<?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: Kavya Saraswat</title>
    <description>The latest articles on DEV Community by Kavya Saraswat (@kavya_saraswat_64c132105f).</description>
    <link>https://dev.to/kavya_saraswat_64c132105f</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%2F4188153%2Fae8ace81-9af7-4c93-b1ff-b36bb9e8aa40.png</url>
      <title>DEV Community: Kavya Saraswat</title>
      <link>https://dev.to/kavya_saraswat_64c132105f</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/kavya_saraswat_64c132105f"/>
    <language>en</language>
    <item>
      <title>Spiritus 🕯️🌼: An Air-Gapped Virtual Ofrenda &amp; Remembrance Sanctuary Powered by Local Gemma 3</title>
      <dc:creator>Kavya Saraswat</dc:creator>
      <pubDate>Sat, 31 Oct 2026 18:35:22 +0000</pubDate>
      <link>https://dev.to/kavya_saraswat_64c132105f/spiritus-an-air-gapped-virtual-ofrenda-remembrance-sanctuary-powered-by-local-gemma-3-4dki</link>
      <guid>https://dev.to/kavya_saraswat_64c132105f/spiritus-an-air-gapped-virtual-ofrenda-remembrance-sanctuary-powered-by-local-gemma-3-4dki</guid>
      <description>&lt;p&gt;&lt;em&gt;This is a submission for the &lt;a href="https://dev.to/challenges/hacktoberfest-week4-2026-10-26"&gt;Hacktoberfest Open-Source AI Challenge Week 4: Ghost in the Machine&lt;/a&gt;&lt;/em&gt;&lt;/p&gt;




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

&lt;p&gt;In Mexican tradition, Día de los Muertos (November 1 &amp;amp; 2) reminds us that &lt;em&gt;“the dead are never truly gone as long as they are remembered.”&lt;/em&gt; Families assemble &lt;strong&gt;ofrendas&lt;/strong&gt; (altars) adorned with marigolds (&lt;em&gt;cempasúchil&lt;/em&gt;), flickering candles (&lt;em&gt;veladoras&lt;/em&gt;), sweet &lt;em&gt;pan de muerto&lt;/em&gt;, copal incense, and personal keepsakes to welcome spirits back to the earthly plane.&lt;/p&gt;

&lt;p&gt;For the &lt;strong&gt;Ghost in the Machine&lt;/strong&gt; theme, I built &lt;strong&gt;Spiritus&lt;/strong&gt;: an open-source, air-gapped, privacy-first virtual ofrenda builder and remembrance writer powered by open-weight AI (&lt;strong&gt;Gemma 3&lt;/strong&gt; via &lt;strong&gt;Ollama&lt;/strong&gt;).&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;                      🕯️   🌼   🕯️
               ┌───────────────────────────┐
               │    S P I R I T U S        │
               │   Virtual Ofrenda Altar   │
               │      &amp;amp; Memory Vault       │
               └─────────────┬─────────────┘
                             │
     Local Open-Weights      ▼      Real-Time SSE
      (Gemma 3 / Ollama) ─────────► Token Streaming
                             ▲      + Web Speech TTS
                             │
               ┌─────────────┴─────────────┐
               │  Zero Telemetry · 100%    │
               │  Air-Gapped &amp;amp; Offline     │
               └───────────────────────────┘
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;

&lt;h3&gt;
  
  
  Key Highlights:
&lt;/h3&gt;

&lt;ul&gt;
&lt;li&gt;🕯️ &lt;strong&gt;Interactive Living Altar&lt;/strong&gt;: Dynamically arrange traditional offerings (candles, marigold petals, pan de muerto, mezcal, family photos, and handwritten mementos) with a live canvas particle simulation that scatters golden petals across the screen.&lt;/li&gt;
&lt;li&gt;🪶 &lt;strong&gt;Tuned Remembrance Generator&lt;/strong&gt;: Feeds offering details, beloved nicknames, milestones, and quirks into &lt;strong&gt;Google Gemma 3&lt;/strong&gt; to compose authentic eulogies and poems across 4 distinct cultural tones (&lt;em&gt;Poetic&lt;/em&gt;, &lt;em&gt;Celebratory / ¡Que viva el recuerdo!&lt;/em&gt;, &lt;em&gt;Solemn&lt;/em&gt;, and &lt;em&gt;Storyteller&lt;/em&gt;).&lt;/li&gt;
&lt;li&gt;⚡ &lt;strong&gt;Real-Time Token Streaming&lt;/strong&gt;: Server-Sent Events (SSE) stream the AI's words character-by-character, capturing the solemnity of a handwritten letter unfolding before your eyes.&lt;/li&gt;
&lt;li&gt;🔊 &lt;strong&gt;Recite Aloud (Speech Synthesis)&lt;/strong&gt;: Built-in atmospheric narration reads the remembrance aloud with reverent pacing.&lt;/li&gt;
&lt;li&gt;💬 &lt;strong&gt;Multi-Turn Spirit Chat&lt;/strong&gt;: Converse with the remembrance model afterwards—ask it to translate the tribute into Spanish, add a verse about their favorite vinyl record, or reminisce about a shared recipe.&lt;/li&gt;
&lt;li&gt;💾 &lt;strong&gt;Local File Persistence &amp;amp; Export&lt;/strong&gt;: Saves full altar records to &lt;code&gt;data/ofrendas.json&lt;/code&gt; and exports mementos as &lt;code&gt;.txt&lt;/code&gt; files with zero cloud dependencies.&lt;/li&gt;
&lt;li&gt;🛡️ &lt;strong&gt;Zero-Config Demo Mode&lt;/strong&gt;: For visitors without a local Ollama instance running, a built-in fallback mode generates complete poetic tributes so anyone can test the interface instantly.&lt;/li&gt;
&lt;/ul&gt;


