<?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: Ashish Anand</title>
    <description>The latest articles on DEV Community by Ashish Anand (@ashishanand).</description>
    <link>https://dev.to/ashishanand</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%2F4152864%2F77e209ca-899c-4008-b783-8d08a110d706.png</url>
      <title>DEV Community: Ashish Anand</title>
      <link>https://dev.to/ashishanand</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/ashishanand"/>
    <language>en</language>
    <item>
      <title>Corvus - Hacktoberfest '26</title>
      <dc:creator>Ashish Anand</dc:creator>
      <pubDate>Sat, 31 Oct 2026 12:10:04 +0000</pubDate>
      <link>https://dev.to/ashishanand/corvus-hacktoberfest-26-2mhg</link>
      <guid>https://dev.to/ashishanand/corvus-hacktoberfest-26-2mhg</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;&lt;strong&gt;Corvus&lt;/strong&gt; is a haunted-terminal AI workspace that runs &lt;strong&gt;Gemma&lt;/strong&gt; entirely on your own computer. You open it, pick a Gemma model, and talk to it. The reply streams in word by word from your own machine, inside an interface that looks like a forgotten laboratory terminal decorated for the Day of the Dead.&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%2Fmdt7cdijzdjti3w2tf4i.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%2Fmdt7cdijzdjti3w2tf4i.png" alt="Corvus Homepage" width="800" height="364"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;I read "Ghost in the Machine" literally. The ghost is Gemma, an intelligence that lives inside your laptop instead of in a distant data center. The look blends two traditions. Halloween brings the mystery: obsidian black, scanlines, ravens and ember light. Día de los Muertos brings the warmth: marigolds, papel picado banners and the idea that remembering is a way of keeping someone alive. That second idea shaped the &lt;strong&gt;Memory Vault&lt;/strong&gt;, where every conversation is kept locally so you can reopen it later.&lt;/p&gt;

&lt;p&gt;What it does:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Streams real Gemma responses through Ollama, with stop, regenerate and retry&lt;/li&gt;
&lt;li&gt;Detects the models you have installed and prefers a Gemma model automatically&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Works with other Ollama models too.&lt;/strong&gt; Gemma is the default, but any chat model you've pulled into Ollama appears in the model picker, so you can switch to Llama, Mistral, Qwen or others without changing any code&lt;/li&gt;
&lt;li&gt;Saves, searches, renames, reopens and exports every session in the Memory Vault&lt;/li&gt;
&lt;li&gt;Includes &lt;strong&gt;Whisper Studio&lt;/strong&gt; for prompt experiments with a custom system instruction, temperature and token limit&lt;/li&gt;
&lt;li&gt;Shows honest diagnostics: real connection status, latency, request counts and the last error&lt;/li&gt;
&lt;li&gt;Lets you slide the whole interface between Halloween and Día de los Muertos&lt;/li&gt;
&lt;li&gt;Gives you slots for your own images and videos, all controlled from one config file&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Known hauntings 👻&lt;/strong&gt; The machine isn’t fully at peace. On any dimension (device), there are some scary mood changes of machine (bugs added intentionally). I’ve documented few honestly instead of hiding them, so if the ghost misbehaves for you, it’s not your setup.&lt;/li&gt;
&lt;/ul&gt;

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

&lt;p&gt;  &lt;iframe src="https://www.youtube.com/embed/tdJlAea8G5I" width="710" height="399"&gt;
  &lt;/iframe&gt;
&lt;/p&gt;

