:8000" rel="noopener noreferrer"&gt;http://127.0.0.1:8000&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Health check:&lt;br&gt;
&lt;a href="http://127.0.0.1:8000/health" rel="noopener noreferrer"&gt;http://127.0.0.1:8000/health&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;The project was developed and tested locally before deployment.&lt;br&gt;
Live Project&lt;br&gt;
Frontend&lt;br&gt;
&lt;a href="https://collectify-7xcu.vercel.app" rel="noopener noreferrer"&gt;https://collectify-7xcu.vercel.app&lt;/a&gt;&lt;br&gt;
Backend API&lt;br&gt;
&lt;a href="https://collectify-api-udxk.onrender.com" rel="noopener noreferrer"&gt;https://collectify-api-udxk.onrender.com&lt;/a&gt;&lt;br&gt;
API Documentation&lt;br&gt;
&lt;a href="https://collectify-api-udxk.onrender.com/docs" rel="noopener noreferrer"&gt;https://collectify-api-udxk.onrender.com/docs&lt;/a&gt;&lt;br&gt;
GitHub Repository&lt;br&gt;
&lt;a href="https://github.com/Ezequie1Sc/collectify" rel="noopener noreferrer"&gt;https://github.com/Ezequie1Sc/collectify&lt;/a&gt;&lt;br&gt;
Why This Project Matters&lt;br&gt;
Small businesses generate valuable information every day.&lt;br&gt;
But that information is often distributed across:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Spreadsheets&lt;/li&gt;
&lt;li&gt;Messages&lt;/li&gt;
&lt;li&gt;Notebooks&lt;/li&gt;
&lt;li&gt;Manual records&lt;/li&gt;
&lt;li&gt;Different applications&lt;/li&gt;
&lt;li&gt;Conversations between partners
The problem is not necessarily the lack of data.
The problem is the lack of understanding.
Collectify attempts to create a bridge between operational data and business decisions.
Inventory
↓
Sales
↓
Business Data
↓
AI
↓
Insights
↓
Recommendations
↓
Action&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;That is the direction I believe AI can bring real value to small businesses.&lt;br&gt;
What Makes Collectify Different?&lt;br&gt;
Collectify is not trying to compete with large enterprise ERP systems.&lt;br&gt;
The goal is different.&lt;br&gt;
It is about taking a real small-business problem and asking:&lt;br&gt;
What could happen if the business's own data became something the application could understand?&lt;/p&gt;

&lt;p&gt;The project combines:&lt;/p&gt;

&lt;h2&gt;
  
  
  Inventory management
&lt;/h2&gt;

&lt;h2&gt;
  
  
  Multi-partner management
&lt;/h2&gt;

&lt;h2&gt;
  
  
  Sales and earnings
&lt;/h2&gt;

&lt;h2&gt;
  
  
  AI analysis
&lt;/h2&gt;

&lt;h1&gt;
  
  
  AI-powered supplier discovery
&lt;/h1&gt;

&lt;p&gt;An AI-first business platform&lt;br&gt;
And the most important part is that this architecture was shaped by an actual business problem.&lt;br&gt;
What's Next?&lt;br&gt;
Collectify is still evolving.&lt;br&gt;
Some of the features I want to explore next include:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Sales forecasting&lt;/li&gt;
&lt;li&gt;Automatic restock recommendations&lt;/li&gt;
&lt;li&gt;Partner-specific analytics&lt;/li&gt;
&lt;li&gt;Supplier comparison&lt;/li&gt;
&lt;li&gt;Purchase recommendations&lt;/li&gt;
&lt;li&gt;Profitability analysis&lt;/li&gt;
&lt;li&gt;Automated business insights&lt;/li&gt;
&lt;li&gt;Role-based access control&lt;/li&gt;
&lt;li&gt;Advanced dashboards&lt;/li&gt;
&lt;li&gt;More open-weight AI models&lt;/li&gt;
&lt;li&gt;More autonomous AI workflows
The long-term goal is to move from:
Inventory Management&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;towards:&lt;br&gt;
AI-Powered Business Intelligence&lt;br&gt;
for Small Retail Businesses&lt;/p&gt;

