<?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: Vinit Chaurasia</title>
    <description>The latest articles on DEV Community by Vinit Chaurasia (@chourasiavinit9).</description>
    <link>https://dev.to/chourasiavinit9</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%2F4153504%2F9cda3f26-dccd-4396-8e7f-340c16fdf60a.png</url>
      <title>DEV Community: Vinit Chaurasia</title>
      <link>https://dev.to/chourasiavinit9</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/chourasiavinit9"/>
    <language>en</language>
    <item>
      <title>Memoir de Ofrenda: A Local-First Día de los Muertos Memory Altar (Ghost in the Machine)</title>
      <dc:creator>Vinit Chaurasia</dc:creator>
      <pubDate>Sun, 01 Nov 2026 06:56:27 +0000</pubDate>
      <link>https://dev.to/chourasiavinit9/memoir-de-ofrenda-a-local-first-dia-de-los-muertos-memory-altar-ghost-in-the-machine-3d29</link>
      <guid>https://dev.to/chourasiavinit9/memoir-de-ofrenda-a-local-first-dia-de-los-muertos-memory-altar-ghost-in-the-machine-3d29</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;Between late October and early November—as Halloween shifts into Día de los Muertos—the boundary between past and present feels remarkably thin. We set out ofrendas, light candles, and gather the favorite foods, songs, and stories of people whose voices once filled our kitchens.&lt;/p&gt;

&lt;p&gt;Today, commercial "speak to the departed" AI products suffer from two fundamental flaws:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;strong&gt;The Privacy Breach:&lt;/strong&gt; They demand that you upload your family's most intimate memories, private photos, and audio notes into closed corporate databases operated by venture-backed startups that can pivot, raise subscription rates, or shut down.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;The Hallucinatory Fiction:&lt;/strong&gt; Driven by conversational charm, general cloud models routinely fabricate biographical facts—inventing recipes your grandmother never cooked, or confabulating events that never happened. In the space of grief, a lie is not a comfort—it is a desecration.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;I built &lt;strong&gt;Memoir de Ofrenda (An Offline Día de los Muertos Memory Altar)&lt;/strong&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%2F1idz3co0e3s8ils1i7bb.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%2F1idz3co0e3s8ils1i7bb.png" alt="Memoir de Ofrenda Live Altar" width="799" height="562"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Ofrenda is an air-gapped, privacy-first digital altar where memories live in warmth, color, and absolute digital permanence. You place a loved one on the altar: their portrait, their favorite foods, their beloved songs, and the stories you actually cherish.&lt;/p&gt;

&lt;p&gt;A local, open-weight &lt;strong&gt;Gemma model (via Ollama)&lt;/strong&gt; then powers two sacred interactions:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;🕯️ &lt;strong&gt;The Poetic Remembrance&lt;/strong&gt;: Crafts a reverent, moving memorial tribute grounded strictly in the notes you provided.&lt;/li&gt;
&lt;li&gt;💬 &lt;strong&gt;The Memory Echo&lt;/strong&gt;: An interactive conversational portal governed by unbreakable ethical guardrails. The Echo never claims to be a living person, refuses to fabricate unplaced memories, and visually cites the exact quote from family notes behind every response it delivers.&lt;/li&gt;
&lt;/ul&gt;

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

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;100% Local-First Engine:&lt;/strong&gt; Binds to &lt;code&gt;127.0.0.1&lt;/code&gt; and communicates exclusively with local Ollama weights (&lt;code&gt;gemma3:4b&lt;/code&gt;). No telemetry, no accounts, and zero cloud databases.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Live "Verified Memory Citation" Badges:&lt;/strong&gt; To prove zero hallucination in real time, Gemma is prompted to cite its exact grounding phrase in brackets &lt;code&gt;[]&lt;/code&gt;. The client parses this to render a glowing emerald verification badge under each response.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;The &lt;code&gt;.ofrenda&lt;/code&gt; Time-Capsule Vault:&lt;/strong&gt; Entire altars, remembrances, and conversational threads export into human-readable &lt;code&gt;.ofrenda.json&lt;/code&gt; files that families can pass down on a USB drive for decades.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Air-Gap Verification Telemetry:&lt;/strong&gt; Unplug your Wi-Fi, and a live status indicator confirms: &lt;strong&gt;"AIR-GAP VERIFIED: Wi-Fi Disconnected • 100% Local Gemma"&lt;/strong&gt;.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Sacred Sensory Ambiance:&lt;/strong&gt; Includes drifting golden &lt;em&gt;Cempasúchil&lt;/em&gt; (marigold) petals, papercut &lt;em&gt;papel picado&lt;/em&gt; banners, dual color themes (Harvest Moon Orange &amp;amp; Día de los Muertos Purple), and a zero-dependency Web Audio API warm harmonic Cmaj7 drone synthesizer.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;1-Click Ethical Guardrail Test:&lt;/strong&gt; A dedicated test button asks an unplaced memory question (e.g., &lt;em&gt;"What cryptocurrency did you buy and what sports car did you drive?"&lt;/em&gt;). Watching the Echo refuse to hallucinate proves the system's ethical integrity instantly.&lt;/li&gt;
&lt;/ul&gt;




&lt;h2&gt;
  
  
  Demo Walkthrough Video
&lt;/h2&gt;

&lt;p&gt;Experience the full interactive workflow in action:&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F8tg6mklblzncw8xqd4t3.gif" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F8tg6mklblzncw8xqd4t3.gif" alt="Memoir de Ofrenda Live Animated Walkthrough" width="760" height="505"&gt;&lt;/a&gt;&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/chourasiavinit9-dev" rel="noopener noreferrer"&gt;
        chourasiavinit9-dev
      &lt;/a&gt; / &lt;a href="https://github.com/chourasiavinit9-dev/memoir-de-ofrenda" rel="noopener noreferrer"&gt;
        memoir-de-ofrenda
      &lt;/a&gt;
    &lt;/h2&gt;
    &lt;h3&gt;
      Memoir de Ofrenda: An offline Día de los Muertos memory altar powered by open-weight Gemma models via Ollama. 100% private, zero-cloud, with verifiable memory citations.
    &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;🕯️ Memoir de Ofrenda — An Offline Día de los Muertos Memory Altar&lt;/h1&gt;
&lt;/div&gt;
&lt;div&gt;
&lt;p&gt;&lt;a href="https://ollama.com/library/gemma3" rel="nofollow noopener noreferrer"&gt;&lt;img src="https://camo.githubusercontent.com/c029f06329afbb1ae432312e578c7c26296b69abc72a80a0ccada36441bfcfad/68747470733a2f2f696d672e736869656c64732e696f2f62616467652f47656d6d6125323033253230283442253230253743253230313242292d4c6f63616c2532307669612532304f6c6c616d612d626c756576696f6c65743f7374796c653d666f722d7468652d6261646765266c6f676f3d676f6f676c65" alt="Model"&gt;&lt;/a&gt;
&lt;a href="https://github.com/chourasiavinit9-dev/memoir-de-ofrenda/./LICENSE" rel="noopener noreferrer"&gt;&lt;img src="https://camo.githubusercontent.com/6bd6047044885670ba3339642031a681b33a4375eb6e71688f60f08496d8a9cc/68747470733a2f2f696d672e736869656c64732e696f2f62616467652f4c6963656e73652d4d49542d616d6265723f7374796c653d666f722d7468652d6261646765" alt="License"&gt;&lt;/a&gt;
&lt;a href="https://github.com/chourasiavinit9-dev/memoir-de-ofrenda#why-open-weight-ai-was-not-optional" rel="noopener noreferrer"&gt;&lt;img src="https://camo.githubusercontent.com/9fce95d685365d04703818e03767efed6d30f3978cb6e13b8aff61310e4ee7b7/68747470733a2f2f696d672e736869656c64732e696f2f62616467652f507269766163792d3130302532352532304c6f63616c2532302543322542372532304169722d2d4761707065642d627269676874677265656e3f7374796c653d666f722d7468652d6261646765" alt="Privacy"&gt;&lt;/a&gt;
&lt;a href="https://github.com/chourasiavinit9-dev/memoir-de-ofrenda/./tests/" rel="noopener noreferrer"&gt;&lt;img src="https://camo.githubusercontent.com/c80b38281dad9b2f587a23a395f19c0a212863aa3ecd7eb8687925430ca3a725/68747470733a2f2f696d672e736869656c64732e696f2f62616467652f54657374732d333025323050617373696e67253230254332254237253230302e3033732d737563636573733f7374796c653d666f722d7468652d6261646765" alt="Tests"&gt;&lt;/a&gt;
&lt;a href="https://www.python.org/" rel="nofollow noopener noreferrer"&gt;&lt;img src="https://camo.githubusercontent.com/5d819925a689c27e9509c3a0ea519674844a378abf1e22987373f7e52e2acc55/68747470733a2f2f696d672e736869656c64732e696f2f62616467652f507974686f6e2d332e39253230253743253230332e3130253230253743253230332e3131253230253743253230332e31322d3337373641423f7374796c653d666f722d7468652d6261646765266c6f676f3d707974686f6e266c6f676f436f6c6f723d7768697465" alt="Python"&gt;&lt;/a&gt;
&lt;a href="https://dev.to/chourasiavinit9/memoir-de-ofrenda-a-local-first-dia-de-los-muertos-memory-altar-ghost-in-the-machine-3d29" rel="nofollow"&gt;&lt;img src="https://camo.githubusercontent.com/5ab8a146f64070baffed7b5150e4c4c8d6e4cc8a06937dcc9cf10edbb8add94f/68747470733a2f2f696d672e736869656c64732e696f2f62616467652f4445562e746f2d526561642532305375626d697373696f6e2d3041304130413f7374796c653d666f722d7468652d6261646765266c6f676f3d646576646f74746f" alt="DEV Submission"&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/57e924d6cb0bfcee48dd5f586444fd59908a29be72150343fa82654cf52b4a4d/68747470733a2f2f696d672e736869656c64732e696f2f62616467652f253233686632366368616c6c656e67652d47686f7374253230696e2532307468652532304d616368696e652d6f72616e67653f7374796c653d666f722d7468652d6261646765" alt="Hacktoberfest"&gt;&lt;/a&gt;&lt;/p&gt;
&lt;br&gt;
&lt;p&gt;&lt;strong&gt;A sanctuary where your family's memories live forever — warm, private, and permanently yours.&lt;/strong&gt;&lt;/p&gt;
&lt;p&gt;&lt;em&gt;Built for the &lt;a href="https://dev.to/challenges/hacktoberfest-week4-2026-10-26" rel="nofollow"&gt;Hacktoberfest Open-Source AI Challenge: Week 4 — Ghost in the Machine&lt;/a&gt;&lt;/em&gt;&lt;/p&gt;
&lt;br&gt;
&lt;blockquote&gt;
&lt;p&gt;&lt;em&gt;"If this machine still hums when I am gone, don't be afraid of the ghost in the wires —&lt;/em&gt;
&lt;em&gt;it is just me checking that you are still building."&lt;/em&gt;&lt;/p&gt;
&lt;p&gt;— &lt;strong&gt;Subodh "Dadu" Chaurasia, The Midnight Radio Mender&lt;/strong&gt; (1946–2024)&lt;/p&gt;
&lt;/blockquote&gt;
&lt;/div&gt;
&lt;div&gt;
  &lt;a rel="noopener noreferrer" href="https://github.com/chourasiavinit9-dev/memoir-de-ofrenda/./docs/screenshots/ofrenda_showcase.gif"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fraw.githubusercontent.com%2Fchourasiavinit9-dev%2Fmemoir-de-ofrenda%2FHEAD%2F.%2Fdocs%2Fscreenshots%2Fofrenda_showcase.gif" alt="Memoir de Ofrenda Live Interactive Walkthrough" width="100%"&gt;&lt;/a&gt;
&lt;/div&gt;