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


&lt;div class="ltag-github-readme-tag"&gt;
  &lt;div class="readme-overview"&gt;
    &lt;h2&gt;
      &lt;img src="https://assets.dev.to/assets/github-logo-5a155e1f9a670af7944dd5e12375bc76ed542ea80224905ecaf878b9157cdefc.svg" alt="GitHub logo"&gt;
      &lt;a href="https://github.com/ASAN-11" rel="noopener noreferrer"&gt;
        ASAN-11
      &lt;/a&gt; / &lt;a href="https://github.com/ASAN-11/Corvus" rel="noopener noreferrer"&gt;
        Corvus
      &lt;/a&gt;
    &lt;/h2&gt;
    &lt;h3&gt;
      Meet the ghosts of intelligence with open AI.
    &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;Corvus&lt;/h1&gt;
&lt;/div&gt;
&lt;blockquote&gt;
&lt;p&gt;&lt;em&gt;Meet the ghosts of intelligence with open AI.&lt;/em&gt;&lt;/p&gt;
&lt;/blockquote&gt;
&lt;p&gt;Corvus is a local-first AI workspace for &lt;a href="https://ollama.com" rel="nofollow noopener noreferrer"&gt;Ollama&lt;/a&gt; and open-weight models such as Gemma. It pairs a haunted-terminal Halloween look with the marigolds and papel picado of Día de los Muertos, and it is built with plain HTML, CSS and JavaScript plus a small Python launcher that has no dependencies.&lt;/p&gt;
&lt;p&gt;Chats stream from a model running on your own computer. There are no API keys, no accounts, no analytics and no hosted AI service involved.&lt;/p&gt;
&lt;p&gt;Built for Hacktoberfest under the challenge &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;/p&gt;
&lt;p&gt;&lt;a rel="noopener noreferrer" href="https://github.com/ASAN-11/Corvus/screenshot.png"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fraw.githubusercontent.com%2FASAN-11%2FCorvus%2FHEAD%2Fscreenshot.png" alt="corvus screenshot"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;div class="markdown-heading"&gt;
&lt;h2 class="heading-element"&gt;Table of contents&lt;/h2&gt;
&lt;/div&gt;
&lt;ul&gt;
&lt;li&gt;&lt;a href="https://github.com/ASAN-11/Corvus#features" rel="noopener noreferrer"&gt;Features&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/ASAN-11/Corvus#how-it-works" rel="noopener noreferrer"&gt;How it works&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/ASAN-11/Corvus#quick-start" rel="noopener noreferrer"&gt;Quick start&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/ASAN-11/Corvus#using-corvus" rel="noopener noreferrer"&gt;Using Corvus&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/ASAN-11/Corvus#adding-your-own-media" rel="noopener noreferrer"&gt;Adding your own media&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/ASAN-11/Corvus#configuration" rel="noopener noreferrer"&gt;Configuration&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/ASAN-11/Corvus#project-structure" rel="noopener noreferrer"&gt;Project structure&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/ASAN-11/Corvus#privacy-and-security" rel="noopener noreferrer"&gt;Privacy and security&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/ASAN-11/Corvus#offline-use-and-pwa-install" rel="noopener noreferrer"&gt;Offline use and PWA install&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/ASAN-11/Corvus#troubleshooting" rel="noopener noreferrer"&gt;Troubleshooting&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/ASAN-11/Corvus#known-limitations" rel="noopener noreferrer"&gt;Known limitations&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/ASAN-11/Corvus#roadmap" rel="noopener noreferrer"&gt;Roadmap&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/ASAN-11/Corvus#contributing" rel="noopener noreferrer"&gt;Contributing&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/ASAN-11/Corvus#acknowledgements" rel="noopener noreferrer"&gt;Acknowledgements&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;

&lt;div class="markdown-heading"&gt;
&lt;h2 class="heading-element"&gt;Features&lt;/h2&gt;
&lt;/div&gt;
&lt;p&gt;&lt;strong&gt;AI workspace (The Séance)&lt;/strong&gt;&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;Real streaming chat with Ollama's &lt;code&gt;/api/chat&lt;/code&gt; endpoint, parsing its newline-delimited JSON as…&lt;/li&gt;
&lt;/ul&gt;&lt;/div&gt;
  &lt;/div&gt;
  &lt;div class="gh-btn-container"&gt;&lt;a class="gh-btn" href="https://github.com/ASAN-11/Corvus" rel="noopener noreferrer"&gt;View on GitHub&lt;/a&gt;&lt;/div&gt;
&lt;/div&gt;