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

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;GitHub Repository&lt;/strong&gt;: &lt;a href="https://github.com/KavyaSaraswat23/Hactober_week4" rel="noopener noreferrer"&gt;github.com/KavyaSaraswat23/Hactober_week4&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Live Local Setup Preview&lt;/strong&gt;:&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%2F3qsygeon03zub35ys5bz.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%2F3qsygeon03zub35ys5bz.jpg" alt="Spiritus Hero Banner" width="800" height="447"&gt;&lt;/a&gt;&lt;/p&gt;
&lt;h3&gt;
  
  
  Quick 30-Second Local Run:
&lt;/h3&gt;


&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;&lt;span class="c"&gt;# 1. Clone the repository&lt;/span&gt;
git clone https://github.com/KavyaSaraswat23/Hactober_week4.git
&lt;span class="nb"&gt;cd &lt;/span&gt;Hactober_week4

&lt;span class="c"&gt;# 2. Install lightweight dependencies (Express, CORS, Node-Fetch)&lt;/span&gt;
npm &lt;span class="nb"&gt;install&lt;/span&gt;

&lt;span class="c"&gt;# 3. Start the application&lt;/span&gt;
npm start
&lt;span class="c"&gt;# -&amp;gt; Visit http://localhost:3000&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;


&lt;p&gt;&lt;em&gt;(If Ollama is running with &lt;code&gt;gemma3:4b&lt;/code&gt; or &lt;code&gt;gemma3:1b&lt;/code&gt;, it connects automatically. If not, toggle Demo Mode with one click to test the entire interactive flow!)&lt;/em&gt;&lt;/p&gt;


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


&lt;div class="ltag-github-readme-tag"&gt;
  &lt;div class="readme-overview"&gt;
    &lt;h2&gt;
      &lt;img src="https://assets.dev.to/assets/github-logo-5a155e1f9a670af7944dd5e12375bc76ed542ea80224905ecaf878b9157cdefc.svg" alt="GitHub logo"&gt;
      &lt;a href="https://github.com/KavyaSaraswat23" rel="noopener noreferrer"&gt;
        KavyaSaraswat23
      &lt;/a&gt; / &lt;a href="https://github.com/KavyaSaraswat23/Hactober_week4" rel="noopener noreferrer"&gt;
        Hactober_week4
      &lt;/a&gt;
    &lt;/h2&gt;
    &lt;h3&gt;
      
    &lt;/h3&gt;
  &lt;/div&gt;
  &lt;div class="ltag-github-body"&gt;
    
&lt;div id="readme" class="md"&gt;&lt;div class="markdown-heading"&gt;
&lt;h1 class="heading-element"&gt;💀 Spiritus — Virtual Ofrenda Builder &amp;amp; Remembrance Generator&lt;/h1&gt;
&lt;/div&gt;

&lt;div&gt;
&lt;p&gt;&lt;a href="https://opensource.org/licenses/MIT" rel="nofollow noopener noreferrer"&gt;&lt;img src="https://camo.githubusercontent.com/fdf2982b9f5d7489dcf44570e714e3a15fce6253e0cc6b5aa61a075aac2ff71b/68747470733a2f2f696d672e736869656c64732e696f2f62616467652f4c6963656e73652d4d49542d79656c6c6f772e737667" alt="License: MIT"&gt;&lt;/a&gt;
&lt;a href="https://nodejs.org/" rel="nofollow noopener noreferrer"&gt;&lt;img src="https://camo.githubusercontent.com/dbc863a2f32d5f4a24b8f57467e7a69e81a882b215e80be4ec92f4f922dc12c8/68747470733a2f2f696d672e736869656c64732e696f2f62616467652f4e6f64652e6a732d7631382532422d677265656e2e737667" alt="Node.js"&gt;&lt;/a&gt;
&lt;a href="https://ollama.com/" rel="nofollow noopener noreferrer"&gt;&lt;img src="https://camo.githubusercontent.com/bfd4c9c8b5bbfacb1aa701728febfed8910a374f9012e145633ae021ea14f17c/68747470733a2f2f696d672e736869656c64732e696f2f62616467652f4f6c6c616d612d4c6f63616c25323041492d626c75652e737667" alt="Ollama"&gt;&lt;/a&gt;
&lt;a href="https://deepmind.google/technologies/gemma/" rel="nofollow noopener noreferrer"&gt;&lt;img src="https://camo.githubusercontent.com/02c651299cf62bbd80f3dfa6e812bdcc8778d05763aeacc67e2ef2349b4a7485/68747470733a2f2f696d672e736869656c64732e696f2f62616467652f4d6f64656c2d47656d6d61253230332532302532462532304d69737472616c2532302532462532304c6c616d612d707572706c652e737667" alt="Gemma 3"&gt;&lt;/a&gt;
&lt;a href="https://dev.to/challenges/hacktoberfest-week4-2026-10-26" rel="nofollow"&gt;&lt;img src="https://camo.githubusercontent.com/fb44f55c6f3e2ad8402b784613566771df193685b69d7d95865287524960bab6/68747470733a2f2f696d672e736869656c64732e696f2f62616467652f4861636b746f626572666573742d5765656b253230342532304368616c6c656e67652d6f72616e67652e737667" alt="Hacktoberfest 2026"&gt;&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;&lt;em&gt;“The dead are never truly gone as long as they are remembered.”&lt;/em&gt;&lt;/p&gt;
&lt;br&gt;
&lt;p&gt;&lt;a rel="noopener noreferrer" href="https://github.com/KavyaSaraswat23/Hactober_week4/./public/hero.jpg"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fraw.githubusercontent.com%2FKavyaSaraswat23%2FHactober_week4%2FHEAD%2F.%2Fpublic%2Fhero.jpg" alt="Spiritus Hero Image"&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/div&gt;