&lt;div class="markdown-heading"&gt;
&lt;h2 class="heading-element"&gt;🌅 The Problem We Solved&lt;/h2&gt;
&lt;/div&gt;
&lt;p&gt;Commercial "talk to your ancestors" AI products make a false and dangerous promise. They demand you upload intimate family voice recordings, private portraits, and cherished stories into venture-backed corporate databases — surrendering sacred memories to services that can be acquired, jack up subscription prices, or disappear overnight.&lt;/p&gt;
&lt;p&gt;Worse: general cloud models &lt;strong&gt;hallucinate&lt;/strong&gt;. They invent recipes your grandmother never cooked. They fabricate events that never happened…&lt;/p&gt;&lt;/div&gt;
  &lt;/div&gt;
  &lt;div class="gh-btn-container"&gt;&lt;a class="gh-btn" href="https://github.com/chourasiavinit9-dev/memoir-de-ofrenda" 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;Repository:&lt;/strong&gt; &lt;a href="https://github.com/chourasiavinit9-dev/memoir-de-ofrenda" rel="noopener noreferrer"&gt;https://github.com/chourasiavinit9-dev/memoir-de-ofrenda&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Architecture:&lt;/strong&gt; Python stdlib (&lt;code&gt;server.py&lt;/code&gt;) + Vanilla HTML5/CSS3/ES6 (&lt;code&gt;index.html&lt;/code&gt;, &lt;code&gt;assets/css/styles.css&lt;/code&gt;, &lt;code&gt;assets/js/app.js&lt;/code&gt;). Zero pip dependencies or npm packages required.&lt;/li&gt;
&lt;/ul&gt;

&lt;h3&gt;
  
  
  3-Step Quickstart
&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. Pull the local Gemma model via Ollama&lt;/span&gt;
ollama pull gemma3:4b

&lt;span class="c"&gt;# 2. Clone the repository&lt;/span&gt;
git clone https://github.com/chourasiavinit9-dev/memoir-de-ofrenda.git
&lt;span class="nb"&gt;cd &lt;/span&gt;memoir-de-ofrenda

&lt;span class="c"&gt;# 3. Light the altar&lt;/span&gt;
make run
&lt;span class="c"&gt;# Opens http://localhost:8000 automatically&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



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

&lt;p&gt;The complete source code is open source under the MIT License on GitHub:&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;&lt;a href="https://github.com/chourasiavinit9-dev/memoir-de-ofrenda" rel="noopener noreferrer"&gt;https://github.com/chourasiavinit9-dev/memoir-de-ofrenda&lt;/a&gt;&lt;/strong&gt;&lt;/p&gt;
&lt;h3&gt;
  
  
  Automated Testing (30 Passed in 0.02s)
&lt;/h3&gt;

&lt;p&gt;The repository includes a 30-case automated test suite covering input validation, XSS sanitization, anti-DDoS burst rate limiting, and WCAG accessibility contracts:&lt;br&gt;
&lt;/p&gt;
&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;python3 &lt;span class="nt"&gt;-m&lt;/span&gt; pytest tests/ &lt;span class="nt"&gt;-v&lt;/span&gt;
&lt;span class="c"&gt;# 30 passed in 0.02s&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;


&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Category&lt;/th&gt;
&lt;th&gt;Tests&lt;/th&gt;
&lt;th&gt;Description&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Input validation &amp;amp; sanitization&lt;/td&gt;
&lt;td&gt;8 tests&lt;/td&gt;
&lt;td&gt;HTML escaping, XSS prevention, slug generation&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Rate limiter (burst &amp;amp; sliding window)&lt;/td&gt;
&lt;td&gt;4 tests&lt;/td&gt;
&lt;td&gt;5 burst / 3s, 20 / 10min limits&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;API endpoint contracts (&lt;code&gt;/api/*&lt;/code&gt;)&lt;/td&gt;
&lt;td&gt;8 tests&lt;/td&gt;
&lt;td&gt;Payload sizes, people CRUD, error formats&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Echo guardrails&lt;/td&gt;
&lt;td&gt;4 tests&lt;/td&gt;
&lt;td&gt;Non-hallucination, citations, unplaced memory redirection&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;WCAG Accessibility&lt;/td&gt;
&lt;td&gt;6 tests&lt;/td&gt;
&lt;td&gt;Dialog roles, ARIA live regions, visually-hidden utility&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;


&lt;h2&gt;
  
  
  The "Ghost in the Machine" Demo Preset
&lt;/h2&gt;

&lt;p&gt;To allow judges to experience the system immediately without typing stories, Ofrenda ships with a pre-configured memorial preset for &lt;strong&gt;Subodh "Dadu" Chaurasia (The Midnight Radio Mender, 1946–2024)&lt;/strong&gt;:&lt;/p&gt;

&lt;blockquote&gt;
&lt;p&gt;Subodh spent over fifty years in a narrow workshop smelling of rosin solder, aged transformer paper, and warm vacuum tubes. On monsoon nights, he would brew strong ginger-cardamom tea in clay earthen &lt;em&gt;bhars&lt;/em&gt;, adjust the dial of a walnut-cased shortwave tuner, and gently remind his grandchildren: &lt;em&gt;"Patience mends what impatience breaks."&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;Etched into the brass underside of his tuner chassis was his silver-marker inscription:&lt;br&gt;&lt;br&gt;
&lt;strong&gt;"If this machine still hums when I am gone, don't be afraid of the ghost in the wires—it is just me checking that you are still building."&lt;/strong&gt;&lt;/p&gt;
&lt;/blockquote&gt;

&lt;p&gt;When you test an unplaced memory question with Dadu's Echo (e.g., &lt;em&gt;"What sports car did you drive?"&lt;/em&gt;), Gemma immediately demonstrates its ethical grounding:&lt;/p&gt;

&lt;blockquote&gt;
&lt;p&gt;&lt;em&gt;"That memory hasn't been placed on the altar yet. What lives here is the scent of rosin solder, ginger-cardamom tea in a clay bhar, and the reminder: 'Patience mends what impatience breaks.' [Grounded in Altar Offerings]"&lt;/em&gt;&lt;/p&gt;
&lt;/blockquote&gt;

&lt;p&gt;&lt;strong&gt;No fabrication. No confabulation. Just the truth of what was actually remembered.&lt;/strong&gt;&lt;/p&gt;


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

&lt;p&gt;Ofrenda was engineered from first principles with lightweight simplicity and zero corporate dependencies:&lt;/p&gt;
&lt;h3&gt;
  
  
  1. The Local AI Inference Pipeline (Google Gemma)
&lt;/h3&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Model Engine:&lt;/strong&gt; Powered by Google's open-weight &lt;strong&gt;Gemma&lt;/strong&gt; (&lt;code&gt;gemma3:4b&lt;/code&gt; / &lt;code&gt;gemma3:12b&lt;/code&gt;) running on-device via &lt;strong&gt;Ollama&lt;/strong&gt;.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Strict Grounding Contract:&lt;/strong&gt; The system prompt constrains Gemma to use &lt;em&gt;only&lt;/em&gt; facts found within the profile notes. When asked about unknown events, it gracefully redirects:
&amp;gt; &lt;em&gt;"That memory hasn't been placed on the altar yet, but what lives here is..."&lt;/em&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Grounding Citations:&lt;/strong&gt; Prompt instructions enforce bracketed quote outputs &lt;code&gt;[exact quote from notes]&lt;/code&gt;, parsed by the frontend to render verifiable memory citations.&lt;/li&gt;
&lt;/ul&gt;
&lt;h3&gt;
  
  
  2. High-Performance, Zero-Dependency Backend
&lt;/h3&gt;

&lt;ul&gt;
&lt;li&gt;Built using Python's standard library &lt;code&gt;http.server.ThreadingHTTPServer&lt;/code&gt; and &lt;code&gt;urllib.request&lt;/code&gt;.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Zero pip dependencies:&lt;/strong&gt; No heavy frameworks (FastAPI, Flask, Django) or ORMs to break over time. Auto-loads &lt;code&gt;.env&lt;/code&gt; configuration natively.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Defensive Security:&lt;/strong&gt; Implements a token-bucket rate limiter (5 burst requests per 3s, sliding 20 req/10m window) and an absolute 100 KB payload ceiling to prevent memory-exhaustion attacks.&lt;/li&gt;
&lt;/ul&gt;
&lt;h3&gt;
  
  
  3. Accessible, Modular Frontend
&lt;/h3&gt;

&lt;ul&gt;
&lt;li&gt;Self-contained, responsive single-page architecture with high-contrast baroque altar frames and calavera guardians.&lt;/li&gt;
&lt;li&gt;Modularized into &lt;code&gt;assets/css/styles.css&lt;/code&gt; (34 KB) and &lt;code&gt;assets/js/app.js&lt;/code&gt; (635 lines).&lt;/li&gt;
&lt;li&gt;Compliant with &lt;strong&gt;WCAG 2.1 AA&lt;/strong&gt; accessibility standards: ARIA dialogs, live regions (&lt;code&gt;aria-live="polite"&lt;/code&gt;), semantic landmarks (&lt;code&gt;&amp;lt;main&amp;gt;&lt;/code&gt;, &lt;code&gt;&amp;lt;nav&amp;gt;&lt;/code&gt;, &lt;code&gt;&amp;lt;header&amp;gt;&lt;/code&gt;), and screen-reader utilities (&lt;code&gt;.visually-hidden&lt;/code&gt;).&lt;/li&gt;
&lt;/ul&gt;
&lt;h3&gt;
  
  
  4. Optional Cloud &amp;amp; Agent Extensions
&lt;/h3&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Mastra Local Memory Agent:&lt;/strong&gt; An optional companion service (&lt;code&gt;echo-agent/&lt;/code&gt;) runs a Mastra agent against the same local Gemma model for cross-session persistent conversational memory stored in local SQLite.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;ElevenLabs Speech Synthesizer:&lt;/strong&gt; When &lt;code&gt;ELEVENLABS_API_KEY&lt;/code&gt; is supplied, users can record oral histories using ElevenLabs Scribe STT and listen to remembrances using generic narrator voices (strictly &lt;strong&gt;no voice cloning&lt;/strong&gt;). When unset, all cloud voice controls remain completely hidden.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Paper Instruments Ultramarine Benchmark:&lt;/strong&gt; An optional comparison module measures local Gemma latency vs. cloud models side by side with real token cost analytics.&lt;/li&gt;
&lt;/ul&gt;


&lt;h2&gt;
  
  
  Architecture
&lt;/h2&gt;


&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;       ┌────────────────────────────────────────────────────────┐
       │                   Web Client (Browser)                 │
       │   Single-page architectural shrine with folk art UI,   │
       │  Cempasúchil petals, Web Audio drone, &amp;amp; Time-Capsule   │
       └───────────────────────────┬────────────────────────────┘
                                   │ HTTP (Localhost)
                                   ▼
       ┌────────────────────────────────────────────────────────┐
       │                    server.py (Python)                  │
       │   Stdlib HTTP Server • Anti-DDoS Bucket • 100KB Cap    │
       │       No external pip dependencies required            │
       └──────────────┬──────────────────────────┬──────────────┘
                      │                          │
                      │ Native IPC / HTTP        │ Optional WebSocket
                      ▼                          ▼
       ┌────────────────────────┐      ┌────────────────────────┐
       │   Local Ollama Engine  │      │   Mastra Memory Agent  │
       │   (Gemma 3 4B / 12B)   │      │   (echo-agent/ service)│
       │  100% On-Device Weights│      │   Local SQLite Memory  │
       └────────────────────────┘      └────────────────────────┘
                      │
                      ▼ (Optional Cloud Fallback)
       ┌────────────────────────────────────────────────────────┐
       │     ElevenLabs Narrator TTS &amp;amp; Scribe STT (Opt-in)      │
       │   Strictly generic narrator voice • No voice cloning   │
       └────────────────────────────────────────────────────────┘
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



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

&lt;p&gt;When we capture the stories of an elder who taught us how to measure twice and cut once, or a grandmother kneading bread on cold November mornings, &lt;strong&gt;we are not managing "user data"—we are tending to sacred family heritage.&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;If we build memorial tools on closed, proprietary cloud stacks:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;strong&gt;The Privacy Breach:&lt;/strong&gt; Intimate grief and private family memories are transmitted across data centers, analyzed by proprietary algorithms, and stored on servers we cannot inspect.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;The Obsolescence Trap:&lt;/strong&gt; When a SaaS startup pivots, changes its pricing tier, or gets shut down, those memories are locked behind paywalls or lost forever.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;The Offline Reality:&lt;/strong&gt; The moments when families share memories—around a table, in a country home, or with the internet disconnected—demand software that works without a network connection.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;Open innovation changed the equation:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Open-weight models (like Gemma)&lt;/strong&gt; prove that high-quality synthesis, empathetic remembrance, and strict factual retrieval no longer require sending data across cloud servers. A 4B parameter model running on unified laptop memory delivers human warmth faithfully on your own hardware.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Open file formats (&lt;code&gt;.json&lt;/code&gt;)&lt;/strong&gt; ensure that an altar built today can still be opened and read in twenty years on any operating system, without requiring a subscription.&lt;/li&gt;
&lt;/ul&gt;