&lt;p&gt;Final Thoughts&lt;br&gt;
This project changed the way I think about AI applications.&lt;br&gt;
AI does not have to be a chat window placed next to an existing application.&lt;br&gt;
It can become part of the application's actual workflow.&lt;br&gt;
With Collectify, the goal is:&lt;br&gt;
Organize Information&lt;br&gt;
        ↓&lt;br&gt;
Understand the Business&lt;br&gt;
        ↓&lt;br&gt;
Discover Opportunities&lt;br&gt;
        ↓&lt;br&gt;
Make Better Decisions&lt;/p&gt;

&lt;p&gt;And most importantly, this is not a hypothetical problem.&lt;br&gt;
Collectify was built around a real local business, with real operational needs and a real multi-partner inventory problem.&lt;br&gt;
That made the project much more meaningful to build.&lt;br&gt;
I did not start with:&lt;br&gt;
"What AI feature can I add to my application?"&lt;/p&gt;

&lt;p&gt;I started with:&lt;br&gt;
"What problem does this business actually have, and where can AI genuinely help?"&lt;/p&gt;

&lt;p&gt;That question shaped the entire project.&lt;br&gt;
AI should not be decoration. It should solve a real problem.&lt;/p&gt;

&lt;p&gt;Built for Hacktoberfest 2026&lt;br&gt;
Hacktoberfest Weekend Challenge — Build for a Friend&lt;br&gt;
Collectify — AI-powered business management for real-world retail.&lt;br&gt;
GitHub:&lt;br&gt;
&lt;a href="https://github.com/Ezequie1Sc/collectify" rel="noopener noreferrer"&gt;https://github.com/Ezequie1Sc/collectify&lt;/a&gt;&lt;br&gt;
Live Demo:&lt;br&gt;
&lt;a href="https://collectify-7xcu.vercel.app" rel="noopener noreferrer"&gt;https://collectify-7xcu.vercel.app&lt;/a&gt;&lt;br&gt;
API Documentation:&lt;br&gt;
&lt;a href="https://collectify-api-udxk.onrender.com/docs" rel="noopener noreferrer"&gt;https://collectify-api-udxk.onrender.com/docs&lt;/a&gt;&lt;/p&gt;

</description>
      <category>devchallenge</category>
      <category>weekendchallenge</category>
      <category>hf26challenge</category>
    </item>
    <item>
      <title>Collectify: Building an AI-First Business Platform for a Real Local Store</title>
      <dc:creator>Orlando Ezequiel Salazar Cruz</dc:creator>
      <pubDate>Mon, 05 Oct 2026 04:32:05 +0000</pubDate>
      <link>https://dev.to/ezequie1sc/collectify-building-an-ai-first-business-platform-for-a-real-local-store-2c1e</link>
      <guid>https://dev.to/ezequie1sc/collectify-building-an-ai-first-business-platform-for-a-real-local-store-2c1e</guid>
      <description>&lt;p&gt;What if AI wasn't a chatbot added to a business application, but part of the business itself?&lt;/p&gt;

&lt;p&gt;For Hacktoberfest 2026, I decided to build something around a problem that is actually happening in my community.&lt;br&gt;
I built Collectify, an AI-first inventory and business management platform designed for a real local store that sells anime and pop-culture collectibles such as Funko, Pokémon products, plushies, pins, posters, and other merchandise.&lt;/p&gt;

&lt;p&gt;This is not a hypothetical business scenario.&lt;br&gt;
The project is being developed around the needs of a real local store with multiple partners, where keeping track of products, ownership, sales, inventory, earnings, and purchasing decisions can become difficult when information is handled manually or spread across different places.&lt;br&gt;
And that is where the idea for Collectify started.&lt;br&gt;
The Real Problem&lt;br&gt;
A traditional inventory system can tell you:&lt;br&gt;
"You have 5 units of this product."&lt;/p&gt;