&lt;div class="markdown-heading"&gt;
&lt;h2 class="heading-element"&gt;📖 About Spiritus&lt;/h2&gt;
&lt;/div&gt;

&lt;p&gt;&lt;strong&gt;Spiritus&lt;/strong&gt; is an open-source, privacy-first virtual &lt;em&gt;ofrenda&lt;/em&gt; (altar) builder and remembrance writer inspired by the rich cultural tradition of &lt;strong&gt;Día de los Muertos&lt;/strong&gt; (Day of the Dead) and built for the &lt;strong&gt;Hacktoberfest 2026: Ghost in the Machine&lt;/strong&gt; challenge.&lt;/p&gt;

&lt;p&gt;In a digital age where memories are routinely scraped, monetized, and trapped behind paid cloud subscriptions, &lt;strong&gt;Spiritus&lt;/strong&gt; takes the opposite stance: &lt;strong&gt;your most sacred memories belong to you alone&lt;/strong&gt;. Powered by open-weights local AI (Gemma, Mistral, LLaMA) via &lt;strong&gt;Ollama&lt;/strong&gt;, all token generation happens directly on your machine. &lt;strong&gt;Zero telemetry. Zero external API keys. Zero cloud fees.&lt;/strong&gt;&lt;/p&gt;




&lt;div class="markdown-heading"&gt;
&lt;h2 class="heading-element"&gt;✨ Features&lt;/h2&gt;
&lt;/div&gt;

&lt;ul&gt;
&lt;li&gt;🕯️ &lt;strong&gt;Interactive Digital Ofrenda&lt;/strong&gt;: Place marigold petals (&lt;em&gt;cempasúchil&lt;/em&gt;), light candles, offer pan de muerto, copal incense, and favorite foods while seeing the live altar update…&lt;/li&gt;
&lt;/ul&gt;&lt;/div&gt;
  &lt;/div&gt;
  &lt;div class="gh-btn-container"&gt;&lt;a class="gh-btn" href="https://github.com/KavyaSaraswat23/Hactober_week4" rel="noopener noreferrer"&gt;View on GitHub&lt;/a&gt;&lt;/div&gt;
&lt;/div&gt;


&lt;p&gt;The core stack is purposely lean and inspectable:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Frontend&lt;/strong&gt;: Vanilla HTML5, CSS3 with gothic-warm glassmorphism, floating canvas particle emitters, and native Web APIs (Fetch, SSE &lt;code&gt;EventSource&lt;/code&gt;, Web Speech API).&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Backend&lt;/strong&gt;: Node.js &amp;amp; Express proxying requests to Ollama's native endpoint (&lt;code&gt;http://localhost:11434/api/generate&lt;/code&gt;).&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Storage&lt;/strong&gt;: Local filesystem JSON storage (&lt;code&gt;data/ofrendas.json&lt;/code&gt;).&lt;/li&gt;
&lt;/ul&gt;




&lt;h2&gt;
  
  
  How I Built It
&lt;/h2&gt;

&lt;h3&gt;
  
  
  1. The Open-Weight Core: Google Gemma 3 via Ollama
&lt;/h3&gt;

&lt;p&gt;The brain of Spiritus runs on &lt;strong&gt;Google Gemma 3&lt;/strong&gt; (&lt;code&gt;gemma3:4b&lt;/code&gt; or &lt;code&gt;gemma3:1b&lt;/code&gt;). Gemma's literary nuance and multilingual understanding make it particularly suited for the lyrical, respectful register required for Día de los Muertos tributes.&lt;/p&gt;

&lt;p&gt;The Node.js backend streams output from Ollama straight to the client via Server-Sent Events (SSE):&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight javascript"&gt;&lt;code&gt;&lt;span class="c1"&gt;// server.js - Real-time streaming from local Ollama&lt;/span&gt;
&lt;span class="nx"&gt;app&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;post&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;/api/generate&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="k"&gt;async &lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="nx"&gt;req&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="nx"&gt;res&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="o"&gt;=&amp;gt;&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
  &lt;span class="kd"&gt;const&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="nx"&gt;model&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;gemma3:4b&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="nx"&gt;prompt&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="nx"&gt;context&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="p"&gt;[]&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="nx"&gt;req&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nx"&gt;body&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;

  &lt;span class="nx"&gt;res&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;setHeader&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;Content-Type&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;text/event-stream&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;);&lt;/span&gt;
  &lt;span class="nx"&gt;res&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;setHeader&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;Cache-Control&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;no-cache&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;);&lt;/span&gt;
  &lt;span class="nx"&gt;res&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;setHeader&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;Connection&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;keep-alive&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;);&lt;/span&gt;