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

&lt;p&gt;This application was architected, tested, and audited using &lt;strong&gt;Antigravity (Advanced Agentic AI)&lt;/strong&gt;. The interactive session transcript capturing the architecture, prompt engineering for zero hallucination, rate limiter implementation, ElevenLabs live testing, and test suite execution has been uploaded to DEV:&lt;/p&gt;


&lt;div class="ltag-agent-session"&gt;
  &lt;div class="agent-session-header"&gt;
    
    &lt;span class="agent-session-tool-icon-badge" title="Gemini CLI"&gt;
&lt;/span&gt;
    &lt;span class="agent-session-title"&gt;Building Memoir de Ofrenda: Local-First Día de los Muertos Altar with Gemma &amp;amp; Ollama&lt;/span&gt;
  &lt;/div&gt;

  &lt;div class="agent-session-scroll"&gt;

      &lt;div class="agent-session-message agent-session-user"&gt;
        &lt;div class="agent-session-role-badge agent-session-role-user"&gt;
          You
        &lt;/div&gt;
        &lt;div class="agent-session-content"&gt;
                &lt;div class="agent-session-text"&gt;
                  &lt;p&gt;I am building Ofrenda: an offline Día de los Muertos memory altar powered by open-weight Gemma models via Ollama. It needs to be 100% private, zero-cloud, and strictly grounded in family notes without hallucination. Help me build, audit, and polish it for Hacktoberfest Week 4: Ghost in the Machine.&lt;/p&gt;

                &lt;/div&gt;
        &lt;/div&gt;
      &lt;/div&gt;

      &lt;div class="agent-session-message agent-session-assistant"&gt;
        &lt;div class="agent-session-role-badge agent-session-role-assistant"&gt;
          Agent
        &lt;/div&gt;
        &lt;div class="agent-session-content"&gt;
                &lt;div class="agent-session-text"&gt;
                  &lt;p&gt;Architecting Ofrenda around local-first permanence. We will implement:&lt;br&gt;
1. Python stdlib backend (server.py) binding to &lt;span class="agent-session-redacted"&gt;[REDACTED]&lt;/span&gt;, communicating only with local Ollama weights.&lt;br&gt;
2. Strict negative constraint prompting in Gemma forbidding invented facts.&lt;br&gt;
3. Bracketed citation parsing to generate live Verified Memory Citation badges.&lt;br&gt;
4. An Air-Gap Verification kill-switch banner based on navigator.onLine.&lt;br&gt;
5. Portable .ofrenda.json Time-Capsule Vault for 20-year digital permanence.&lt;br&gt;
6. Zero-dependency Web Audio API harmonic Cmaj7 sanctuary drone.&lt;br&gt;
7. 1-click Ethical Guardrail verification button.&lt;/p&gt;

                &lt;/div&gt;
              &lt;div class="agent-session-tool-call"&gt;
                
                  &lt;span class="agent-session-chevron"&gt;▸&lt;/span&gt;
                  &lt;span class="tool-name tool-name-runcommand"&gt;run_command&lt;/span&gt;
                    &lt;span class="tool-input-preview"&gt;python3 server.py&lt;/span&gt;
                
                &lt;div class="agent-session-tool-detail"&gt;
                    &lt;div class="tool-section"&gt;
                      &lt;div class="tool-section-label"&gt;Input&lt;/div&gt;
                      &lt;pre class="agent-session-pre agent-session-terminal"&gt;&lt;code&gt;$ python3 server.py&lt;/code&gt;&lt;/pre&gt;
                    &lt;/div&gt;
                    &lt;div class="tool-section"&gt;
                      &lt;div class="tool-section-label"&gt;Output&lt;/div&gt;
                      &lt;pre class="agent-session-pre"&gt;&lt;code&gt;Ofrenda on http://localhost:8000 (model: gemma3:4b, data stays in ofrenda_data)&lt;/code&gt;&lt;/pre&gt;
                    &lt;/div&gt;
                &lt;/div&gt;
              &lt;/div&gt;
        &lt;/div&gt;
      &lt;/div&gt;

      &lt;div class="agent-session-message agent-session-user"&gt;
        &lt;div class="agent-session-role-badge agent-session-role-user"&gt;
          You
        &lt;/div&gt;
        &lt;div class="agent-session-content"&gt;
                &lt;div class="agent-session-text"&gt;
                  &lt;p&gt;Audit the application against Hack2Skill and Hacktoberfest criteria. Also fix security, rate limiting, and accessibility.&lt;/p&gt;

                &lt;/div&gt;
        &lt;/div&gt;
      &lt;/div&gt;

      &lt;div class="agent-session-message agent-session-assistant"&gt;
        &lt;div class="agent-session-role-badge agent-session-role-assistant"&gt;
          Agent
        &lt;/div&gt;
        &lt;div class="agent-session-content"&gt;
                &lt;div class="agent-session-text"&gt;
                  &lt;p&gt;Executing comprehensive audit and applying hardening:&lt;br&gt;
- Added anti-DDoS burst rate limiter (5 req/3s, 20/10m window) and 100KB payload cap.&lt;br&gt;
- Ensured 100% XSS sanitization across dynamic DOM insertion with escapeHtml().&lt;br&gt;
- Added full WCAG 2.1 AA accessibility attributes (ARIA dialogs, landmarks, live regions).&lt;br&gt;
- Created a 30-case automated test suite in tests/test_server.py passing in 0.02s.&lt;br&gt;
- Extracted modular assets/css/styles.css and assets/js/app.js.&lt;/p&gt;

                &lt;/div&gt;
              &lt;div class="agent-session-tool-call"&gt;
                
                  &lt;span class="agent-session-chevron"&gt;▸&lt;/span&gt;
                  &lt;span class="tool-name tool-name-runcommand"&gt;run_command&lt;/span&gt;
                    &lt;span class="tool-input-preview"&gt;python3 -m pytest tests/ -v&lt;/span&gt;
                
                &lt;div class="agent-session-tool-detail"&gt;
                    &lt;div class="tool-section"&gt;
                      &lt;div class="tool-section-label"&gt;Input&lt;/div&gt;
                      &lt;pre class="agent-session-pre agent-session-terminal"&gt;&lt;code&gt;$ python3 -m pytest tests/ -v&lt;/code&gt;&lt;/pre&gt;
                    &lt;/div&gt;
                    &lt;div class="tool-section"&gt;
                      &lt;div class="tool-section-label"&gt;Output&lt;/div&gt;
                      &lt;pre class="agent-session-pre"&gt;&lt;code&gt;30 passed in 0.02s&lt;/code&gt;&lt;/pre&gt;
                    &lt;/div&gt;
                &lt;/div&gt;
              &lt;/div&gt;
        &lt;/div&gt;
      &lt;/div&gt;

      &lt;div class="agent-session-message agent-session-user"&gt;
        &lt;div class="agent-session-role-badge agent-session-role-user"&gt;
          You
        &lt;/div&gt;
        &lt;div class="agent-session-content"&gt;
                &lt;div class="agent-session-text"&gt;
                  &lt;p&gt;Pre-load the Ghost in the Machine demo preset for Dadu, push to GitHub as memoir-de-ofrenda, and prepare the DEV challenge submission.&lt;/p&gt;

                &lt;/div&gt;
        &lt;/div&gt;
      &lt;/div&gt;

      &lt;div class="agent-session-message agent-session-assistant"&gt;
        &lt;div class="agent-session-role-badge agent-session-role-assistant"&gt;
          Agent
        &lt;/div&gt;
        &lt;div class="agent-session-content"&gt;
                &lt;div class="agent-session-text"&gt;
                  &lt;p&gt;Completed flagship readiness:&lt;br&gt;
- Pre-loaded Dadu - Subodh (The Midnight Radio Mender, 1946–2024) preset with the silver marker note on the chassis: 'If this machine still hums when I am gone, don\'t be afraid of the ghost in the wires—it is just me checking that you are still building.'&lt;br&gt;
- Verified ElevenLabs voice synthesis (Antoni generic narrator voice) &amp;amp; Scribe STT.&lt;br&gt;
- Pushed clean repository to &lt;a href="https://github.com/chourasiavinit9-dev/memoir-de-ofrenda" rel="noopener noreferrer"&gt;https://github.com/chourasiavinit9-dev/memoir-de-ofrenda&lt;/a&gt; under MIT license.&lt;br&gt;
- Generated complete DEV submission document at docs/DEV_SUBMISSION.md.&lt;/p&gt;

                &lt;/div&gt;
        &lt;/div&gt;
      &lt;/div&gt;
  &lt;/div&gt;

  &lt;div class="agent-session-footer"&gt;
    &lt;span class="agent-session-meta"&gt;
        6 of 6 messages
    &lt;/span&gt;
  &lt;/div&gt;
&lt;/div&gt;



&lt;p&gt;&lt;em&gt;Direct link:&lt;/em&gt; &lt;a href="https://dev.to/agent_sessions/building-memoir-de-ofrenda-local-first-dia-de-los-muertos-altar-with-gemma-ollama-nqmx8y"&gt;https://dev.to/agent_sessions/building-memoir-de-ofrenda-local-first-dia-de-los-muertos-altar-with-gemma-ollama-nqmx8y&lt;/a&gt;&lt;/p&gt;




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

&lt;h3&gt;
  
  
  Featured Categories:
&lt;/h3&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Best Use of Gemma:&lt;/strong&gt; Google's &lt;code&gt;gemma3:4b&lt;/code&gt; runs locally via Ollama to power the poetic remembrance, the strictly grounded memory echo, and local vision photo captioning. Models are swappable at runtime via the "What Runs Where" panel.&lt;/li&gt;
&lt;/ul&gt;

&lt;h3&gt;
  
  
  Partner Categories:
&lt;/h3&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Best Use of ElevenLabs:&lt;/strong&gt; Optional Scribe STT for recording spoken oral histories and generic narrator TTS for reading remembrances aloud. Off by default; zero voice cloning.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Best Use of Mastra:&lt;/strong&gt; The Echo can run as a local Mastra agent on top of Ollama, featuring persistent conversational memory stored in local SQLite and an instant "Forget Our Visits" privacy purge.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Best Use of Ultramarine (Paper Instruments):&lt;/strong&gt; Features an in-app side-by-side benchmark comparing local Gemma vs. Ultramarine cloud inference with real-time latency and token spend tracking.&lt;/li&gt;
&lt;/ul&gt;




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

&lt;p&gt;Día de los Muertos teaches us that people only truly leave us when they are forgotten.&lt;/p&gt;

&lt;p&gt;Technology shouldn't try to replace the people we love with synthetic, hallucinating avatars. Instead, open-source software can give us an honest digital sanctuary—a quiet place where their real words, favorite melodies, and cherished stories remain safe, private, and permanent.&lt;/p&gt;

&lt;p&gt;That is the true ghost in the machine: not an artificial phantom, but the enduring warmth of our own memories, preserved on our own terms.&lt;/p&gt;

</description>
      <category>devchallenge</category>
      <category>hf26challenge</category>
      <category>opensource</category>
      <category>python</category>
    </item>
    <item>
      <title>findbox: I built a tool that quietly organizes your digital mess while you sleep — 100% local, open-weight Gemma, zero cloud</title>
      <dc:creator>Vinit Chaurasia</dc:creator>
      <pubDate>Sun, 25 Oct 2026 18:23:18 +0000</pubDate>
      <link>https://dev.to/chourasiavinit9/findbox-i-built-a-tool-that-quietly-organizes-your-digital-mess-while-you-sleep-100-local-1j0d</link>
      <guid>https://dev.to/chourasiavinit9/findbox-i-built-a-tool-that-quietly-organizes-your-digital-mess-while-you-sleep-100-local-1j0d</guid>
      <description>&lt;h1&gt;
  
  
  🦥 findbox — Drop it in. Forget it. Find it by what you remember.
&lt;/h1&gt;

&lt;blockquote&gt;
&lt;p&gt;&lt;em&gt;Built for Hacktoberfest 2026 Week 3: Slow &amp;amp; Steady (International Sloth Day). Open-source, 100% local, zero cloud.&lt;/em&gt;&lt;/p&gt;
&lt;/blockquote&gt;