&lt;p&gt;But a real business needs to ask much more:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Who owns those products?&lt;/li&gt;
&lt;li&gt;Which partner's products are selling?&lt;/li&gt;
&lt;li&gt;Which products are performing better?&lt;/li&gt;
&lt;li&gt;What should be restocked?&lt;/li&gt;
&lt;li&gt;What products are becoming slow-moving?&lt;/li&gt;
&lt;li&gt;How much is each partner generating?&lt;/li&gt;
&lt;li&gt;Where could we find suppliers?&lt;/li&gt;
&lt;li&gt;What should we consider purchasing next?&lt;/li&gt;
&lt;li&gt;What does the business data actually tell us?&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;For a store with multiple partners, inventory is not just about counting products.&lt;br&gt;
It is also about understanding the business behind the inventory.&lt;br&gt;
That is the problem Collectify is designed to address.&lt;br&gt;
Built for a Real Business&lt;br&gt;
Collectify is being developed around a real local store, not an imaginary customer created specifically for the hackathon.&lt;br&gt;
The store has multiple partners who contribute and sell products through the same physical location.&lt;br&gt;
That creates a very practical challenge:&lt;br&gt;
How do you maintain control over everyone's products and sales while still having a unified view of the business?&lt;br&gt;
Collectify addresses this by connecting:&lt;br&gt;
Products → Partners → Sales → Earnings → Inventory → AI&lt;br&gt;
The objective is to give the people running the store a centralized place where they can understand what is happening and make better decisions.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fbj7gf3sjc7te238wrdw6.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fbj7gf3sjc7te238wrdw6.jpg" alt=" " width="800" height="1067"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F4yyt26wndv4dp57d6ddv.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F4yyt26wndv4dp57d6ddv.jpg" alt=" " width="800" height="1067"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;The Idea Behind Collectify&lt;br&gt;
I wanted to avoid a very common pattern in AI applications:&lt;br&gt;
Inventory System&lt;br&gt;
       +&lt;br&gt;
Generic Chatbot&lt;br&gt;
       =&lt;br&gt;
"AI-powered" application&lt;/p&gt;

&lt;p&gt;That was not the direction I wanted to take.&lt;br&gt;
Instead, Collectify follows this model:&lt;br&gt;
Business Data&lt;br&gt;
      ↓&lt;br&gt;
      AI&lt;br&gt;
      ↓&lt;br&gt;
Understanding&lt;br&gt;
      ↓&lt;br&gt;
Recommendations&lt;br&gt;
      ↓&lt;br&gt;
Business Action&lt;/p&gt;

&lt;p&gt;The AI is not there simply to answer questions.&lt;br&gt;
It is designed to interact with the information generated by the business and turn that information into something useful.&lt;br&gt;
Why AI Is Actually Useful Here&lt;br&gt;
Collectify uses AI in workflows that have a direct relationship with the business.&lt;br&gt;
For example:&lt;br&gt;
Inventory Data&lt;br&gt;
      ↓&lt;br&gt;
AI Analysis&lt;br&gt;
      ↓&lt;br&gt;
Restocking Insights&lt;/p&gt;

&lt;p&gt;Sales Data&lt;br&gt;
      ↓&lt;br&gt;
AI Analysis&lt;br&gt;
      ↓&lt;br&gt;
Product Performance&lt;/p&gt;

&lt;p&gt;Partner Data&lt;br&gt;
      ↓&lt;br&gt;
AI Analysis&lt;br&gt;
      ↓&lt;br&gt;
Individual Business Insights&lt;/p&gt;