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

&lt;p&gt;&lt;strong&gt;Gemma is the engine of the whole project.&lt;/strong&gt; Everything else in Corvus exists to give it a home.&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Model:&lt;/strong&gt; Google's open-weight &lt;strong&gt;Gemma&lt;/strong&gt;, run locally through &lt;strong&gt;Ollama&lt;/strong&gt;. Corvus reads the models you have installed and prefers any Gemma tag it finds, so it works with whichever size fits your hardware instead of hardcoding one.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Other models:&lt;/strong&gt; because Corvus talks to Ollama's standard &lt;code&gt;/api/chat&lt;/code&gt; endpoint, it isn't tied to Gemma. Pull any chat model with &lt;code&gt;ollama pull &amp;lt;model&amp;gt;&lt;/code&gt;, click &lt;strong&gt;Refresh models&lt;/strong&gt;, and it shows up in the picker. Behavior varies by model, since each has its own style, context size and speed, and some Whisper Studio options may be handled differently depending on the model and your Ollama version. Gemma stays the default and the model the app is designed around.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Streaming:&lt;/strong&gt; the app calls &lt;code&gt;/api/chat&lt;/code&gt; and parses the newline-delimited JSON stream as it arrives. Network chunks don't always line up with JSON lines, so the parser buffers partial lines.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Steering the model:&lt;/strong&gt; the system prompt is editable, and Whisper Studio passes real Ollama options (&lt;code&gt;temperature&lt;/code&gt;, &lt;code&gt;num_predict&lt;/code&gt;) so you can see how settings change its voice.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Frontend:&lt;/strong&gt; plain HTML, CSS and vanilla JavaScript with no framework. It has custom SVG art, theme variables, hash-based routing, and a small Markdown renderer that escapes model output before formatting it, because model text is untrusted.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Backend:&lt;/strong&gt; a dependency-free Python launcher serves the app on &lt;code&gt;127.0.0.1&lt;/code&gt; and proxies only three allow-listed Ollama routes. This avoids browser CORS problems without exposing Ollama to the network, and it never logs conversations.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Memory:&lt;/strong&gt; chats are stored in the browser's local storage. They are unencrypted, and the app says so.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Honest status:&lt;/strong&gt; the "Connected" badge only appears after a real request succeeds. There are separate states for Disconnected and Model unavailable. If a model isn't installed, Corvus shows the install command and never silently switches to another service or downloads anything without your consent.&lt;/li&gt;
&lt;/ul&gt;

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

&lt;p&gt;A closed API would have made this a different, smaller project. Because Gemma and the other models Ollama serves are open-weight:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Your conversations never leave your machine.&lt;/strong&gt; For an app built around memory and remembrance, that is the point.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;It costs nothing per message.&lt;/strong&gt; There is no API key, quota or bill, so anyone with a laptop can run it.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;You choose the model, not a vendor.&lt;/strong&gt; Use Gemma at whatever size fits your hardware, or switch to another Ollama model you prefer. Nobody can retire it, reprice it or change it underneath you.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;It can work without internet.&lt;/strong&gt; Once Ollama, a model and the app files are installed, chatting needs no connection.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Anyone can inspect and modify it.&lt;/strong&gt; The app is plain files and a short Python script, so people can see exactly where their prompts go.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;There are honest limits. You need enough RAM for the model you choose. Open-weight is not the same as fully open source, and every model has its own terms of use that you should read before commercial use. Corvus is local-first software built around open models, and it makes no claim that everything is unrestricted.&lt;/p&gt;

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

&lt;p&gt;Día de los Muertos is about keeping the people we love close by remembering them. I wanted a machine that felt like that: not a cold chat box, but a candlelit archive where conversations are kept instead of discarded, and where the intelligence answering you lives in your own home, on your own hardware.&lt;/p&gt;