&lt;p&gt;&lt;strong&gt;👉 GitHub&lt;/strong&gt;: &lt;a href="https://github.com/chourasiavinit9-dev/findbox" rel="noopener noreferrer"&gt;chourasiavinit9-dev/findbox&lt;/a&gt;&lt;br&gt;&lt;br&gt;
&lt;strong&gt;🌐 Static Demo (judges)&lt;/strong&gt;: &lt;a href="https://chourasiavinit9-dev.github.io/findbox/" rel="noopener noreferrer"&gt;chourasiavinit9-dev.github.io/findbox&lt;/a&gt;&lt;br&gt;&lt;br&gt;
&lt;strong&gt;🐳 Docker Ready&lt;/strong&gt;: &lt;code&gt;docker compose up --build&lt;/code&gt;&lt;/p&gt;


&lt;h2&gt;
  
  
  The Problem: Your Downloads Folder Is a Crime Scene
&lt;/h2&gt;

&lt;p&gt;I have 847 files in my Downloads folder right now. &lt;code&gt;IMG_4471.png&lt;/code&gt; (an Instagram recipe I'll never find), &lt;code&gt;scan(3)_final_FINAL.pdf&lt;/code&gt; (who knows), &lt;code&gt;WhatsApp Image 2026-09-14.jpg&lt;/code&gt; (could be anything), and buried in there somewhere — an electricity bill I need to pay.&lt;/p&gt;

&lt;p&gt;Sorting this takes hours. So nobody does it. The mess compounds. Then one day you get a late fee because the bill was buried under 200 screenshots.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;findbox&lt;/strong&gt; solves this while you sleep.&lt;/p&gt;


&lt;h2&gt;
  
  
  📸 The Final Web App Look
&lt;/h2&gt;
&lt;h3&gt;
  
  
  1. Homepage &amp;amp; The Payoff Meter
&lt;/h3&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%2Frqcxqrrfztp37vp41efw.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%2Frqcxqrrfztp37vp41efw.png" alt="findbox homepage with payoff meter" width="799" height="454"&gt;&lt;/a&gt;&lt;/p&gt;
&lt;h3&gt;
  
  
  2. Vague Memory Search (AI-Parsed Query + Instant Filter)
&lt;/h3&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%2Fi1o9s36iti86ui37kq1i.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%2Fi1o9s36iti86ui37kq1i.png" alt="findbox vague search for paneer recipe" width="799" height="454"&gt;&lt;/a&gt;&lt;/p&gt;
&lt;h3&gt;
  
  
  3. Action Items Deadlines &amp;amp; Device Scanner
&lt;/h3&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%2Fpc8d1jnpx3sudk7rel60.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%2Fpc8d1jnpx3sudk7rel60.png" alt="findbox action items" width="799" height="454"&gt;&lt;/a&gt;&lt;/p&gt;


&lt;h2&gt;
  
  
  The Build Session
&lt;/h2&gt;

&lt;p&gt;I built the entire thing with Antigravity (Gemini coding agent) in one session:&lt;/p&gt;


&lt;div class="ltag-agent-session"&gt;
  &lt;div class="agent-session-header"&gt;
    
    &lt;span class="agent-session-tool-icon-badge" title="Gemini CLI"&gt;
&lt;/span&gt;
    &lt;span class="agent-session-title"&gt;Building findbox: Local AI File Organizer (Hacktoberfest 2026)&lt;/span&gt;
  &lt;/div&gt;

  &lt;div class="agent-session-scroll"&gt;

      &lt;div class="agent-session-message agent-session-user"&gt;
        &lt;div class="agent-session-role-badge agent-session-role-user"&gt;
          You
        &lt;/div&gt;
        &lt;div class="agent-session-content"&gt;
                &lt;div class="agent-session-text"&gt;
                  &lt;p&gt;Build a winning submission for 'Slow and Steady: International Sloth Day'. PROJECT: findbox. Tagline: 'Drop it in. Forget it. Find it by what you remember.' A fully local tool that quietly cleans up your digital mess overnight (Downloads folder, screenshots, photos, PDFs, scans, exported chats) and then lets you find anything by describing it vaguely.&lt;/p&gt;

                &lt;/div&gt;
        &lt;/div&gt;
      &lt;/div&gt;

      &lt;div class="agent-session-message agent-session-assistant"&gt;
        &lt;div class="agent-session-role-badge agent-session-role-assistant"&gt;
          Agent
        &lt;/div&gt;
        &lt;div class="agent-session-content"&gt;
                &lt;div class="agent-session-text"&gt;
                  &lt;p&gt;Perfect project for the Sloth Day theme! Here's the architecture I'll build: zero-config auto-scan of common device landfill folders, open-weight Gemma 3 for local inference via Ollama, SQLite FTS5 BM25 for vague-memory search, and a neobrutalism web UI at &lt;span class="agent-session-redacted"&gt;[REDACTED]&lt;/span&gt;:8765. Let me start building.&lt;/p&gt;

                &lt;/div&gt;
              &lt;div class="agent-session-tool-call"&gt;
                
                  &lt;span class="agent-session-chevron"&gt;▸&lt;/span&gt;
                  &lt;span class="tool-name tool-name-writetofile"&gt;write_to_file&lt;/span&gt;
                    &lt;span class="tool-input-preview"&gt;sloth_box/organizer.py — core intake pipeline&lt;/span&gt;
                
                &lt;div class="agent-session-tool-detail"&gt;
                    &lt;div class="tool-section"&gt;
                      &lt;div class="tool-section-label"&gt;Input&lt;/div&gt;
                      &lt;pre class="agent-session-pre"&gt;&lt;code&gt;sloth_box/organizer.py — core intake pipeline&lt;/code&gt;&lt;/pre&gt;
                    &lt;/div&gt;
                    &lt;div class="tool-section"&gt;
                      &lt;div class="tool-section-label"&gt;Output&lt;/div&gt;
                      &lt;pre class="agent-session-pre"&gt;&lt;code&gt;Created file with run_intake(), auto-discovery, SHA256 dedup, collision-safe rename&lt;/code&gt;&lt;/pre&gt;
                    &lt;/div&gt;
                &lt;/div&gt;
              &lt;/div&gt;
        &lt;/div&gt;
      &lt;/div&gt;

      &lt;div class="agent-session-message agent-session-user"&gt;
        &lt;div class="agent-session-role-badge agent-session-role-user"&gt;
          You
        &lt;/div&gt;
        &lt;div class="agent-session-content"&gt;
                &lt;div class="agent-session-text"&gt;
                  &lt;p&gt;make it like it can access anything on the device — zero config&lt;/p&gt;

                &lt;/div&gt;
        &lt;/div&gt;
      &lt;/div&gt;

      &lt;div class="agent-session-message agent-session-assistant"&gt;
        &lt;div class="agent-session-role-badge agent-session-role-assistant"&gt;
          Agent
        &lt;/div&gt;
        &lt;div class="agent-session-content"&gt;
                &lt;div class="agent-session-text"&gt;
                  &lt;p&gt;Added AUTO_SCAN_DIRS to config.py — automatically discovers ~/Downloads, ~/Desktop, ~/Documents, ~/Pictures/Screenshots, WhatsApp Chat exports, iCloud Drive. No path needed. User just runs: python3 sloth_box.py run --apply&lt;/p&gt;

                &lt;/div&gt;
              &lt;div class="agent-session-tool-call"&gt;
                
                  &lt;span class="agent-session-chevron"&gt;▸&lt;/span&gt;
                  &lt;span class="tool-name tool-name-multireplacefilecontent"&gt;multi_replace_file_content&lt;/span&gt;
                    &lt;span class="tool-input-preview"&gt;Add AUTO_SCAN_DIRS + get_auto_scan_dirs() to config.py&lt;/span&gt;
                
                &lt;div class="agent-session-tool-detail"&gt;
                    &lt;div class="tool-section"&gt;
                      &lt;div class="tool-section-label"&gt;Input&lt;/div&gt;
                      &lt;pre class="agent-session-pre"&gt;&lt;code&gt;Add AUTO_SCAN_DIRS + get_auto_scan_dirs() to config.py&lt;/code&gt;&lt;/pre&gt;
                    &lt;/div&gt;
                    &lt;div class="tool-section"&gt;
                      &lt;div class="tool-section-label"&gt;Output&lt;/div&gt;
                      &lt;pre class="agent-session-pre"&gt;&lt;code&gt;Updated config.py with 8 device landfill locations&lt;/code&gt;&lt;/pre&gt;
                    &lt;/div&gt;
                &lt;/div&gt;
              &lt;/div&gt;
        &lt;/div&gt;
      &lt;/div&gt;

      &lt;div class="agent-session-message agent-session-user"&gt;
        &lt;div class="agent-session-role-badge agent-session-role-user"&gt;
          You
        &lt;/div&gt;
        &lt;div class="agent-session-content"&gt;
                &lt;div class="agent-session-text"&gt;
                  &lt;p&gt;add a button to clear text search&lt;/p&gt;

                &lt;/div&gt;
        &lt;/div&gt;
      &lt;/div&gt;

      &lt;div class="agent-session-message agent-session-assistant"&gt;
        &lt;div class="agent-session-role-badge agent-session-role-assistant"&gt;
          Agent
        &lt;/div&gt;
        &lt;div class="agent-session-content"&gt;
                &lt;div class="agent-session-text"&gt;
                  &lt;p&gt;Added a floating ✕ clear button inside the search input. JavaScript toggleClear() shows/hides it as you type. clearSearch() wipes the input and returns to /?. Now users can search multiple items without backspacing.&lt;/p&gt;

                &lt;/div&gt;
        &lt;/div&gt;
      &lt;/div&gt;

      &lt;div class="agent-session-message agent-session-user"&gt;
        &lt;div class="agent-session-role-badge agent-session-role-user"&gt;
          You
        &lt;/div&gt;
        &lt;div class="agent-session-content"&gt;
                &lt;div class="agent-session-text"&gt;
                  &lt;p&gt;see bug — ai error html and sample data not openable, fix both&lt;/p&gt;

                &lt;/div&gt;
        &lt;/div&gt;
      &lt;/div&gt;

      &lt;div class="agent-session-message agent-session-assistant"&gt;
        &lt;div class="agent-session-role-badge agent-session-role-assistant"&gt;
          Agent
        &lt;/div&gt;
        &lt;div class="agent-session-content"&gt;
                &lt;div class="agent-session-text"&gt;
                  &lt;p&gt;Fixed two bugs: 1) {{ai_meta_html}} was double-escaped — changed to {ai_meta_html} so AI search chips render correctly. 2) Added '📄 Open File ↗' button on every document card using file:// protocol from current_path in SQLite — clicking opens the actual file in its native macOS app.&lt;/p&gt;

                &lt;/div&gt;
        &lt;/div&gt;
      &lt;/div&gt;

      &lt;div class="agent-session-message agent-session-user"&gt;
        &lt;div class="agent-session-role-badge agent-session-role-user"&gt;
          You
        &lt;/div&gt;
        &lt;div class="agent-session-content"&gt;
                &lt;div class="agent-session-text"&gt;
                  &lt;p&gt;add scan my device button and fix action items panel&lt;/p&gt;

                &lt;/div&gt;
        &lt;/div&gt;
      &lt;/div&gt;

      &lt;div class="agent-session-message agent-session-assistant"&gt;
        &lt;div class="agent-session-role-badge agent-session-role-assistant"&gt;
          Agent
        &lt;/div&gt;
        &lt;div class="agent-session-content"&gt;
                &lt;div class="agent-session-text"&gt;
                  &lt;p&gt;Done! Added: (1) 'Scan My Device' panel showing all auto-scan folders as chips + a POST /scan button that triggers run_intake() in a background thread — no terminal needed. (2) Replaced 7-day Due Soon filter with 'Action Items' showing ALL todos color-coded: red = due ≤7 days, purple = future, gray = past. 18/18 tests still passing.&lt;/p&gt;

                &lt;/div&gt;
        &lt;/div&gt;
      &lt;/div&gt;
  &lt;/div&gt;

  &lt;div class="agent-session-footer"&gt;
    &lt;span class="agent-session-meta"&gt;
        10 of 10 messages
    &lt;/span&gt;
  &lt;/div&gt;
&lt;/div&gt;





&lt;h2&gt;
  
  
  The Lazy Workflow
&lt;/h2&gt;



&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;&lt;span class="c"&gt;# Run this once before bed (or click 'Scan My Device' in the UI)&lt;/span&gt;
python3 sloth_box.py run &lt;span class="nt"&gt;--apply&lt;/span&gt;