&lt;p&gt;Product Search&lt;br&gt;
      ↓&lt;br&gt;
AI Supplier Discovery&lt;br&gt;
      ↓&lt;br&gt;
Potential Suppliers&lt;br&gt;
      ↓&lt;br&gt;
Purchasing Decision&lt;/p&gt;

&lt;p&gt;The important part is that the AI is connected to the application's actual domain.&lt;br&gt;
It is not simply sitting beside the application as a chat window.&lt;br&gt;
AI Business Analysis&lt;br&gt;
Collectify includes an AI analysis workflow where users can ask questions about their business.&lt;br&gt;
For example:&lt;br&gt;
"Which products are performing better and which ones should I consider restocking?"&lt;/p&gt;

&lt;p&gt;The goal is to transform operational information into business insight.&lt;br&gt;
Instead of manually going through multiple records, the user can interact with the business data through the AI layer.&lt;br&gt;
The long-term idea is to make Collectify capable of continuously helping the business understand:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Product performance&lt;/li&gt;
&lt;li&gt;Sales behavior&lt;/li&gt;
&lt;li&gt;Inventory conditions&lt;/li&gt;
&lt;li&gt;Partner performance&lt;/li&gt;
&lt;li&gt;Potential restocking needs&lt;/li&gt;
&lt;li&gt;Business opportunities&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fc67sugkdvf01ejp9g7ak.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fc67sugkdvf01ejp9g7ak.png" alt=" " width="800" height="407"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F2qr45jowhpyni3x87ksl.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F2qr45jowhpyni3x87ksl.png" alt=" " width="800" height="400"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fwexwxp6vmwgu0rfa9hyu.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fwexwxp6vmwgu0rfa9hyu.png" alt=" " width="799" height="407"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;AI-Powered Supplier Discovery&lt;br&gt;
One of the most important AI workflows in Collectify is supplier discovery.&lt;br&gt;
A user can search for something like:&lt;br&gt;
"Funko Pokémon"&lt;/p&gt;

&lt;p&gt;Collectify sends the request through the backend AI workflow and returns structured supplier information.&lt;br&gt;
Instead of returning only a paragraph of text, the system can produce application-ready information such as:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Supplier name&lt;/li&gt;
&lt;li&gt;Description&lt;/li&gt;
&lt;li&gt;Category&lt;/li&gt;
&lt;li&gt;Location&lt;/li&gt;
&lt;li&gt;Website&lt;/li&gt;
&lt;li&gt;Relevance&lt;/li&gt;
&lt;li&gt;Rating when available&lt;/li&gt;
&lt;li&gt;Verification information when available&lt;/li&gt;
&lt;li&gt;Supplier tags
The frontend then consumes those results and renders them as supplier cards.
This distinction is important.
The AI is not only generating language.
It is generating structured information that the application can use.
User
↓
Supplier Search
↓
FastAPI
↓
AI / Supplier Discovery
↓
Structured Results
↓
Angular
↓
Supplier Cards&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fkc2z3xpeze1dtsc2c4nj.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fkc2z3xpeze1dtsc2c4nj.png" alt=" " width="799" height="382"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fxrrutx5wbowyc5yv2ek8.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fxrrutx5wbowyc5yv2ek8.png" alt=" " width="800" height="380"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fad9j6ozmjllfg070uhu5.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fad9j6ozmjllfg070uhu5.png" alt=" " width="800" height="371"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;AI Is a Business Tool, Not a Decoration&lt;br&gt;
This is probably the most important design decision behind Collectify.&lt;br&gt;
I did not want to be able to say:&lt;br&gt;
"My inventory application has AI."&lt;/p&gt;

&lt;p&gt;I wanted to be able to say:&lt;br&gt;
"The AI changes what the application can do."&lt;/p&gt;