&lt;p&gt;The moment I built it for is the first reply. You press Enter, three small lights pulse, and words begin to appear from a model running on the computer in front of you, with no cloud in between.&lt;/p&gt;

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

&lt;p&gt;&lt;strong&gt;Best Use of Gemma.&lt;/strong&gt; Gemma is the core of the project, not an add-on. It runs locally through Ollama, Corvus automatically prefers and selects installed Gemma models, and every feature (chat, Whisper Studio experiments, the Memory Vault) is built around talking to it. Other Ollama models are supported as a flexibility bonus, but Gemma is the default and the model the experience is designed for. The app is built around what makes an open-weight Gemma valuable: privacy, zero per-message cost, and full user control over where inference happens.&lt;/p&gt;

</description>
      <category>devchallenge</category>
      <category>hf26challenge</category>
    </item>
    <item>
      <title>Pigritia - Hacktoberfest '26</title>
      <dc:creator>Ashish Anand</dc:creator>
      <pubDate>Mon, 26 Oct 2026 04:05:48 +0000</pubDate>
      <link>https://dev.to/ashishanand/pigritia-hacktoberfest-26-21k6</link>
      <guid>https://dev.to/ashishanand/pigritia-hacktoberfest-26-21k6</guid>
      <description>&lt;p&gt;&lt;em&gt;This is a submission for the &lt;a href="https://dev.to/challenges/hacktoberfest-week3-2026-10-19"&gt;Hacktoberfest Open-Source AI Challenge Week 3: Slow and Steady&lt;/a&gt;&lt;/em&gt;&lt;/p&gt;

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

&lt;p&gt;&lt;strong&gt;Pigritia&lt;/strong&gt; (Latin for sloth, in the laziness sense) is a set of four lazy-day helpers that run entirely on my own machine. It's a single HTML file that talks to Ollama and Gemma. There's no backend, no account and no API key.&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%2Fjtgueukoifrob0zbmp7q.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%2Fjtgueukoifrob0zbmp7q.png" alt="Pigritia homepage" width="800" height="364"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Autopilot:&lt;/strong&gt; paste an email thread, meeting notes or a group chat. It gives back a short summary with a priority tag, the questions waiting on me, a task list with owners and due dates, a drafted reply in the tone I pick, and a calendar file (&lt;code&gt;.ics&lt;/code&gt;) for anything with a deadline.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Fridge list:&lt;/strong&gt; photograph the fridge and get a tickable shopping checklist, so nobody has to stand there with the door open.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Wind down:&lt;/strong&gt; a calm bedtime story, read aloud, with generated rain, ocean or forest sound underneath. It works offline.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;File tidier:&lt;/strong&gt; paste or pick messy file names and get a rename-and-sort plan, plus a ready-to-run &lt;code&gt;.sh&lt;/code&gt; or &lt;code&gt;.ps1&lt;/code&gt; script. Only names are read, and nothing moves until I run the script myself.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;The people who get to be lazier are anyone who drowns in group chats and inboxes, hates writing "sounds good, I'll send that Friday" for the hundredth time, or has a Downloads folder full of &lt;code&gt;final_FINAL_v2.pdf&lt;/code&gt;.&lt;/p&gt;

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

&lt;p&gt;&lt;a href="https://youtu.be/rWFGOD2IJyQ" rel="noopener noreferrer"&gt;https://youtu.be/rWFGOD2IJyQ&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Pigritia calls Ollama on &lt;code&gt;localhost&lt;/code&gt;, so a hosted copy can't reach a visitor's own model. A short screen recording is the honest way to show it. The repo has setup instructions if you want to run it yourself.&lt;/p&gt;