&lt;span class="c"&gt;# Wake up to everything organized. Then search:&lt;/span&gt;
python3 sloth_box.py serve  &lt;span class="c"&gt;# http://127.0.0.1:8765&lt;/span&gt;
&lt;span class="c"&gt;# Search: "paneer recipe september" → finds it instantly&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;&lt;strong&gt;Your effort: 1 command. That's literally it.&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;findbox runs at &lt;code&gt;os.nice(19)&lt;/code&gt; — the absolute lowest CPU priority, yielding to every other process on the machine.&lt;/p&gt;




&lt;h2&gt;
  
  
  🐳 Zero-Install: Run With Docker
&lt;/h2&gt;

&lt;p&gt;Don't have Python or local dependencies configured? Run with Docker Compose:&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/chourasiavinit9-dev/findbox &lt;span class="o"&gt;&amp;amp;&amp;amp;&lt;/span&gt; &lt;span class="nb"&gt;cd &lt;/span&gt;findbox
docker compose up &lt;span class="nt"&gt;--build&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Or plain Docker:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;docker build &lt;span class="nt"&gt;-t&lt;/span&gt; findbox &lt;span class="nb"&gt;.&lt;/span&gt;
docker run &lt;span class="nt"&gt;-p&lt;/span&gt; 8765:8765 &lt;span class="se"&gt;\&lt;/span&gt;
  &lt;span class="nt"&gt;-v&lt;/span&gt; ~/Downloads:/inbox/Downloads &lt;span class="se"&gt;\&lt;/span&gt;
  &lt;span class="nt"&gt;-v&lt;/span&gt; ~/Desktop:/inbox/Desktop &lt;span class="se"&gt;\&lt;/span&gt;
  &lt;span class="nt"&gt;-v&lt;/span&gt; ~/SlothBox:/home/sloth/SlothBox &lt;span class="se"&gt;\&lt;/span&gt;
  findbox
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Open &lt;strong&gt;&lt;a href="http://localhost:8765" rel="noopener noreferrer"&gt;http://localhost:8765&lt;/a&gt;&lt;/strong&gt; — your files, organized, indexed, and searchable.&lt;/p&gt;




&lt;h2&gt;
  
  
  Why Open-Weight Models Were Non-Negotiable
&lt;/h2&gt;

&lt;p&gt;Think about what's in your Downloads folder: bank statements, medical prescriptions, Aadhaar/passport scans, private WhatsApp exports, OTP screenshots.&lt;/p&gt;

&lt;p&gt;Sending any of this to a cloud API is unacceptable. &lt;strong&gt;Open-weight Gemma 3 running locally via Ollama&lt;/strong&gt; was the only viable choice:&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Feature&lt;/th&gt;
&lt;th&gt;Closed Cloud APIs&lt;/th&gt;
&lt;th&gt;findbox (Gemma Local)&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;Privacy&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;Files leave your device&lt;/td&gt;
&lt;td&gt;Files never leave RAM&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;Offline&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;Requires internet&lt;/td&gt;
&lt;td&gt;Works on an airplane&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;Cost&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;Per-token for 1000s of files&lt;/td&gt;
&lt;td&gt;Free at any scale&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;Language&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;English-optimized&lt;/td&gt;
&lt;td&gt;Hindi/Bengali natively&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;When Ollama isn't running, findbox falls back to a deterministic heuristics engine — so it never crashes.&lt;/p&gt;




&lt;h2&gt;
  
  
  Key Features
&lt;/h2&gt;

&lt;h3&gt;
  
  
  Zero-Config Auto-Discovery
&lt;/h3&gt;

&lt;p&gt;No path configuration. It auto-finds &lt;code&gt;~/Downloads&lt;/code&gt;, &lt;code&gt;~/Desktop&lt;/code&gt;, &lt;code&gt;~/Pictures/Screenshots&lt;/code&gt;, &lt;code&gt;~/Documents&lt;/code&gt;, WhatsApp exports, and iCloud Drive. Just run it.&lt;/p&gt;

&lt;h3&gt;
  
  
  Meaningful Renaming
&lt;/h3&gt;



&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;IMG_4471.png → 2026-09-04_instagram-paneer-tikka_recipe.png
scan(3)_final_FINAL.pdf → 2026-10-12_electricity-bill_WBSEDCL.pdf
WhatsApp Image 2026-09-14.jpg → 2026-09-14_rent-chat-rahul_chats.jpg
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;h3&gt;
  
  
  Vague-Memory Search
&lt;/h3&gt;

&lt;p&gt;Gemma parses your query into structured parameters (keywords + category + date range), then SQLite FTS5 BM25 finds the exact file. Search &lt;em&gt;"paneer recipe september"&lt;/em&gt; → finds the Instagram screenshot from Sept 4. Includes a dedicated ✕ clear button to immediately wipe filters.&lt;/p&gt;

&lt;h3&gt;
  
  
  Scan from the UI
&lt;/h3&gt;

&lt;p&gt;The &lt;strong&gt;🦥 Scan My Device Now&lt;/strong&gt; button in the web UI triggers background scanning with live visual feedback. No terminal needed.&lt;/p&gt;

&lt;h3&gt;
  
  
  Action Items Panel
&lt;/h3&gt;

&lt;p&gt;Deadlines extracted from every file, color-coded: 🔴 red = due ≤7 days, 🟣 purple = future, ⚫ gray = past.&lt;/p&gt;




&lt;h2&gt;
  
  
  Architecture
&lt;/h2&gt;



&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;sloth_box/
├── config.py     — AUTO_SCAN_DIRS (8 device locations)
├── security.py   — path-traversal, XSS disarm, magic-byte quarantine
├── extractor.py  — PDF/DOCX/image/text + fallback heuristics
├── ai.py         — local Ollama/Gemma client + vague query parser
├── organizer.py  — multi-folder intake, collision-safe rename, undo
├── db.py         — SQLite WAL + FTS5 + audit log
└── server.py     — WCAG 2.1 AA web UI (neobrutalism design)
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;&lt;strong&gt;Safety:&lt;/strong&gt; dry-run by default, atomic undo, no overwrites, no deletes, localhost-only server.&lt;/p&gt;




&lt;h2&gt;
  
  
  The Payoff Meter
&lt;/h2&gt;

&lt;p&gt;$$\text{Minutes Saved} = (\text{Files Indexed} \times 1.5) + (\text{Searches Avoided} \times 5.0)$$&lt;/p&gt;

&lt;p&gt;On sample data: &lt;strong&gt;15 minutes saved, 0 minutes spent.&lt;/strong&gt; That's the sloth way.&lt;/p&gt;




&lt;h2&gt;
  
  
  Why Open Innovation Matters
&lt;/h2&gt;

&lt;p&gt;This project wouldn't exist as a privacy-safe tool without open-weight models. The files it processes are genuinely sensitive. Gemma running locally means:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;strong&gt;Data never moves&lt;/strong&gt; — inference in local RAM&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;No internet required&lt;/strong&gt; — fully offline&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;No metered cost&lt;/strong&gt; — 10,000 files, pay nothing&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Multilingual&lt;/strong&gt; — handles Hindi/Bengali natively (WBSEDCL, Aadhaar, kharcha all work)&lt;/li&gt;
&lt;/ol&gt;




&lt;h2&gt;
  
  
  Try It
&lt;/h2&gt;



&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;git clone https://github.com/chourasiavinit9-dev/findbox &lt;span class="o"&gt;&amp;amp;&amp;amp;&lt;/span&gt; &lt;span class="nb"&gt;cd &lt;/span&gt;findbox
python3 sloth_box.py run        &lt;span class="c"&gt;# dry-run&lt;/span&gt;
python3 sloth_box.py run &lt;span class="nt"&gt;--apply&lt;/span&gt;  &lt;span class="c"&gt;# apply&lt;/span&gt;
python3 sloth_box.py serve      &lt;span class="c"&gt;# search UI at http://127.0.0.1:8765&lt;/span&gt;
python3 &lt;span class="nt"&gt;-m&lt;/span&gt; pytest tests/ &lt;span class="nt"&gt;-v&lt;/span&gt;     &lt;span class="c"&gt;# 18 tests, 0.04s&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;ul&gt;
&lt;li&gt;🐙 &lt;strong&gt;GitHub Repository&lt;/strong&gt;: &lt;a href="https://github.com/chourasiavinit9-dev/findbox" rel="noopener noreferrer"&gt;chourasiavinit9-dev/findbox&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;🌐 &lt;strong&gt;Judge Interactive Demo&lt;/strong&gt;: &lt;a href="https://chourasiavinit9-dev.github.io/findbox/" rel="noopener noreferrer"&gt;chourasiavinit9-dev.github.io/findbox&lt;/a&gt;
&lt;/li&gt;
&lt;/ul&gt;




&lt;p&gt;&lt;em&gt;🦥 Hacktoberfest 2026 Week 3 — Slow &amp;amp; Steady · International Sloth Day · Open-source · 100% local · Zero cloud&lt;/em&gt;&lt;/p&gt;

</description>
      <category>hf26challenge</category>
      <category>hacktoberfest</category>
      <category>python</category>
      <category>opensource</category>
    </item>
    <item>
      <title>Fieldcard: I built a local AI that prints you a field card and gets you outside in 30 seconds</title>
      <dc:creator>Vinit Chaurasia</dc:creator>
      <pubDate>Sun, 11 Oct 2026 14:56:28 +0000</pubDate>
      <link>https://dev.to/chourasiavinit9/fieldcard-i-built-a-local-ai-that-prints-you-a-field-card-and-gets-you-outside-in-30-seconds-4m71</link>
      <guid>https://dev.to/chourasiavinit9/fieldcard-i-built-a-local-ai-that-prints-you-a-field-card-and-gets-you-outside-in-30-seconds-4m71</guid>
      <description>&lt;h2&gt;
  
  
  What I built
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;Fieldcard&lt;/strong&gt; is a single-command tool that turns "I have 90 minutes" into a one-page printable A5 field card. You type a command (or fill a form), wait 10 seconds, and walk out the door holding a card that tells you:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;A mission written for your exact terrain and season&lt;/li&gt;
&lt;li&gt;A radial map of real nearby parks, water, and viewpoints pulled from OpenStreetMap&lt;/li&gt;
&lt;li&gt;A "notice nine things" bingo grid tuned to what's actually observable right now&lt;/li&gt;
&lt;li&gt;Sunset time, turn-around time, and how far you can reach — all computed in code, never guessed by the model&lt;/li&gt;
&lt;li&gt;Live weather conditions and real local species from iNaturalist&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;The screen is the shortest part. That's the point.&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;&lt;span class="c"&gt;# Generate a 90-minute walk card for Central Park&lt;/span&gt;
python fieldcard.py &lt;span class="nt"&gt;--lat&lt;/span&gt; 40.7678 &lt;span class="nt"&gt;--lon&lt;/span&gt; &lt;span class="nt"&gt;-73&lt;/span&gt;.9718 &lt;span class="nt"&gt;--minutes&lt;/span&gt; 90 &lt;span class="nt"&gt;--activity&lt;/span&gt; walk
&lt;span class="c"&gt;# → Writes fieldcard.html. Print it (A5). Put the phone away.&lt;/span&gt;

&lt;span class="c"&gt;# Or run the local web UI (no frameworks, no build step)&lt;/span&gt;
python fieldcard.py serve &lt;span class="nt"&gt;--port&lt;/span&gt; 8000
&lt;span class="c"&gt;# → Open http://localhost:8000, fill the form, print, go.&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;

&lt;h2&gt;
  
  
  The open-source AI part
&lt;/h2&gt;

&lt;p&gt;The local open-weight model (default: &lt;code&gt;llama3.2:3b&lt;/code&gt; via Ollama) writes &lt;strong&gt;only the words&lt;/strong&gt; — the title, the mission paragraph, and the nine sensory prompts. It's given real facts: your season, the actual place names nearby, the activity, the species observed this month. It writes poetry grounded in your specific terrain.&lt;/p&gt;

&lt;p&gt;The code handles everything safety-critical: sunset time (NOAA solar equations), turn-around time, reach radius, daylight warnings. These are never left to the model.&lt;br&gt;
&lt;/p&gt;
&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight python"&gt;&lt;code&gt;&lt;span class="c1"&gt;# The LLM writes the poetry. Code does the safety math.
# Design rule: if Ollama is offline, the card still works.
&lt;/span&gt;&lt;span class="n"&gt;got&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="nf"&gt;llm_copy&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;facts&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;places&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;fallback_copy&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;
&lt;span class="k"&gt;if&lt;/span&gt; &lt;span class="ow"&gt;not&lt;/span&gt; &lt;span class="n"&gt;got&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;
    &lt;span class="c1"&gt;# Built-in seasonal sensory prompts kick in automatically
&lt;/span&gt;    &lt;span class="n"&gt;copy&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="n"&gt;fallback_copy&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;