&lt;p&gt;Without the AI layer, Collectify can store business information.&lt;br&gt;
With the AI layer, the system can begin to interpret that information and help transform it into decisions.&lt;br&gt;
That creates a much more interesting relationship:&lt;br&gt;
Store&lt;br&gt;
 ↓&lt;br&gt;
Data&lt;br&gt;
 ↓&lt;br&gt;
Understanding&lt;br&gt;
 ↓&lt;br&gt;
Decision&lt;br&gt;
 ↓&lt;br&gt;
Action&lt;/p&gt;

&lt;p&gt;This is the direction I want to continue developing.&lt;br&gt;
Multi-Partner Management&lt;br&gt;
Another important part of Collectify is the multi-partner model.&lt;br&gt;
The store is shared by multiple partners, which means the system needs to understand that not every product belongs to the same person.&lt;br&gt;
Collectify allows the business to organize information around individual partners and their products.&lt;br&gt;
This helps provide visibility into:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Product ownership&lt;/li&gt;
&lt;li&gt;Individual inventory&lt;/li&gt;
&lt;li&gt;Sales associated with products&lt;/li&gt;
&lt;li&gt;Partner information&lt;/li&gt;
&lt;li&gt;Earnings&lt;/li&gt;
&lt;li&gt;Business performance
Instead of having one large inventory with no context, Collectify creates a more structured representation of the business.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fzv2q2t4t0xxgbvhlpzov.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fzv2q2t4t0xxgbvhlpzov.png" alt=" " width="799" height="382"&gt;&lt;/a&gt;&lt;br&gt;
&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fz3kn86v4awvx5oz90rxg.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fz3kn86v4awvx5oz90rxg.png" alt=" " width="800" height="380"&gt;&lt;/a&gt;&lt;br&gt;
&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F9zqjqaqf1b6ky0iurepj.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F9zqjqaqf1b6ky0iurepj.png" alt=" " width="800" height="371"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Inventory Management&lt;br&gt;
Collectify provides the operational foundation required by the AI layer.&lt;br&gt;
Products can contain information such as:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Name&lt;/li&gt;
&lt;li&gt;Description&lt;/li&gt;
&lt;li&gt;Category&lt;/li&gt;
&lt;li&gt;SKU&lt;/li&gt;
&lt;li&gt;Cost&lt;/li&gt;
&lt;li&gt;Sale price&lt;/li&gt;
&lt;li&gt;Stock&lt;/li&gt;
&lt;li&gt;Status&lt;/li&gt;
&lt;li&gt;Ownership
The inventory is not intended to be the final product.
It is the data foundation that allows the rest of the platform to work.
[ESPAÑOL — AQUÍ AGREGA LA IMAGEN DEL DASHBOARD]
Coloca aquí public/dashboard/dashboard.png.
Sales and Earnings
Sales are connected to products and business information so that the platform can later analyze what is happening.
The earnings section provides another layer of visibility into the operation.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fnkavc69s6vl0zydyyvk9.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fnkavc69s6vl0zydyyvk9.png" alt=" " width="800" height="386"&gt;&lt;/a&gt;&lt;br&gt;
&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fzfiva0jd6jsuwp6i1mc5.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fzfiva0jd6jsuwp6i1mc5.png" alt=" " width="800" height="386"&gt;&lt;/a&gt;&lt;br&gt;
&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Ff4xt8pt9y3elyff6e3xz.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Ff4xt8pt9y3elyff6e3xz.png" alt=" " width="800" height="407"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fc9xtqglaj3r2jx952psf.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fc9xtqglaj3r2jx952psf.png" alt=" " width="799" height="384"&gt;&lt;/a&gt;&lt;br&gt;
&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F486axwc9z6bfjp2civ5d.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F486axwc9z6bfjp2civ5d.png" alt=" " width="800" height="386"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;These sections are important because they provide the data that future AI workflows can use for deeper business analysis.&lt;br&gt;
Open AI and Local Development&lt;br&gt;
One of the requirements of this Hacktoberfest challenge is to build something with open-source AI at its core.&lt;br&gt;
Collectify was designed around an AI layer that can work with open/open-weight models and different providers, rather than making the entire application permanently dependent on a single proprietary model.&lt;br&gt;
The application was also developed and tested locally, including both the Angular frontend and the FastAPI backend.&lt;br&gt;
This is important to me because I want the AI layer to remain replaceable.&lt;br&gt;
Collectify&lt;br&gt;
     ↓&lt;br&gt;
AI Service&lt;br&gt;
     ↓&lt;br&gt;
Model / Provider&lt;/p&gt;