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


&lt;div class="ltag-github-readme-tag"&gt;
  &lt;div class="readme-overview"&gt;
    &lt;h2&gt;
      &lt;img src="https://assets.dev.to/assets/github-logo-5a155e1f9a670af7944dd5e12375bc76ed542ea80224905ecaf878b9157cdefc.svg" alt="GitHub logo"&gt;
      &lt;a href="https://github.com/ASAN-11" rel="noopener noreferrer"&gt;
        ASAN-11
      &lt;/a&gt; / &lt;a href="https://github.com/ASAN-11/Pigritia" rel="noopener noreferrer"&gt;
        Pigritia
      &lt;/a&gt;
    &lt;/h2&gt;
    &lt;h3&gt;
      Experience AI core but slow and steady.
    &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;Pigritia&lt;/h1&gt;
&lt;/div&gt;
&lt;p&gt;&lt;strong&gt;Experience AI core but slow and steady.&lt;/strong&gt;&lt;/p&gt;
&lt;p&gt;Pigritia is a small set of lazy-day helpers that run entirely on your own machine. It is a single HTML file that talks to &lt;a href="https://ollama.com" rel="nofollow noopener noreferrer"&gt;Ollama&lt;/a&gt; and an open-weight &lt;a href="https://ai.google.dev/gemma" rel="nofollow noopener noreferrer"&gt;Gemma&lt;/a&gt; model. There is no backend, no account, no build step and no API key. Your messages, photos and file names never leave your computer.&lt;/p&gt;
&lt;p&gt;Built for the &lt;a href="https://dev.to/challenges/hacktoberfest-week3-2026-10-19" rel="nofollow"&gt;Hacktoberfest Open-Source AI Challenge Week 3: Slow and Steady&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;&lt;a rel="noopener noreferrer" href="https://github.com/ASAN-11/Pigritia/screenshot.png"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fraw.githubusercontent.com%2FASAN-11%2FPigritia%2FHEAD%2Fscreenshot.png" alt="pigritia screenshot"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;div class="markdown-heading"&gt;
&lt;h2 class="heading-element"&gt;Contents&lt;/h2&gt;
&lt;/div&gt;
&lt;ul&gt;
&lt;li&gt;&lt;a href="https://github.com/ASAN-11/Pigritia#what-it-does" rel="noopener noreferrer"&gt;What it does&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/ASAN-11/Pigritia#why-open-models" rel="noopener noreferrer"&gt;Why open models&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/ASAN-11/Pigritia#requirements" rel="noopener noreferrer"&gt;Requirements&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/ASAN-11/Pigritia#quick-start" rel="noopener noreferrer"&gt;Quick start&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/ASAN-11/Pigritia#the-site-section-by-section" rel="noopener noreferrer"&gt;The site, section by section&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/ASAN-11/Pigritia#using-each-tool" rel="noopener noreferrer"&gt;Using each tool&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/ASAN-11/Pigritia#how-it-works" rel="noopener noreferrer"&gt;How it works&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/ASAN-11/Pigritia#privacy-and-data-handling" rel="noopener noreferrer"&gt;Privacy and data handling&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/ASAN-11/Pigritia#choosing-a-model" rel="noopener noreferrer"&gt;Choosing a model&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/ASAN-11/Pigritia#troubleshooting" rel="noopener noreferrer"&gt;Troubleshooting&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/ASAN-11/Pigritia#limitations" rel="noopener noreferrer"&gt;Limitations&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/ASAN-11/Pigritia#customising" rel="noopener noreferrer"&gt;Customising&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/ASAN-11/Pigritia#project-structure" rel="noopener noreferrer"&gt;Project structure&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;

&lt;div class="markdown-heading"&gt;
&lt;h2 class="heading-element"&gt;What it does&lt;/h2&gt;
&lt;/div&gt;
&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Tool&lt;/th&gt;
&lt;th&gt;You give it&lt;/th&gt;
&lt;th&gt;You get back&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;Autopilot&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;An email thread, meeting notes or a group chat&lt;/td&gt;
&lt;td&gt;A short summary, a priority tag, questions waiting on you, a task list with owners and due dates, a calendar file and a drafted&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;…&lt;/div&gt;
  &lt;/div&gt;
  &lt;div class="gh-btn-container"&gt;&lt;a class="gh-btn" href="https://github.com/ASAN-11/Pigritia" rel="noopener noreferrer"&gt;View on GitHub&lt;/a&gt;&lt;/div&gt;
&lt;/div&gt;