  &lt;span class="kd"&gt;const&lt;/span&gt; &lt;span class="nx"&gt;ollamaRes&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="k"&gt;await&lt;/span&gt; &lt;span class="nf"&gt;fetch&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="s2"&gt;`&lt;/span&gt;&lt;span class="p"&gt;${&lt;/span&gt;&lt;span class="nx"&gt;OLLAMA_BASE&lt;/span&gt;&lt;span class="p"&gt;}&lt;/span&gt;&lt;span class="s2"&gt;/api/generate`&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
    &lt;span class="na"&gt;method&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;POST&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
    &lt;span class="na"&gt;headers&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;Content-Type&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;application/json&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt; &lt;span class="p"&gt;},&lt;/span&gt;
    &lt;span class="na"&gt;body&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="nx"&gt;JSON&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;stringify&lt;/span&gt;&lt;span class="p"&gt;({&lt;/span&gt; &lt;span class="nx"&gt;model&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="nx"&gt;prompt&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="nx"&gt;context&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;stream&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="kc"&gt;true&lt;/span&gt; &lt;span class="p"&gt;}),&lt;/span&gt;
  &lt;span class="p"&gt;});&lt;/span&gt;

  &lt;span class="nx"&gt;ollamaRes&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nx"&gt;body&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;on&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;data&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="nx"&gt;chunk&lt;/span&gt; &lt;span class="o"&gt;=&amp;gt;&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
    &lt;span class="kd"&gt;const&lt;/span&gt; &lt;span class="nx"&gt;lines&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="nx"&gt;chunk&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;toString&lt;/span&gt;&lt;span class="p"&gt;().&lt;/span&gt;&lt;span class="nf"&gt;split&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="se"&gt;\n&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;).&lt;/span&gt;&lt;span class="nf"&gt;filter&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="nb"&gt;Boolean&lt;/span&gt;&lt;span class="p"&gt;);&lt;/span&gt;
    &lt;span class="k"&gt;for &lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="kd"&gt;const&lt;/span&gt; &lt;span class="nx"&gt;line&lt;/span&gt; &lt;span class="k"&gt;of&lt;/span&gt; &lt;span class="nx"&gt;lines&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
      &lt;span class="k"&gt;try&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
        &lt;span class="kd"&gt;const&lt;/span&gt; &lt;span class="nx"&gt;parsed&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="nx"&gt;JSON&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;parse&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="nx"&gt;line&lt;/span&gt;&lt;span class="p"&gt;);&lt;/span&gt;
        &lt;span class="nx"&gt;res&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;write&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="s2"&gt;`data: &lt;/span&gt;&lt;span class="p"&gt;${&lt;/span&gt;&lt;span class="nx"&gt;JSON&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;stringify&lt;/span&gt;&lt;span class="p"&gt;({&lt;/span&gt; &lt;span class="na"&gt;token&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="nx"&gt;parsed&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nx"&gt;response&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;done&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="nx"&gt;parsed&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nx"&gt;done&lt;/span&gt; &lt;span class="p"&gt;})}&lt;/span&gt;&lt;span class="s2"&gt;\n\n`&lt;/span&gt;&lt;span class="p"&gt;);&lt;/span&gt;
      &lt;span class="p"&gt;}&lt;/span&gt; &lt;span class="k"&gt;catch&lt;/span&gt; &lt;span class="p"&gt;{}&lt;/span&gt;
    &lt;span class="p"&gt;}&lt;/span&gt;
  &lt;span class="p"&gt;});&lt;/span&gt;

  &lt;span class="nx"&gt;ollamaRes&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nx"&gt;body&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;on&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;end&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="p"&gt;()&lt;/span&gt; &lt;span class="o"&gt;=&amp;gt;&lt;/span&gt; &lt;span class="nx"&gt;res&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;end&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;h3&gt;
  
  
  2. Tone Architecture
&lt;/h3&gt;

&lt;p&gt;Rather than generic praise, Spiritus injects cultural context into the system prompt based on user-selected tone:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Poetic&lt;/strong&gt;: Focuses on natural metaphors, cempasúchil petals, evening breezes, and eternal light.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Celebratory&lt;/strong&gt;: Focuses on laughter, music, festive memories, and the joy of a life fully lived (&lt;em&gt;"No llores por mí..."&lt;/em&gt;).&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Solemn&lt;/strong&gt;: Traditional, reverent prayers and dignified gratitude.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Storyteller&lt;/strong&gt;: Weaves specific remembered anecdotes (favorite meals, funny quirks, treasured advice) into a cohesive biographical narrative.&lt;/li&gt;
&lt;/ul&gt;




&lt;h2&gt;
  
  
  Why Does Open Innovation Matter?
&lt;/h2&gt;

&lt;p&gt;When we build AI applications around grief, heritage, and family remembrance, the question of &lt;strong&gt;who owns the infrastructure&lt;/strong&gt; is not an academic debate—it is an ethical boundary.&lt;/p&gt;

&lt;h3&gt;
  
  
  1. Sacred Memories Must Never Be Scraped
&lt;/h3&gt;

&lt;p&gt;Commercial cloud LLM APIs train on user inputs or log transcripts on remote servers. Grief is not training data. By running &lt;strong&gt;Gemma 3 locally on your laptop&lt;/strong&gt;, what you write about your grandmother, your departed mentor, or your childhood friend never leaves your device. Not a single packet crosses the open internet.&lt;/p&gt;

&lt;h3&gt;
  
  
  2. Permanence Beyond Corporate Lifecycles
&lt;/h3&gt;

&lt;p&gt;Closed APIs deprecate endpoints, change terms of service, raise token prices, or shut down entirely. A local open-weight model (&lt;code&gt;.gguf&lt;/code&gt;) will run on your hardware 5, 10, or 20 years from now. As long as your machine powers on, your ofrenda remains accessible.&lt;/p&gt;

&lt;h3&gt;
  
  
  3. Air-Gapped &amp;amp; Accessible Everywhere
&lt;/h3&gt;

&lt;p&gt;Whether you are sitting in a rural cemetery with no cellular signal, an off-grid cabin, or an area with expensive bandwidth, open-source AI runs entirely offline.&lt;/p&gt;