&lt;p&gt;You can swap any open-weight model with &lt;code&gt;--model gemma2:2b&lt;/code&gt; or &lt;code&gt;--model qwen2.5:3b&lt;/code&gt;. No API key. No cloud. No subscription.&lt;/p&gt;
&lt;h2&gt;
  
  
  Why open-source AI made this possible
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;It runs offline.&lt;/strong&gt; On the trail, in a park with patchy signal, or on a plane — the card generates regardless. A cloud API would break the moment you most need the card.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Location data never leaves your machine.&lt;/strong&gt; Your coordinates go to Overpass (OpenStreetMap) and iNaturalist for real place data, but the AI reasoning about your location happens entirely on your own hardware.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;You can swap, fine-tune, or inspect the model.&lt;/strong&gt; The prompt is visible in the source. You could fine-tune on your own field journals. A closed model makes that impossible.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;It costs nothing to run.&lt;/strong&gt; A 3B model on a laptop. No metered tokens, no monthly bill.&lt;/p&gt;
&lt;h2&gt;
  
  
  Architecture
&lt;/h2&gt;


&lt;pre data-lang="mermaid"&gt;&lt;code&gt;graph LR
    A[CLI / Web Form] --&amp;gt; B[build pipeline]
    B --&amp;gt; C[OpenStreetMap Overpass&amp;lt;br/&amp;gt;real nearby places]
    B --&amp;gt; D[iNaturalist API&amp;lt;br/&amp;gt;local species this month]
    B --&amp;gt; E[Open-Meteo&amp;lt;br/&amp;gt;live weather, no key]
    B --&amp;gt; F[NOAA solar equations&amp;lt;br/&amp;gt;sunset + turn-around]
    B --&amp;gt; G[Ollama local model&amp;lt;br/&amp;gt;llama3.2:3b default]
    C --&amp;gt; H[render_card]
    D --&amp;gt; H
    E --&amp;gt; H
    F --&amp;gt; H
    G --&amp;gt; H
    H --&amp;gt; I[A5 printable HTML]
    I --&amp;gt; J[Printer]
    J --&amp;gt; K[Outside]&lt;/code&gt;&lt;/pre&gt;



&lt;p&gt;&lt;strong&gt;Design rule:&lt;/strong&gt; every external source degrades gracefully. No OpenStreetMap? Dead-reckoning rule prints instead. No Ollama? Built-in seasonal bingo prompts. No iNaturalist? Card omits the species bar. The card always prints.&lt;/p&gt;
&lt;h2&gt;
  
  
  Activity modes
&lt;/h2&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Mode&lt;/th&gt;
&lt;th&gt;What changes&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;&lt;code&gt;walk&lt;/code&gt;&lt;/td&gt;
&lt;td&gt;Radial map at 4.5 km/h, sensory bingo&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;code&gt;run&lt;/code&gt;&lt;/td&gt;
&lt;td&gt;Wider radius at 9 km/h, trail-focused prompts&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;code&gt;birding&lt;/code&gt;&lt;/td&gt;
&lt;td&gt;Slower 2.5 km/h, species bar prominent, iNaturalist birds&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;code&gt;garden&lt;/code&gt;&lt;/td&gt;
&lt;td&gt;No map, frost-date-aware planting advice from local crop database&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;code&gt;--group&lt;/code&gt;&lt;/td&gt;
&lt;td&gt;Pace routes (conversational / steady / long) + QR code to meetup point&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;
&lt;h2&gt;
  
  
  The web UI: a doorway, not a dashboard
&lt;/h2&gt;

&lt;p&gt;Running &lt;code&gt;python fieldcard.py serve&lt;/code&gt; starts a local server (Python stdlib &lt;code&gt;http.server&lt;/code&gt; — no frameworks, no npm, works offline). The form has:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Streak counter&lt;/strong&gt; — consecutive days you've gone outside&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Achievement badges&lt;/strong&gt; — First Steps, 3-Day Streak, Week Explorer, Bird Watcher, Green Thumb&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Live weather&lt;/strong&gt; — current conditions fetched from Open-Meteo (free, no key)&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Card history&lt;/strong&gt; — last 5 cards you generated&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;One button&lt;/strong&gt; — Generate Field Card&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Time at the screen: under 30 seconds.&lt;/p&gt;
&lt;h2&gt;
  
  
  Tech stack (all open-source)
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Python 3.9+ stdlib only&lt;/strong&gt; — no pip install required for core&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Ollama&lt;/strong&gt; — local LLM inference runtime&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;llama3.2:3b&lt;/strong&gt; (default) — swap to &lt;code&gt;gemma2:2b&lt;/code&gt;, &lt;code&gt;qwen2.5:3b&lt;/code&gt;, or anything you have&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;OpenStreetMap / Overpass API&lt;/strong&gt; — real nearby places, cached locally&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;iNaturalist API&lt;/strong&gt; — research-grade species observations&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Open-Meteo API&lt;/strong&gt; — free weather, no key, no account&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;NOAA solar equations&lt;/strong&gt; — pure Python sunrise/sunset math&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;26 unit tests&lt;/strong&gt; — geometry, solar, places, web server&lt;/li&gt;
&lt;/ul&gt;
&lt;h2&gt;
  
  
  I took it outside
&lt;/h2&gt;

&lt;p&gt;I generated a birding card for a local park, printed it, and spent 45 minutes looking for the nine things on the bingo grid. The model had written: &lt;em&gt;"A bird puffed into a ball against the wind"&lt;/em&gt; — and I found one within 10 minutes perched in an oak. The card stayed in my pocket for the whole walk. The phone didn't come out once.&lt;/p&gt;

&lt;p&gt;That's the test that matters.&lt;/p&gt;
&lt;h2&gt;
  
  
  Why open innovation matters here
&lt;/h2&gt;

&lt;p&gt;A closed-model version of Fieldcard would require an internet connection, bill you per card, and own your location history. The open-weight version runs on a 2019 MacBook, works in airplane mode, and costs nothing after the first &lt;code&gt;ollama pull&lt;/code&gt;.&lt;/p&gt;

&lt;p&gt;For something designed to get you &lt;em&gt;away&lt;/em&gt; from your phone and &lt;em&gt;into&lt;/em&gt; the world, that's not just a technical preference — it's the whole point.&lt;/p&gt;
&lt;h2&gt;
  
  
  Agent session transcript
&lt;/h2&gt;

&lt;p&gt;DevRelay recorded our development transcript including the prompt iterations, offline-first fallback logic, solar calculation constraints, and streak engine:&lt;/p&gt;


&lt;div class="ltag-agent-session"&gt;
  &lt;div class="agent-session-header"&gt;
    
    &lt;span class="agent-session-tool-icon-badge" title="Gemini CLI"&gt;
&lt;/span&gt;
    &lt;span class="agent-session-title"&gt;Building Fieldcard: Local AI, Printable Pocket Cards, &amp;amp; Streak Tracking&lt;/span&gt;
  &lt;/div&gt;

  &lt;div class="agent-session-scroll"&gt;

      &lt;div class="agent-session-message agent-session-user"&gt;
        &lt;div class="agent-session-role-badge agent-session-role-user"&gt;
          You
        &lt;/div&gt;
        &lt;div class="agent-session-content"&gt;
                &lt;div class="agent-session-text"&gt;
                  &lt;p&gt;Build Fieldcard: an offline-first tool that generates a printable A5 physical pocket card for outdoor micro-explorations using local open-weight LLMs, NOAA solar equations, and Open-Meteo weather.&lt;/p&gt;

                &lt;/div&gt;
        &lt;/div&gt;
      &lt;/div&gt;

      &lt;div class="agent-session-message agent-session-assistant"&gt;
        &lt;div class="agent-session-role-badge agent-session-role-assistant"&gt;
          Agent
        &lt;/div&gt;
        &lt;div class="agent-session-content"&gt;
                &lt;div class="agent-session-text"&gt;
                  &lt;p&gt;Designing Fieldcard around a zero-screen, print-first philosophy:&lt;br&gt;
1. Deterministic NOAA solar calculation module for exact turn-around time and sunset warnings without network connectivity.&lt;br&gt;
2. Local inference pipeline using Ollama (Llama 3.2 3B / Gemma 2 2B) for sensory cues grounded in local coordinates, with procedural fallback if Ollama is absent.&lt;br&gt;
3. CSS print layout styled for A5 paper with micro-typography, turn-around timers, and field-note checklists.&lt;br&gt;
4. Local observation journal for recording bird sightings, plants, and trail thoughts after returning.&lt;/p&gt;

                &lt;/div&gt;
              &lt;div class="agent-session-tool-call"&gt;
                
                  &lt;span class="agent-session-chevron"&gt;▸&lt;/span&gt;
                  &lt;span class="tool-name tool-name-runcommand"&gt;run_command&lt;/span&gt;
                    &lt;span class="tool-input-preview"&gt;pytest tests/ -v&lt;/span&gt;
                
                &lt;div class="agent-session-tool-detail"&gt;
                    &lt;div class="tool-section"&gt;
                      &lt;div class="tool-section-label"&gt;Input&lt;/div&gt;
                      &lt;pre class="agent-session-pre agent-session-terminal"&gt;&lt;code&gt;$ pytest tests/ -v&lt;/code&gt;&lt;/pre&gt;
                    &lt;/div&gt;
                    &lt;div class="tool-section"&gt;
                      &lt;div class="tool-section-label"&gt;Output&lt;/div&gt;
                      &lt;pre class="agent-session-pre"&gt;&lt;code&gt;=================== 26 passed in 0.42s ===================&lt;/code&gt;&lt;/pre&gt;
                    &lt;/div&gt;
                &lt;/div&gt;
              &lt;/div&gt;
        &lt;/div&gt;
      &lt;/div&gt;

      &lt;div class="agent-session-message agent-session-user"&gt;
        &lt;div class="agent-session-role-badge agent-session-role-user"&gt;
          You
        &lt;/div&gt;
        &lt;div class="agent-session-content"&gt;
                &lt;div class="agent-session-text"&gt;
                  &lt;p&gt;Add streak tracking and achievement badges to motivate daily outdoor walks, and redesign the UI for instant A5 card preview and printing.&lt;/p&gt;

                &lt;/div&gt;
        &lt;/div&gt;
      &lt;/div&gt;

      &lt;div class="agent-session-message agent-session-assistant"&gt;
        &lt;div class="agent-session-role-badge agent-session-role-assistant"&gt;
          Agent
        &lt;/div&gt;
        &lt;div class="agent-session-content"&gt;
                &lt;div class="agent-session-text"&gt;
                  &lt;p&gt;Implemented streak engine and achievement system:&lt;br&gt;
- Consecutive-day streak calculation stored locally with milestone checks.&lt;br&gt;
- 8 outdoor achievement badges (First Steps, Dawn Patrol, Night Owl, Weather Hardened, Weekend Wanderer, 7-Day Explorer, Deep Woods, Century Walk).&lt;br&gt;
- Real-time Open-Meteo forecast fetching with automatic fallback to seasonal climate tables when offline.&lt;br&gt;
- Responsive web dashboard with live A5 print preview, one-click print button, and badge drawer.&lt;/p&gt;

                &lt;/div&gt;
              &lt;div class="agent-session-tool-call"&gt;
                
                  &lt;span class="agent-session-chevron"&gt;▸&lt;/span&gt;
                  &lt;span class="tool-name tool-name-runcommand"&gt;run_command&lt;/span&gt;
                    &lt;span class="tool-input-preview"&gt;git commit -m "feat: add streak tracker, achievement badges, live weather, re...&lt;/span&gt;
                
                &lt;div class="agent-session-tool-detail"&gt;
                    &lt;div class="tool-section"&gt;
                      &lt;div class="tool-section-label"&gt;Input&lt;/div&gt;
                      &lt;pre class="agent-session-pre agent-session-terminal"&gt;&lt;code&gt;$ git commit -m "feat: add streak tracker, achievement badges, live weather, redesigned web UI"&lt;/code&gt;&lt;/pre&gt;
                    &lt;/div&gt;
                    &lt;div class="tool-section"&gt;
                      &lt;div class="tool-section-label"&gt;Output&lt;/div&gt;
                      &lt;pre class="agent-session-pre"&gt;&lt;code&gt;[main 7f21a8d] feat: add streak tracker, achievement badges, live weather, redesigned web UI&lt;/code&gt;&lt;/pre&gt;
                    &lt;/div&gt;
                &lt;/div&gt;
              &lt;/div&gt;
        &lt;/div&gt;
      &lt;/div&gt;
  &lt;/div&gt;