&lt;p&gt;The application should be able to evolve as models evolve.&lt;br&gt;
That means being able to:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Experiment with different models.&lt;/li&gt;
&lt;li&gt;Replace providers.&lt;/li&gt;
&lt;li&gt;Test different AI capabilities.&lt;/li&gt;
&lt;li&gt;Adapt the AI layer without redesigning the entire application.&lt;/li&gt;
&lt;li&gt;Run the core application locally.
An AI-first application should not necessarily mean being locked into one model forever.
Evidence of the AI Workflow
The screenshots below document the actual development and functionality of the AI features.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F236tgzpflv5morrbek2d.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F236tgzpflv5morrbek2d.png" alt=" " width="800" height="432"&gt;&lt;/a&gt;&lt;br&gt;
&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F4pyk92c19o3bktc67ny6.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F4pyk92c19o3bktc67ny6.png" alt=" " width="800" height="417"&gt;&lt;/a&gt;&lt;br&gt;
&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fyuvis0ur13rm6p9i14uz.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fyuvis0ur13rm6p9i14uz.png" alt=" " width="800" height="400"&gt;&lt;/a&gt;&lt;br&gt;
&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F8nd304ng0ytv7cbn3v23.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F8nd304ng0ytv7cbn3v23.png" alt=" " width="800" height="391"&gt;&lt;/a&gt;&lt;br&gt;
Technology Stack&lt;br&gt;
Frontend&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Angular&lt;/li&gt;
&lt;li&gt;TypeScript&lt;/li&gt;
&lt;li&gt;SCSS&lt;/li&gt;
&lt;li&gt;Angular HttpClient&lt;/li&gt;
&lt;li&gt;Responsive UI
Backend&lt;/li&gt;
&lt;li&gt;Python&lt;/li&gt;
&lt;li&gt;FastAPI&lt;/li&gt;
&lt;li&gt;Pydantic&lt;/li&gt;
&lt;li&gt;Uvicorn&lt;/li&gt;
&lt;li&gt;REST API
Database&lt;/li&gt;
&lt;li&gt;Supabase&lt;/li&gt;
&lt;li&gt;PostgreSQL
AI&lt;/li&gt;
&lt;li&gt;OpenRouter&lt;/li&gt;
&lt;li&gt;Open/open-weight AI models&lt;/li&gt;
&lt;li&gt;AI-powered business analysis&lt;/li&gt;
&lt;li&gt;Structured AI supplier discovery
Deployment&lt;/li&gt;
&lt;li&gt;Vercel&lt;/li&gt;
&lt;li&gt;Render&lt;/li&gt;
&lt;li&gt;Supabase
Architecture
              ┌─────────────────────────┐
              │       Collectify        │
              │     Angular Frontend    │
              └────────────┬────────────┘
                           │
                        HTTP/REST
                           │
                           ▼
              ┌─────────────────────────┐
              │     FastAPI Backend     │
              │                         │
              │ Products                │
              │ Sales                   │
              │ Partners                │
              │ Earnings                │
              │ AI Analysis             │
              │ Supplier Discovery      │
              └────────────┬────────────┘
                           │
                ┌──────────┴──────────┐
                │                     │
                ▼                     ▼
      ┌──────────────────┐   ┌──────────────────┐
      │    Supabase      │   │    OpenRouter    │
      │    PostgreSQL    │   │    AI Models     │
      └──────────────────┘   └──────────────────┘&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Running Locally&lt;br&gt;
Clone the repository:&lt;br&gt;
git clone &lt;a href="https://github.com/Ezequie1Sc/collectify.git" rel="noopener noreferrer"&gt;https://github.com/Ezequie1Sc/collectify.git&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Install the frontend dependencies:&lt;br&gt;
cd frontend/collectify&lt;br&gt;
pnpm install&lt;/p&gt;