&lt;h2&gt;
  
  
  The Spirit of It
&lt;/h2&gt;

&lt;p&gt;During testing, I set up an ofrenda for my grandfather, listing his signature wool cardigan, black cardamom chai, and the old brass transistor radio he listened to every morning.&lt;/p&gt;

&lt;p&gt;When local Gemma streamed the closing lines:&lt;/p&gt;

&lt;blockquote&gt;
&lt;p&gt;&lt;em&gt;“Let the scent of roasted cardamom lead you back through the mist tonight. Your chair by the radio is not empty; it is simply waiting for your song to finish.”&lt;/em&gt;&lt;/p&gt;
&lt;/blockquote&gt;

&lt;p&gt;The room went completely still. It wasn't a sterile corporate autocomplete; it was a deeply personal, comforting echo. That moment captured the entire essence of the &lt;strong&gt;Ghost in the Machine&lt;/strong&gt;: using open silicon and weights to rekindle the human warmth that technology so often forgets.&lt;/p&gt;




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

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Featured Category: Best Use of Gemma&lt;/strong&gt; ($200)

&lt;ul&gt;
&lt;li&gt;Spiritus is built specifically around &lt;strong&gt;Google Gemma 3&lt;/strong&gt; (&lt;code&gt;gemma3:4b&lt;/code&gt; / &lt;code&gt;gemma3:1b&lt;/code&gt;) running through Ollama. Gemma's tone control, cultural empathy, and lightweight footprint enable zero-telemetry, on-device remembrance generation with low latency on everyday consumer hardware.&lt;/li&gt;
&lt;/ul&gt;
&lt;/li&gt;
&lt;/ul&gt;




&lt;p&gt;&lt;strong&gt;Author&lt;/strong&gt;: Kavya Saraswat (&lt;a href="https://github.com/KavyaSaraswat23" rel="noopener noreferrer"&gt;@KavyaSaraswat23&lt;/a&gt;)&lt;/p&gt;

</description>
      <category>devchallenge</category>
      <category>hf26challenge</category>
    </item>
    <item>
      <title>CraftList AI: Why Open Innovation Powers My 100% Local E-Commerce Listing Studio with Gemma 2</title>
      <dc:creator>Kavya Saraswat</dc:creator>
      <pubDate>Mon, 19 Oct 2026 05:50:06 +0000</pubDate>
      <link>https://dev.to/kavya_saraswat_64c132105f/craftlist-ai-why-open-innovation-powers-my-100-local-e-commerce-listing-studio-with-gemma-2-5bof</link>
      <guid>https://dev.to/kavya_saraswat_64c132105f/craftlist-ai-why-open-innovation-powers-my-100-local-e-commerce-listing-studio-with-gemma-2-5bof</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;h2&gt;
  
  
  What I Built
&lt;/h2&gt;

&lt;p&gt;I built &lt;strong&gt;CraftList AI&lt;/strong&gt; — a privacy-first, zero-cloud e-commerce listing studio and profit margin engine designed for independent artisans, vintage thrifters, and handmade makers selling on Etsy, eBay, Shopify, and Amazon Handmade.&lt;/p&gt;

&lt;h3&gt;
  
  
  Who Would Pay For It &amp;amp; What Problem Does It Solve?
&lt;/h3&gt;

&lt;p&gt;Solo makers and creative side-hustlers love making products, but they universally dread writing marketplace listings. For every handmade mug, leather wallet, or carved wooden utensil, an artisan spends 45–60 minutes wrestling with:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;strong&gt;Algorithmic Titles &amp;amp; SEO Copy:&lt;/strong&gt; Cramming high-volume search phrases into 140 characters without sounding robotic.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;The 13-Tag Etsy Matrix:&lt;/strong&gt; Generating exactly 13 unique tags under Etsy's strict rule of &lt;strong&gt;20 characters or fewer&lt;/strong&gt; per tag.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Pricing &amp;amp; Margin Anxiety:&lt;/strong&gt; Underpricing labor or overlooking marketplace take-rates (6.5% transaction + payment processing + listing fees), accidentally working below minimum wage.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Multi-Channel Social Promotion:&lt;/strong&gt; Writing Instagram captions, Pinterest SEO pins, and TikTok hooks.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;Existing SaaS listing tools charge $30–$50/month on subscription, meter tokens aggressively, and upload unreleased product designs to proprietary cloud servers. &lt;strong&gt;CraftList AI&lt;/strong&gt; runs &lt;strong&gt;100% locally on the seller's laptop using Google Gemma 2&lt;/strong&gt;, costing $0.00 to run and ensuring complete data privacy.&lt;/p&gt;

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

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Local Web App:&lt;/strong&gt; Runs locally via Vite + React (&lt;code&gt;http://localhost:5173&lt;/code&gt;)&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Key Capabilities Demonstrated:&lt;/strong&gt;

&lt;ul&gt;
&lt;li&gt;One-click sample presets (Hand-Carved Walnut Knife, Speckled Ceramic Mug, Horween Leather Wallet, Botanical Soy Candle).&lt;/li&gt;
&lt;li&gt;Multi-marketplace selector (Etsy, eBay, Shopify, Amazon Handmade).&lt;/li&gt;
&lt;li&gt;Built-in Side-Hustle Profit Calculator (Materials + Labor Hours × Rate + Marketplace Fees = Suggested Retail Price &amp;amp; Margin %).&lt;/li&gt;
&lt;li&gt;Real-time Tag Auditor (13 tags, character counter per tag, duplicate keyword rejection).&lt;/li&gt;
&lt;li&gt;Social Commerce Launch suite (Instagram carousel captions, Pinterest SEO pins, TikTok ASMR video scripts).&lt;/li&gt;
&lt;li&gt;One-click bulk CSV and JSON export for instant marketplace store imports.&lt;/li&gt;
&lt;/ul&gt;
&lt;/li&gt;
&lt;/ul&gt;