&lt;p&gt;The whole app is one file, &lt;code&gt;pigritia.html&lt;/code&gt;, plus a detailed README.&lt;/p&gt;

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

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Model:&lt;/strong&gt; Gemma, an open-weight model, served by &lt;strong&gt;Ollama&lt;/strong&gt;. Everything runs through two Ollama endpoints: &lt;code&gt;/api/tags&lt;/code&gt; to list installed models and &lt;code&gt;/api/chat&lt;/code&gt; for generation.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Streaming for free text:&lt;/strong&gt; the fridge list and the bedtime story stream token by token, so the text appears as it's written.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;JSON mode for structured work:&lt;/strong&gt; Autopilot and the File tidier ask Ollama for &lt;code&gt;format: "json"&lt;/code&gt; at a low temperature. They parse the result into tasks, a reply, or a rename plan. The page checks the plan before showing it, discarding file names the model made up.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Vision:&lt;/strong&gt; the fridge photo is shrunk in the browser and sent as a base64 image to a vision-capable Gemma.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Everything else is the browser:&lt;/strong&gt; &lt;code&gt;speechSynthesis&lt;/code&gt; reads the story aloud, the Web Audio API generates the ambient sound, and Blob downloads produce the scripts and calendar files. There are no dependencies, no build step and no external fonts or scripts.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Design:&lt;/strong&gt; a vertical sidebar, serif type, and a line-art sloth hanging from a branch.&lt;/li&gt;
&lt;/ul&gt;

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

&lt;p&gt;Three of the four tools handle things I wouldn't paste into someone else's server: my work email, my group chats and photos of my kitchen. With an open-weight model on Ollama, that data goes to &lt;code&gt;localhost&lt;/code&gt; and nowhere else. You can check this in the browser's network tab.&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;It works offline.&lt;/strong&gt; Once the model is pulled, no internet is needed. The story, the sounds and the voice all work with the Wi-Fi off.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Nothing is metered.&lt;/strong&gt; A tool you use without thinking can't have a per-token bill hanging over it.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;The model is swappable.&lt;/strong&gt; Setup lists whatever Ollama has installed. A smaller model works on an old laptop, and a larger one gives better replies. Nothing in the app is tied to one model.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;The behaviour is mine to change.&lt;/strong&gt; Every tool's prompt is plain text in one readable file, so I can change the tone, the rules or the output format without asking anyone.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;A closed API could do the summarising. It couldn't do it privately, offline and for free, all at once.&lt;/p&gt;

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

&lt;p&gt;I measured this by counting what I no longer do by hand, not by stopwatch.&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Autopilot&lt;/strong&gt; replaces four chores: reading a long thread to find what's being asked, writing down the tasks, putting the deadlines in a calendar, and drafting the reply. Now I paste the text and press one button. The calendar step is one &lt;code&gt;.ics&lt;/code&gt; download that imports every dated task at once.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;File tidier&lt;/strong&gt; replaces renaming and sorting files one at a time. I paste or pick the names, check the before and after table, and run one script. The script never overwrites an existing file, and I read the plan first.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Fridge list&lt;/strong&gt; replaces standing at an open fridge and typing a list into my phone. I take one photo and tick items off at the shop.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Wind down&lt;/strong&gt; replaces searching for a bedtime story or a sound track. I type an idea, press one button, and it reads the story aloud over rain.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;I pasted a 40-message family group chat about a trip and got back three tasks and a drafted reply; it took about 2 minutes including a quick read-through.&lt;/p&gt;