&lt;p&gt;Run the frontend:&lt;br&gt;
pnpm start&lt;/p&gt;

&lt;p&gt;The Angular application will be available at:&lt;br&gt;
&lt;a href="http://localhost:4200" rel="noopener noreferrer"&gt;http://localhost:4200&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;For the backend:&lt;br&gt;
uvicorn app.main:app --reload&lt;/p&gt;

&lt;p&gt;The API will run at:&lt;br&gt;
&lt;a href="http://127.0.0.1:8000" rel="noopener noreferrer"&gt;http://127.0.0.1:8000&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Health check:&lt;br&gt;
&lt;a href="http://127.0.0.1:8000/health" rel="noopener noreferrer"&gt;http://127.0.0.1:8000/health&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;The project was developed and tested locally before deployment.&lt;br&gt;
Live Project&lt;br&gt;
Frontend&lt;br&gt;
&lt;a href="https://collectify-7xcu.vercel.app" rel="noopener noreferrer"&gt;https://collectify-7xcu.vercel.app&lt;/a&gt;&lt;br&gt;
Backend API&lt;br&gt;
&lt;a href="https://collectify-api-udxk.onrender.com" rel="noopener noreferrer"&gt;https://collectify-api-udxk.onrender.com&lt;/a&gt;&lt;br&gt;
API Documentation&lt;br&gt;
&lt;a href="https://collectify-api-udxk.onrender.com/docs" rel="noopener noreferrer"&gt;https://collectify-api-udxk.onrender.com/docs&lt;/a&gt;&lt;br&gt;
GitHub Repository&lt;br&gt;
&lt;a href="https://github.com/Ezequie1Sc/collectify" rel="noopener noreferrer"&gt;https://github.com/Ezequie1Sc/collectify&lt;/a&gt;&lt;br&gt;
Why This Project Matters&lt;br&gt;
Small businesses generate valuable information every day.&lt;br&gt;
But that information is often distributed across:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Spreadsheets&lt;/li&gt;
&lt;li&gt;Messages&lt;/li&gt;
&lt;li&gt;Notebooks&lt;/li&gt;
&lt;li&gt;Manual records&lt;/li&gt;
&lt;li&gt;Different applications&lt;/li&gt;
&lt;li&gt;Conversations between partners
The problem is not necessarily the lack of data.
The problem is the lack of understanding.
Collectify attempts to create a bridge between operational data and business decisions.
Inventory
↓
Sales
↓
Business Data
↓
AI
↓
Insights
↓
Recommendations
↓
Action&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;That is the direction I believe AI can bring real value to small businesses.&lt;br&gt;
What Makes Collectify Different?&lt;br&gt;
Collectify is not trying to compete with large enterprise ERP systems.&lt;br&gt;
The goal is different.&lt;br&gt;
It is about taking a real small-business problem and asking:&lt;br&gt;
What could happen if the business's own data became something the application could understand?&lt;/p&gt;

&lt;p&gt;The project combines:&lt;/p&gt;

&lt;h2&gt;
  
  
  Inventory management
&lt;/h2&gt;

&lt;h2&gt;
  
  
  Multi-partner management
&lt;/h2&gt;

&lt;h2&gt;
  
  
  Sales and earnings
&lt;/h2&gt;

&lt;h2&gt;
  
  
  AI analysis
&lt;/h2&gt;

&lt;h1&gt;
  
  
  AI-powered supplier discovery
&lt;/h1&gt;