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

&lt;p&gt;The project is fully open-source and built with React, Vite, and local open-weight AI integration:&lt;/p&gt;


&lt;div class="ltag-github-readme-tag"&gt;
  &lt;div class="readme-overview"&gt;
    &lt;h2&gt;
      &lt;img src="https://assets.dev.to/assets/github-logo-5a155e1f9a670af7944dd5e12375bc76ed542ea80224905ecaf878b9157cdefc.svg" alt="GitHub logo"&gt;
      &lt;a href="https://github.com/KavyaSaraswat23" rel="noopener noreferrer"&gt;
        KavyaSaraswat23
      &lt;/a&gt; / &lt;a href="https://github.com/KavyaSaraswat23/Hactober_week3" rel="noopener noreferrer"&gt;
        Hactober_week3
      &lt;/a&gt;
    &lt;/h2&gt;
    &lt;h3&gt;
      
    &lt;/h3&gt;
  &lt;/div&gt;
  &lt;div class="ltag-github-body"&gt;
    
&lt;div id="readme" class="md"&gt;&lt;div class="markdown-heading"&gt;
&lt;h1 class="heading-element"&gt;CraftList AI 🪵✨&lt;/h1&gt;
&lt;/div&gt;
&lt;blockquote&gt;
&lt;p&gt;&lt;strong&gt;Local E-Commerce Studio &amp;amp; Listing Engine Powered by Open-Weight Google Gemma 2&lt;/strong&gt;&lt;br&gt;
&lt;em&gt;Built for the Hacktoberfest Open-Source AI Challenge: Week 2 (Side Hustle)&lt;/em&gt;&lt;br&gt;
&lt;em&gt;Target Category: Featured Prize — Best Use of Gemma&lt;/em&gt;&lt;/p&gt;
&lt;/blockquote&gt;
&lt;p&gt;CraftList AI is a privacy-first, client-side listing optimizer and margin engine for Etsy, eBay, and Shopify artisans. It turns product materials, dimensions, and craft heritage notes into high-converting SEO listings, exactly 13 Etsy tags (≤ 20 characters), and social launch copy—running entirely offline with &lt;strong&gt;Google Gemma 2&lt;/strong&gt;.&lt;/p&gt;

&lt;div class="markdown-heading"&gt;
&lt;h2 class="heading-element"&gt;🚀 Key Features&lt;/h2&gt;
&lt;/div&gt;
&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Local Open-Weight AI Core:&lt;/strong&gt; Powered by Google's Gemma 2 (2B / 9B) running locally via Ollama with zero cloud API token costs.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Strict Marketplace Algorithmic Compliance:&lt;/strong&gt; Generates 140-char Etsy titles, 80-char eBay specs, and audits exactly 13 Etsy tags to ensure every tag is 20 characters or fewer with zero keyword duplicates.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Side Hustle Profit &amp;amp; Margin Calculator:&lt;/strong&gt; Computes exact material cost, labor hours…&lt;/li&gt;
&lt;/ul&gt;&lt;/div&gt;
  &lt;/div&gt;
  &lt;div class="gh-btn-container"&gt;&lt;a class="gh-btn" href="https://github.com/KavyaSaraswat23/Hactober_week3" rel="noopener noreferrer"&gt;View on GitHub&lt;/a&gt;&lt;/div&gt;
&lt;/div&gt;


&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;GitHub Repository:&lt;/strong&gt; &lt;a href="https://github.com/KavyaSaraswat23/Hactober_week3" rel="noopener noreferrer"&gt;https://github.com/KavyaSaraswat23/Hactober_week3&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Tech Stack:&lt;/strong&gt; React 19, Vite, Lucide Icons, Canvas-Confetti, Local Ollama / Google Gemma 2 API integration.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  How I Built It
&lt;/h2&gt;

&lt;p&gt;CraftList AI is architected around &lt;strong&gt;open-weight models and local inference&lt;/strong&gt;:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;strong&gt;Model of Choice: Google Gemma 2 (2B / 9B)&lt;/strong&gt;

&lt;ul&gt;
&lt;li&gt;Gemma 2 was chosen for its remarkable instruction-following capabilities, low memory footprint, and rapid generation speed on standard consumer laptops.&lt;/li&gt;
&lt;li&gt;The 2B variant (&lt;code&gt;gemma2:2b&lt;/code&gt;) runs blazingly fast even on CPU with zero dedicated GPU requirements, making it universally accessible to any non-technical maker.&lt;/li&gt;
&lt;/ul&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Turn-Token Prompt Formatting:&lt;/strong&gt;

&lt;ul&gt;
&lt;li&gt;We utilize Gemma's native turn structure:&lt;/li&gt;
&lt;/ul&gt;
&lt;/li&gt;
&lt;/ol&gt;



&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;
text
   &amp;lt;start_of_turn&amp;gt;user
   You are CraftList AI, an expert e-commerce listing optimizer.
   [CONSTRAINTS]
   - Strictly 13 tags, each &amp;lt;= 20 chars.
   - 140-char title limit for Etsy.
   - Target Price: $38.00 (Margin: 68%)
   &amp;lt;end_of_turn&amp;gt;
   &amp;lt;start_of_turn&amp;gt;model
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;