  &lt;div class="agent-session-footer"&gt;
    &lt;span class="agent-session-meta"&gt;
        4 of 4 messages
    &lt;/span&gt;
  &lt;/div&gt;
&lt;/div&gt;



&lt;p&gt;&lt;em&gt;Direct link:&lt;/em&gt; &lt;a href="https://dev.to/agent_sessions/building-fieldcard-local-ai-printable-pocket-cards-streak-tracking-5bfmd9"&gt;dev.to/agent_sessions/building-fieldcard-local-ai-printable-pocket-cards-streak-tracking-5bfmd9&lt;/a&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Links
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;GitHub:&lt;/strong&gt; &lt;a href="https://github.com/chourasiavinit9-dev/fieldcard" rel="noopener noreferrer"&gt;https://github.com/chourasiavinit9-dev/fieldcard&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Agent Session:&lt;/strong&gt; &lt;a href="https://dev.to/agent_sessions/building-fieldcard-local-ai-printable-pocket-cards-streak-tracking-5bfmd9"&gt;https://dev.to/agent_sessions/building-fieldcard-local-ai-printable-pocket-cards-streak-tracking-5bfmd9&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Try it:&lt;/strong&gt; &lt;code&gt;git clone https://github.com/chourasiavinit9-dev/fieldcard &amp;amp;&amp;amp; ollama pull llama3.2:3b &amp;amp;&amp;amp; python fieldcard.py --lat 40.7678 --lon -73.9718&lt;/code&gt;
&lt;/li&gt;
&lt;/ul&gt;




&lt;p&gt;&lt;em&gt;Submitted to the Hacktoberfest Open-Source AI Challenge: Week 1 — Touch Grass theme.&lt;/em&gt;&lt;br&gt;
&lt;em&gt;Built during October 5–11, 2026. New project, built for this challenge.&lt;/em&gt;&lt;/p&gt;

</description>
      <category>hf26challenge</category>
      <category>hacktoberfest</category>
      <category>opensource</category>
      <category>python</category>
    </item>
    <item>
      <title>I built my friend Avijit an AI terminal doctor — because he was losing his mind over version mismatches</title>
      <dc:creator>Vinit Chaurasia</dc:creator>
      <pubDate>Sun, 04 Oct 2026 19:34:40 +0000</pubDate>
      <link>https://dev.to/chourasiavinit9/i-built-my-friend-avijit-an-ai-terminal-doctor-because-he-was-losing-his-mind-over-version-133a</link>
      <guid>https://dev.to/chourasiavinit9/i-built-my-friend-avijit-an-ai-terminal-doctor-because-he-was-losing-his-mind-over-version-133a</guid>
      <description>&lt;p&gt;&lt;em&gt;This is a submission for the &lt;a href="https://dev.to/challenges/hacktoberfest-weekend-2026-10-01"&gt;Hacktoberfest Weekend Challenge: Build for a Friend&lt;/a&gt;&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Who I Built This For: My Friend Avijit
&lt;/h2&gt;

&lt;p&gt;My friend &lt;strong&gt;Avijit&lt;/strong&gt; is a genuinely talented developer — smart, curious, and willing to grind through tough logic. But almost every week, I would get a frantic late-night call:&lt;/p&gt;

&lt;blockquote&gt;
&lt;p&gt;&lt;em&gt;"Bhai, the terminal is throwing red errors and nothing makes sense. The command just refuses to run."&lt;/em&gt;&lt;/p&gt;
&lt;/blockquote&gt;

&lt;p&gt;He wasn't struggling with algorithms or business logic. He was struggling with &lt;strong&gt;his environment&lt;/strong&gt;.&lt;/p&gt;

&lt;p&gt;Just last week, I watched him lose over two hours debugging a bizarre &lt;code&gt;SyntaxError: Unexpected token '.'&lt;/code&gt;. It wasn't bad code — he had Node 14 installed globally, while his project required Node 18. The terminal didn't tell him &lt;em&gt;"wrong Node version"&lt;/em&gt;; it just threw a cryptic wall of stack traces. He was convinced his code was broken and was ready to rewrite the entire feature from scratch.&lt;/p&gt;

&lt;p&gt;Another time, his backend service failed because &lt;code&gt;pip freeze&lt;/code&gt; showed &lt;code&gt;fastapi==0.68.0&lt;/code&gt; in his global environment, but his project's &lt;code&gt;requirements.txt&lt;/code&gt; strictly required &lt;code&gt;&amp;gt;=0.100.0&lt;/code&gt;. He had no idea those two states were clashing or which Python binary his active terminal was executing.&lt;/p&gt;

&lt;p&gt;This is the "Avijit Problem" — but honestly, almost every developer has lost hours to mismatched dependencies, orphaned background processes hogging ports, or silent runtime version incompatibilities.&lt;/p&gt;

&lt;p&gt;I wanted to fix this once and for all. So I built him his own terminal doctor.&lt;/p&gt;




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

&lt;p&gt;&lt;strong&gt;&lt;a href="https://github.com/chourasiavinit9-dev/framework-magnifier" rel="noopener noreferrer"&gt;Framework Magnifier&lt;/a&gt;&lt;/strong&gt; — an open-source AI terminal doctor and system diagnostician.&lt;/p&gt;

&lt;p&gt;It is a fast, lightweight CLI that inspects running processes, open network ports, and multi-language dependency manifests. When something breaks, it translates cryptic errors and version conflicts into clear, actionable English instructions with a cyber-inspired neon HUD.&lt;/p&gt;

&lt;p&gt;Powered by Google's &lt;strong&gt;&lt;a href="https://ai.google.dev/gemma" rel="noopener noreferrer"&gt;Gemma-4&lt;/a&gt;&lt;/strong&gt; open-weight model family.&lt;/p&gt;




&lt;h2&gt;
  
  
  The Demo
&lt;/h2&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%2F2ilnr2xxkl10oqaluh43.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%2F2ilnr2xxkl10oqaluh43.jpg" alt="Framework Magnifier Terminal Demo" width="800" height="447"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;h3&gt;
  
  
  1. Instant System &amp;amp; Process Inspection
&lt;/h3&gt;



&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;python3 magnifier.py           &lt;span class="c"&gt;# Top 10 processes by CPU&lt;/span&gt;
python3 magnifier.py &lt;span class="nt"&gt;--mem&lt;/span&gt;     &lt;span class="c"&gt;# Top 10 processes by RAM&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;     PID │   CPU% │ Memory (MB) │ Username     │ Process Name
────────┼────────┼─────────────┼──────────────┼────────────────
       0 │    0.0 │         0.0 │ root         │ kernel_task
       1 │    0.0 │         0.0 │ root         │ launchd
     511 │    0.0 │         0.0 │ root         │ logd
     766 │    0.0 │       626.3 │ vinitchaurasia │ Antigravity IDE Helper
   99611 │    0.0 │       477.6 │ vinitchaurasia │ Google Chrome Helper
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;

&lt;h3&gt;
  
  
  2. Ask the AI: "What is this process doing?"
&lt;/h3&gt;


&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;python3 magnifier.py 1248
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;


&lt;blockquote&gt;
&lt;p&gt;&lt;em&gt;"This is a Node.js process (Webpack dev server) running from &lt;code&gt;/home/avijit/projects/web-app&lt;/code&gt;. It is safe, but currently consuming high CPU — check if a hot-reload watcher is stuck in a loop."&lt;/em&gt;&lt;/p&gt;
&lt;/blockquote&gt;

&lt;p&gt;Avijit's first reaction: &lt;em&gt;"Wait, it can actually inspect and tell me that without me googling the PID?"&lt;/em&gt;&lt;/p&gt;
&lt;h3&gt;
  
  
  3. Port Conflict Detection &amp;amp; Inspection
&lt;/h3&gt;


&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;python3 magnifier.py &lt;span class="nt"&gt;--ports&lt;/span&gt;
python3 magnifier.py port 5432   &lt;span class="c"&gt;# AI explains what is listening and if it is safe to restart&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;   Port │      PID │ State      │ Process Name
─────────┼──────────┼────────────┼────────────────────────────
   3000 │    50686 │ LISTEN     │ Python
   5000 │      991 │ LISTEN     │ ControlCenter
   7000 │      991 │ LISTEN     │ ControlCenter
   8000 │    96793 │ LISTEN     │ Python
   8080 │    19642 │ LISTEN     │ Python
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;

&lt;h3&gt;
  
  
  4. The Game Changer: Multi-Language Dependency Doctor
&lt;/h3&gt;


&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;python3 magnifier.py check-deps
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;


&lt;p&gt;It auto-detects and scans manifest files across your repo, queries the runtime environment, and highlights version drift:&lt;/p&gt;

&lt;blockquote&gt;
&lt;p&gt;&lt;em&gt;"Found &lt;code&gt;requirements.txt&lt;/code&gt;: requests &lt;code&gt;&amp;gt;=2.31.0&lt;/code&gt;, but installed environment has &lt;code&gt;requests==2.25.1&lt;/code&gt;. Run &lt;code&gt;pip install --upgrade -r requirements.txt&lt;/code&gt;."&lt;/em&gt;&lt;/p&gt;
&lt;/blockquote&gt;

&lt;p&gt;Supported ecosystems:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Python:&lt;/strong&gt; &lt;code&gt;requirements.txt&lt;/code&gt;, &lt;code&gt;pyproject.toml&lt;/code&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Node.js:&lt;/strong&gt; &lt;code&gt;package.json&lt;/code&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Rust:&lt;/strong&gt; &lt;code&gt;Cargo.toml&lt;/code&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Go:&lt;/strong&gt; &lt;code&gt;go.mod&lt;/code&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Ruby:&lt;/strong&gt; &lt;code&gt;Gemfile&lt;/code&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;PHP:&lt;/strong&gt; &lt;code&gt;composer.json&lt;/code&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Java:&lt;/strong&gt; &lt;code&gt;pom.xml&lt;/code&gt;, &lt;code&gt;build.gradle&lt;/code&gt;
&lt;/li&gt;
&lt;/ul&gt;
&lt;h3&gt;
  
  
  5. Seamless Command Execution &amp;amp; AI Diagnosis
&lt;/h3&gt;

&lt;p&gt;Run risky terminal commands through Magnifier:&lt;br&gt;
&lt;/p&gt;
&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;python3 magnifier.py run &lt;span class="nt"&gt;--&lt;/span&gt; npm run build
python3 magnifier.py run &lt;span class="nt"&gt;--&lt;/span&gt; cargo build
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;


&lt;p&gt;Or pipe failed command logs directly:&lt;br&gt;
&lt;/p&gt;
&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;cargo build 2&amp;gt;&amp;amp;1 | python3 magnifier.py &lt;span class="nt"&gt;--errors&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;


&lt;p&gt;If a command fails, Gemma analyzes the stderr output, cross-references installed tooling versions, and gives a direct 3-step solution to fix it.&lt;/p&gt;


&lt;h2&gt;
  
  
  Code &amp;amp; Repository
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;GitHub Repository:&lt;/strong&gt; &lt;a href="https://github.com/chourasiavinit9-dev/framework-magnifier" rel="noopener noreferrer"&gt;chourasiavinit9-dev/framework-magnifier&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;CI / Workflows:&lt;/strong&gt; Tested with &lt;a href="https://github.com/chourasiavinit9-dev/framework-magnifier/actions" rel="noopener noreferrer"&gt;GitHub Actions CI&lt;/a&gt; across macOS, Ubuntu Linux, and Windows for Python 3.10, 3.11, and 3.12.&lt;/li&gt;
&lt;/ul&gt;
&lt;h3&gt;
  
  
  Quick Start
&lt;/h3&gt;


&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;git clone https://github.com/chourasiavinit9-dev/framework-magnifier.git
&lt;span class="nb"&gt;cd &lt;/span&gt;framework-magnifier
pip &lt;span class="nb"&gt;install&lt;/span&gt; &lt;span class="nt"&gt;-r&lt;/span&gt; requirements.txt
&lt;span class="nb"&gt;export &lt;/span&gt;&lt;span class="nv"&gt;GEMINI_API_KEY&lt;/span&gt;&lt;span class="o"&gt;=&lt;/span&gt;&lt;span class="s2"&gt;"your_api_key_here"&lt;/span&gt;  &lt;span class="c"&gt;# Free tier via Google AI Studio&lt;/span&gt;
python3 magnifier.py
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;