&lt;p&gt;An AI-first business platform&lt;br&gt;
And the most important part is that this architecture was shaped by an actual business problem.&lt;br&gt;
What's Next?&lt;br&gt;
Collectify is still evolving.&lt;br&gt;
Some of the features I want to explore next include:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Sales forecasting&lt;/li&gt;
&lt;li&gt;Automatic restock recommendations&lt;/li&gt;
&lt;li&gt;Partner-specific analytics&lt;/li&gt;
&lt;li&gt;Supplier comparison&lt;/li&gt;
&lt;li&gt;Purchase recommendations&lt;/li&gt;
&lt;li&gt;Profitability analysis&lt;/li&gt;
&lt;li&gt;Automated business insights&lt;/li&gt;
&lt;li&gt;Role-based access control&lt;/li&gt;
&lt;li&gt;Advanced dashboards&lt;/li&gt;
&lt;li&gt;More open-weight AI models&lt;/li&gt;
&lt;li&gt;More autonomous AI workflows
The long-term goal is to move from:
Inventory Management&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;towards:&lt;br&gt;
AI-Powered Business Intelligence&lt;br&gt;
for Small Retail Businesses&lt;/p&gt;

&lt;p&gt;Final Thoughts&lt;br&gt;
This project changed the way I think about AI applications.&lt;br&gt;
AI does not have to be a chat window placed next to an existing application.&lt;br&gt;
It can become part of the application's actual workflow.&lt;br&gt;
With Collectify, the goal is:&lt;br&gt;
Organize Information&lt;br&gt;
        ↓&lt;br&gt;
Understand the Business&lt;br&gt;
        ↓&lt;br&gt;
Discover Opportunities&lt;br&gt;
        ↓&lt;br&gt;
Make Better Decisions&lt;/p&gt;

&lt;p&gt;And most importantly, this is not a hypothetical problem.&lt;br&gt;
Collectify was built around a real local business, with real operational needs and a real multi-partner inventory problem.&lt;br&gt;
That made the project much more meaningful to build.&lt;br&gt;
I did not start with:&lt;br&gt;
"What AI feature can I add to my application?"&lt;/p&gt;

&lt;p&gt;I started with:&lt;br&gt;
"What problem does this business actually have, and where can AI genuinely help?"&lt;/p&gt;

&lt;p&gt;That question shaped the entire project.&lt;br&gt;
AI should not be decoration. It should solve a real problem.&lt;/p&gt;

&lt;p&gt;Built for Hacktoberfest 2026&lt;br&gt;
Hacktoberfest Weekend Challenge — Build for a Friend&lt;br&gt;
Collectify — AI-powered business management for real-world retail.&lt;br&gt;
GitHub:&lt;br&gt;
&lt;a href="https://github.com/Ezequie1Sc/collectify" rel="noopener noreferrer"&gt;https://github.com/Ezequie1Sc/collectify&lt;/a&gt;&lt;br&gt;
Live Demo:&lt;br&gt;
&lt;a href="https://collectify-7xcu.vercel.app" rel="noopener noreferrer"&gt;https://collectify-7xcu.vercel.app&lt;/a&gt;&lt;br&gt;
API Documentation:&lt;br&gt;
&lt;a href="https://collectify-api-udxk.onrender.com/docs" rel="noopener noreferrer"&gt;https://collectify-api-udxk.onrender.com/docs&lt;/a&gt;&lt;/p&gt;

</description>
      <category>hf26challenge</category>
    </item>
    <item>
      <title>Hi</title>
      <dc:creator>Orlando Ezequiel Salazar Cruz</dc:creator>
      <pubDate>Mon, 05 Oct 2026 04:02:09 +0000</pubDate>
      <link>https://dev.to/ezequie1sc/hi-5chk</link>
      <guid>https://dev.to/ezequie1sc/hi-5chk</guid>
      <description></description>
    </item>
  </channel>
</rss>