</description>
      <category>devchallenge</category>
      <category>hf26challenge</category>
      <category>gemma</category>
      <category>opensource</category>
    </item>
    <item>
      <title>"CraftList AI: Why Open Innovation Powers My 100% Local E-Commerce Listing Studio with Gemma 2"</title>
      <dc:creator>Kavya Saraswat</dc:creator>
      <pubDate>Mon, 19 Oct 2026 05:19:57 +0000</pubDate>
      <link>https://dev.to/kavya_saraswat_64c132105f/craftlist-ai-why-open-innovation-powers-my-100-local-e-commerce-listing-studio-with-gemma-2-55ge</link>
      <guid>https://dev.to/kavya_saraswat_64c132105f/craftlist-ai-why-open-innovation-powers-my-100-local-e-commerce-listing-studio-with-gemma-2-55ge</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;h2&gt;
  
  
  What I Built
&lt;/h2&gt;

&lt;p&gt;I built &lt;strong&gt;CraftList AI&lt;/strong&gt; — a privacy-first, zero-cloud e-commerce studio and profit margin engine designed for independent artisans, vintage thrifters, and handmade makers selling on Etsy, eBay, and Shopify.&lt;/p&gt;

&lt;h3&gt;
  
  
  Who Would Pay For It &amp;amp; Why?
&lt;/h3&gt;

&lt;p&gt;Solo makers and creative side-hustlers love crafting their products, but they universally hate writing listings. For every handmade mug, leather wallet, or carved wooden utensil, a seller must spend 45–60 minutes agonizing over:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;strong&gt;Algorithmic Titles &amp;amp; Descriptions:&lt;/strong&gt; Cramming high-volume search phrases into 140 characters without sounding robotic.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;The 13-Tag Etsy Matrix:&lt;/strong&gt; Generating exactly 13 unique tags under Etsy's strict rule of &lt;strong&gt;20 characters or fewer&lt;/strong&gt; per tag.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Pricing &amp;amp; Margin Anxiety:&lt;/strong&gt; Underpricing their labor or forgetting platform take-rates (6.5% transaction + payment processing + listing fees), accidentally working for less than minimum wage.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Multi-Channel Social Promotion:&lt;/strong&gt; Writing Instagram captions, Pinterest SEO pins, and TikTok hooks.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;Existing commercial listing generators charge $30–$50/month on subscription, meter tokens aggressively, and upload unreleased product designs to proprietary cloud servers. &lt;strong&gt;CraftList AI&lt;/strong&gt; runs &lt;strong&gt;100% locally on the seller's computer with Google Gemma 2&lt;/strong&gt;, costing $0.00 to run and guaranteeing complete data sovereignty.&lt;/p&gt;




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

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Live Application:&lt;/strong&gt; &lt;a href="http://localhost:5173" rel="noopener noreferrer"&gt;http://localhost:5173&lt;/a&gt; (Vite + React)&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Features Included:&lt;/strong&gt;

&lt;ul&gt;
&lt;li&gt;One-click sample presets (Hand-Carved Walnut Knife, Speckled Ceramic Mug, Horween Leather Wallet, Botanical Soy Candle).&lt;/li&gt;
&lt;li&gt;Multi-marketplace selector (Etsy, eBay, Shopify, Amazon Handmade).&lt;/li&gt;
&lt;li&gt;Built-in Side-Hustle Profit Calculator (Materials + Labor Hours × Rate + Marketplace Fees = Suggested Retail Price &amp;amp; Margin %).&lt;/li&gt;
&lt;li&gt;Real-time Tag Auditor (13 tags, character count counter, duplicate detection).&lt;/li&gt;
&lt;li&gt;Social Commerce Launch suite (Instagram, Pinterest, TikTok ASMR scripts).&lt;/li&gt;
&lt;li&gt;Bulk CSV and JSON export for instant store importing.&lt;/li&gt;
&lt;/ul&gt;
&lt;/li&gt;
&lt;/ul&gt;




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

&lt;p&gt;The code is open-source and structured with a modern React + Vite architecture any one can clone git repo and run it there own machine:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;GitHub Repository: &lt;a href="https://github.com/KavyaSaraswat23/Hactober_week3" rel="noopener noreferrer"&gt;https://github.com/KavyaSaraswat23/Hactober_week3&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;Tech Stack: React 19, Vite, Lucide Icons, Canvas-Confetti, Local Ollama / Gemma 2 API integration.&lt;/li&gt;
&lt;/ul&gt;




&lt;h2&gt;
  
  
  How I Built It
&lt;/h2&gt;

&lt;p&gt;CraftList AI is architected around &lt;strong&gt;open-weight models and local inference&lt;/strong&gt;:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;strong&gt;Model of Choice: Google Gemma 2 (2B / 9B)&lt;/strong&gt;

&lt;ul&gt;
&lt;li&gt;Gemma 2 was chosen for its remarkable instruction-following capabilities, low memory footprint, and rapid generation speed on standard consumer laptops.&lt;/li&gt;
&lt;li&gt;The 2B variant (&lt;code&gt;gemma2:2b&lt;/code&gt;) runs blazingly fast even on CPU with zero dedicated GPU requirements, making it universally accessible to any non-technical maker.&lt;/li&gt;
&lt;/ul&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Turn-Token Prompt Formatting:&lt;/strong&gt;