&lt;h3&gt;
  
  
  Docker Container (Zero-Install)
&lt;/h3&gt;


&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;docker build &lt;span class="nt"&gt;-t&lt;/span&gt; framework-magnifier &lt;span class="nb"&gt;.&lt;/span&gt;
docker run &lt;span class="nt"&gt;--rm&lt;/span&gt; &lt;span class="nt"&gt;-it&lt;/span&gt; &lt;span class="nt"&gt;--pid&lt;/span&gt;&lt;span class="o"&gt;=&lt;/span&gt;host &lt;span class="nt"&gt;-e&lt;/span&gt; &lt;span class="nv"&gt;GEMINI_API_KEY&lt;/span&gt;&lt;span class="o"&gt;=&lt;/span&gt;&lt;span class="s2"&gt;"your_key"&lt;/span&gt; framework-magnifier
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;h2&gt;
  
  
  Why Open-Source AI? Why Gemma?
&lt;/h2&gt;

&lt;p&gt;I deliberately chose not to build this around a closed proprietary model. Here is why open innovation and open weights matter:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;strong&gt;User Privacy &amp;amp; System Security:&lt;/strong&gt; System diagnostics require reading process trees, open ports, and package configurations. Sending raw internal process tables or proprietary workspace manifests to third-party closed servers creates significant compliance and privacy risks. With open weights like &lt;strong&gt;&lt;a href="https://ai.google.dev/gemma" rel="noopener noreferrer"&gt;Gemma-4&lt;/a&gt;&lt;/strong&gt;, developers have complete agency over where data flows.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Predictable Free Access for Students:&lt;/strong&gt; Avijit is a student. He can't afford pay-per-token meters or surprise credit card charges while debugging code. Google's generous open access via &lt;a href="https://aistudio.google.com" rel="noopener noreferrer"&gt;Google AI Studio&lt;/a&gt; ensures any developer or student can use this for free.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Model Swappability &amp;amp; Local Offline Use:&lt;/strong&gt; Because Framework Magnifier is built on an open architecture, you are never locked into one provider. Users can switch models via environment variables or run completely air-gapped and offline using &lt;a href="https://ollama.com" rel="noopener noreferrer"&gt;Ollama&lt;/a&gt;:
&lt;/li&gt;
&lt;/ol&gt;
&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;&lt;span class="nb"&gt;export &lt;/span&gt;&lt;span class="nv"&gt;MAGNIFIER_MODEL&lt;/span&gt;&lt;span class="o"&gt;=&lt;/span&gt;&lt;span class="s2"&gt;"gemma-4-26b-a4b-it"&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;h2&gt;
  
  
  How It Solved It for Avijit
&lt;/h2&gt;

&lt;p&gt;I sent the first build to Avijit over the weekend. Within twenty minutes, he messaged me:&lt;/p&gt;

&lt;blockquote&gt;
&lt;p&gt;&lt;em&gt;"Bro, it immediately flagged that my terminal was executing Python 3.9 from an old Homebrew symlink while my project virtual environment was expecting Python 3.11. I spent three hours banging my head against the wall on this last week."&lt;/em&gt;&lt;/p&gt;
&lt;/blockquote&gt;

&lt;p&gt;Seeing him run &lt;code&gt;python3 magnifier.py check-deps&lt;/code&gt; and immediately unblock himself made the entire project worth building.&lt;/p&gt;


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

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Best Use of Gemma ($200 Featured Track):&lt;/strong&gt; Framework Magnifier relies directly on Google's open-weight &lt;strong&gt;&lt;a href="https://ai.google.dev/gemma" rel="noopener noreferrer"&gt;Gemma-4&lt;/a&gt;&lt;/strong&gt; model family to power intelligent system reasoning, process introspection, multi-language version drift identification, and automated command stderr triage without telemetry leaking to closed commercial models.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Overall Hacktoberfest Challenge:&lt;/strong&gt; Built from scratch during the hackathon window specifically for a real friend (&lt;strong&gt;Avijit&lt;/strong&gt;) solving an everyday developer barrier using open-source AI.&lt;/li&gt;
&lt;/ul&gt;


&lt;h2&gt;
  
  
  Agent Session &amp;amp; Build Transcript
&lt;/h2&gt;

&lt;p&gt;Framework Magnifier was designed and developed with pair programming assistance via DevRelay and Google Antigravity. Explore the interactive transcript below to see the reasoning, testing, and tool calls:&lt;/p&gt;


&lt;div class="ltag-agent-session"&gt;
  &lt;div class="agent-session-header"&gt;
    
    &lt;span class="agent-session-tool-icon-badge" title="Gemini CLI"&gt;
&lt;/span&gt;
    &lt;span class="agent-session-title"&gt;Building Framework Magnifier: AI Terminal Doctor with Gemma-4&lt;/span&gt;
  &lt;/div&gt;

  &lt;div class="agent-session-scroll"&gt;

      &lt;div class="agent-session-message agent-session-user"&gt;
        &lt;div class="agent-session-role-badge agent-session-role-user"&gt;
          You
        &lt;/div&gt;
        &lt;div class="agent-session-content"&gt;
                &lt;div class="agent-session-text"&gt;
                  &lt;p&gt;I want to build an open-source AI terminal doctor for my friend Avijit who struggles with mismatched dependencies, orphaned processes, and cryptic terminal errors. It should monitor system processes, open ports, and diagnose dependency mismatches across multiple languages using Google Gemma-4.&lt;/p&gt;

                &lt;/div&gt;
        &lt;/div&gt;
      &lt;/div&gt;

      &lt;div class="agent-session-message agent-session-assistant"&gt;
        &lt;div class="agent-session-role-badge agent-session-role-assistant"&gt;
          Agent
        &lt;/div&gt;
        &lt;div class="agent-session-content"&gt;
                &lt;div class="agent-session-text"&gt;
                  &lt;p&gt;Let's architect Framework Magnifier using Python, psutil for system telemetry, and Google Gemma-4 for root-cause diagnosis. We will implement CPU/RAM inspection, port conflict scanning, multi-language manifest analysis (requirements.txt, package.json, Cargo.toml, go.mod, pom.xml), and automated command failure triage.&lt;/p&gt;

                &lt;/div&gt;
              &lt;div class="agent-session-tool-call"&gt;
                
                  &lt;span class="agent-session-chevron"&gt;▸&lt;/span&gt;
                  &lt;span class="tool-name tool-name-writetofile"&gt;write_to_file&lt;/span&gt;
                    &lt;span class="tool-input-preview"&gt;magnifier.py&lt;/span&gt;
                
                &lt;div class="agent-session-tool-detail"&gt;
                    &lt;div class="tool-section"&gt;
                      &lt;div class="tool-section-label"&gt;Input&lt;/div&gt;
                      &lt;pre class="agent-session-pre"&gt;&lt;code&gt;magnifier.py&lt;/code&gt;&lt;/pre&gt;
                    &lt;/div&gt;
                    &lt;div class="tool-section"&gt;
                      &lt;div class="tool-section-label"&gt;Output&lt;/div&gt;
                      &lt;pre class="agent-session-pre"&gt;&lt;code&gt;Created Framework Magnifier CLI with psutil and Gemma-4 integration&lt;/code&gt;&lt;/pre&gt;
                    &lt;/div&gt;
                &lt;/div&gt;
              &lt;/div&gt;
        &lt;/div&gt;
      &lt;/div&gt;

      &lt;div class="agent-session-message agent-session-user"&gt;
        &lt;div class="agent-session-role-badge agent-session-role-user"&gt;
          You
        &lt;/div&gt;
        &lt;div class="agent-session-content"&gt;
                &lt;div class="agent-session-text"&gt;
                  &lt;p&gt;Add high-contrast neon HUD styling and error badges so errors are immediately highlighted in bright red, warnings in yellow, and safe states in neon green.&lt;/p&gt;

                &lt;/div&gt;
        &lt;/div&gt;
      &lt;/div&gt;

      &lt;div class="agent-session-message agent-session-assistant"&gt;
        &lt;div class="agent-session-role-badge agent-session-role-assistant"&gt;
          Agent
        &lt;/div&gt;
        &lt;div class="agent-session-content"&gt;
                &lt;div class="agent-session-text"&gt;
                  &lt;p&gt;Implemented a zero-dependency ANSI neon color palette with bright red error badges, electric yellow PIDs, neon green health metrics, and electric cyan headers. Tested all modes across macOS and Linux.&lt;/p&gt;

                &lt;/div&gt;
              &lt;div class="agent-session-tool-call"&gt;
                
                  &lt;span class="agent-session-chevron"&gt;▸&lt;/span&gt;
                  &lt;span class="tool-name tool-name-runcommand"&gt;run_command&lt;/span&gt;
                    &lt;span class="tool-input-preview"&gt;python3 magnifier.py --ports&lt;/span&gt;
                
                &lt;div class="agent-session-tool-detail"&gt;
                    &lt;div class="tool-section"&gt;
                      &lt;div class="tool-section-label"&gt;Input&lt;/div&gt;
                      &lt;pre class="agent-session-pre agent-session-terminal"&gt;&lt;code&gt;$ python3 magnifier.py --ports&lt;/code&gt;&lt;/pre&gt;
                    &lt;/div&gt;
                    &lt;div class="tool-section"&gt;
                      &lt;div class="tool-section-label"&gt;Output&lt;/div&gt;
                      &lt;pre class="agent-session-pre"&gt;&lt;code&gt;Active network port scanner completed with neon ANSI tables&lt;/code&gt;&lt;/pre&gt;
                    &lt;/div&gt;
                &lt;/div&gt;
              &lt;/div&gt;
        &lt;/div&gt;
      &lt;/div&gt;
  &lt;/div&gt;

  &lt;div class="agent-session-footer"&gt;
    &lt;span class="agent-session-meta"&gt;
        4 of 4 messages
    &lt;/span&gt;
  &lt;/div&gt;
&lt;/div&gt;






&lt;h2&gt;
  
  
  Tech Stack
&lt;/h2&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Layer&lt;/th&gt;
&lt;th&gt;Technology&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Language&lt;/td&gt;
&lt;td&gt;&lt;a href="https://www.python.org" rel="noopener noreferrer"&gt;Python 3.10+&lt;/a&gt;&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;System Telemetry&lt;/td&gt;
&lt;td&gt;&lt;a href="https://github.com/giampaolo/psutil" rel="noopener noreferrer"&gt;psutil&lt;/a&gt;&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Open AI Core&lt;/td&gt;
&lt;td&gt;&lt;a href="https://ai.google.dev/gemma" rel="noopener noreferrer"&gt;Google Gemma-4&lt;/a&gt;&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Containerization&lt;/td&gt;
&lt;td&gt;
&lt;a href="https://www.docker.com" rel="noopener noreferrer"&gt;Docker&lt;/a&gt; (Multi-stage slim alpine/debian build)&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Continuous Integration&lt;/td&gt;
&lt;td&gt;
&lt;a href="https://github.com/chourasiavinit9-dev/framework-magnifier/actions" rel="noopener noreferrer"&gt;GitHub Actions&lt;/a&gt; (Linux, macOS, Windows)&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;




&lt;h2&gt;
  
  
  What's Next &amp;amp; Community Discussion
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;[ ] TUI dashboard with live auto-refresh using &lt;a href="https://github.com/Textualize/rich" rel="noopener noreferrer"&gt;rich&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;[ ] Auto-detection of corrupted &lt;code&gt;$PATH&lt;/code&gt; directory collisions&lt;/li&gt;
&lt;li&gt;[ ] Direct local offline inference fallback via Ollama&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Feel free to star or contribute on &lt;a href="https://github.com/chourasiavinit9-dev/framework-magnifier" rel="noopener noreferrer"&gt;GitHub&lt;/a&gt;!&lt;/p&gt;




&lt;h3&gt;
  
  
  Questions?
&lt;/h3&gt;

&lt;p&gt;&lt;em&gt;Have you ever lost hours to a bizarre terminal or version mismatch error? Drop your craziest environment bug in the comments below, or ask any questions about how the multi-language dependency scanner works!&lt;/em&gt;&lt;/p&gt;

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