&lt;ul&gt;
&lt;li&gt;We utilize Gemma's native turn structure:
&lt;/li&gt;
&lt;/ul&gt;
&lt;/li&gt;
&lt;/ol&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;   &amp;lt;start_of_turn&amp;gt;user
   You are CraftList AI, an expert e-commerce listing optimizer.
   [CONSTRAINTS]
   - Strictly 13 tags, each &amp;lt;= 20 chars.
   - 140-char title limit for Etsy.
   - Profit Margin: 68%
   &amp;lt;end_of_turn&amp;gt;
   &amp;lt;start_of_turn&amp;gt;model
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;ol&gt;
&lt;li&gt;
&lt;strong&gt;Dual-Mode Inference Resilience:&lt;/strong&gt;

&lt;ul&gt;
&lt;li&gt;The app natively probes &lt;code&gt;http://localhost:11434&lt;/code&gt; for a live Ollama daemon running Gemma 2.&lt;/li&gt;
&lt;li&gt;If Ollama is offline or the user is on an unsupported device, a deterministic local client engine provides instant, high-quality listings with zero downtime or setup friction.&lt;/li&gt;
&lt;/ul&gt;
&lt;/li&gt;
&lt;/ol&gt;




&lt;h2&gt;
  
  
  Why Does Open Innovation Matter?
&lt;/h2&gt;

&lt;p&gt;Open-source AI isn't just an alternative to closed APIs like GPT-4 or Claude—for a side-hustle business, &lt;strong&gt;it is the only architecture that makes economic and operational sense&lt;/strong&gt;:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;strong&gt;Zero Marginal COGS ($0 vs $40/mo):&lt;/strong&gt;
A small maker selling ten $30 items a month cannot afford a $40/month software subscription. With open weights, their AI runs on the hardware they already own. Zero API keys, zero surprise credit card charges.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Protection of Artisan Trade Secrets:&lt;/strong&gt;
Makers invent secret wood finishing mixtures, ceramic glaze recipes, and proprietary silhouettes months before public launch. Sending raw sketches and formulas to a closed cloud API puts proprietary IP at risk of model retraining. Local Gemma 2 inference guarantees complete data privacy.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Rigid Rule Compliance Without Drift:&lt;/strong&gt;
Closed APIs update unpredictably and frequently ignore strict constraints like Etsy's "maximum 20 characters per tag." Open-weight models can be evaluated, controlled, and temperature-locked for deterministic compliance.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Complete Offline Autonomy:&lt;/strong&gt;
Artisans source materials at flea markets, work in barn studios, and sell at craft fairs where cell service is non-existent. CraftList AI works 100% offline on laptop battery power.&lt;/li&gt;
&lt;/ol&gt;




&lt;h2&gt;
  
  
  🌐 Community Wisdom
&lt;/h2&gt;

&lt;p&gt;Before building CraftList AI, we researched how fellow developers tackled local e-commerce and open-weight model integration on DEV:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;In &lt;a href="https://dev.to/yellowvoid/shelfscribe-local-ai-listing-writer-for-etsy-sellers-open-weight-ai-side-hustle-1hc5"&gt;ShelfScribe: Local AI Listing Writer for Etsy Sellers&lt;/a&gt; by &lt;a class="mentioned-user" href="https://dev.to/yellowvoid"&gt;@yellowvoid&lt;/a&gt;, the author demonstrated how small open-weight models easily outperform generic assistants when given tight domain constraints.&lt;/li&gt;
&lt;li&gt;In &lt;a href="https://dev.to/cooldashing24/ollama-cheat-sheet-local-llms-models-api-integration-2026-5a0f"&gt;Ollama Cheat Sheet: Local LLMs, Models, API &amp;amp; Integration (2026)&lt;/a&gt; by &lt;a class="mentioned-user" href="https://dev.to/cooldashing24"&gt;@cooldashing24&lt;/a&gt;, community discussions highlighted the importance of structuring prompt turn-tokens specifically for Gemma 2 to prevent token hallucination and ensure strict JSON schema outputs.&lt;/li&gt;
&lt;/ul&gt;




&lt;h2&gt;
  
  
  My Agent Session
&lt;/h2&gt;

&lt;p&gt;This project was built and debugged pair-programming with Antigravity and DevRelay:&lt;/p&gt;

&lt;p&gt;&lt;code&gt;{% agent_session b07531a1-e075-41f2-bb93-ce0b43b1ab87 %}&lt;/code&gt;&lt;/p&gt;




&lt;h2&gt;
  
  
  My First User
&lt;/h2&gt;

&lt;p&gt;I showed CraftList AI to my friend Maya, an independent ceramist who sells speckled stoneware mugs and planter pots at weekend pop-up markets. &lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Her Reaction:&lt;/strong&gt; &lt;em&gt;"Wait, I don't have to think of 13 tags myself? And it counted the characters so Etsy doesn't reject them in red text?!"&lt;/em&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;The Test:&lt;/strong&gt; We plugged in her 14 oz speckled mug, material cost ($4.80), and labor time (50 mins). The calculator instantly showed she was undercharging at $26 (barely 1.8x markup). CraftList AI recommended $38 based on her $26/hr target rate and platform fees, and generated an aesthetic story description highlighting her hand-pulled three-finger handle.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Did She Pay / Would She?&lt;/strong&gt; She offered to buy me lunch on the spot to keep the tool permanently on her MacBook, saying: &lt;em&gt;"I would easily pay a $29 one-time download fee for this because I refuse to pay another monthly subscription."&lt;/em&gt;
&lt;/li&gt;
&lt;/ul&gt;




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

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Featured Category: Best Use of Gemma&lt;/strong&gt; (Built specifically for Google Gemma 2 2B/9B open weights running locally via Ollama with strict turn-token conditioning).&lt;/li&gt;
&lt;/ul&gt;

</description>
      <category>devchallenge</category>
      <category>hf26challenge</category>
    </item>
  </channel>
</rss>