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

&lt;p&gt;&lt;strong&gt;Best Use of Gemma.&lt;/strong&gt; Gemma is the model behind every tool in Pigritia, running locally through Ollama:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Text generation:&lt;/strong&gt; streamed bedtime stories and fridge lists.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Structured output:&lt;/strong&gt; Autopilot and the File tidier ask Gemma for JSON at a low temperature and turn it into tasks, a drafted reply, calendar events and rename scripts.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Vision:&lt;/strong&gt; the Fridge list sends a photo to a vision-capable Gemma, such as &lt;code&gt;gemma3:4b&lt;/code&gt; or larger, so one open model handles both text and images.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Swappable size:&lt;/strong&gt; a smaller Gemma works on a modest laptop and a larger one gives better replies, all through one dropdown in Setup.&lt;/li&gt;
&lt;/ul&gt;

</description>
      <category>devchallenge</category>
      <category>hf26challenge</category>
    </item>
    <item>
      <title>Cygnus - Hacktoberfest '26</title>
      <dc:creator>Ashish Anand</dc:creator>
      <pubDate>Thu, 15 Oct 2026 13:56:00 +0000</pubDate>
      <link>https://dev.to/ashishanand/cygnus-hacktoberfest-26-1p4i</link>
      <guid>https://dev.to/ashishanand/cygnus-hacktoberfest-26-1p4i</guid>
      <description>&lt;p&gt;&lt;em&gt;This is a submission for the &lt;a href="https://dev.to/challenges/hacktoberfest-week2-2026-10-12"&gt;Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle&lt;/a&gt;&lt;/em&gt;&lt;/p&gt;

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

&lt;p&gt;&lt;strong&gt;Cygnus&lt;/strong&gt; is a small web app with three AI tools for people who handle messy paperwork and can't send it to a cloud AI:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Receipt Sorter&lt;/strong&gt; is for freelancers and bookkeepers. It reads pasted receipt text or a photo and returns vendor, date, total, tax, currency and category as table rows, with CSV export.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Listing Writer&lt;/strong&gt; is for Etsy and eBay sellers. It turns product notes and photos into a title, description, tags and a price range.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Menu Translator&lt;/strong&gt; is for local restaurants. It translates a menu but keeps the restaurant's own dish names and follows its glossary.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fkt3cb8hsm1yztopxvxxv.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%2Fkt3cb8hsm1yztopxvxxv.png" alt="Cygnus Home Section" width="800" height="365"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;The problem is the same in each case. The inputs hold client invoices, supplier prices and customer details, and paid APIs charge per document, which eats the margin of a small monthly subscription. Cygnus runs the model on the user's own computer, so nothing leaves it and there is no per-document bill.&lt;/p&gt;

&lt;p&gt;There is also a &lt;strong&gt;"Will it pay?"&lt;/strong&gt; calculator. You enter your customers, price and volume, and it compares monthly profit on a paid API with running the model locally. Its defaults are example numbers, so replace them with your own.&lt;/p&gt;

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

&lt;p&gt;Cygnus needs a local Ollama to talk to, so there is no hosted demo. Here is a walkthrough instead:&lt;/p&gt;

&lt;p&gt;&lt;a href="https://youtu.be/mmXypTU8JWk" rel="noopener noreferrer"&gt;https://youtu.be/mmXypTU8JWk&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;To run it yourself, follow the steps in the README (about five minutes).&lt;/p&gt;

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

&lt;p&gt;&lt;a href="https://github.com/ASAN-11/Cygnus" rel="noopener noreferrer"&gt;https://github.com/ASAN-11/Cygnus&lt;/a&gt;&lt;/p&gt;

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

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Model:&lt;/strong&gt; Gemma 3 through &lt;a href="https://ollama.com" rel="noopener noreferrer"&gt;Ollama&lt;/a&gt; is the default. The page calls Ollama's &lt;code&gt;/api/chat&lt;/code&gt; endpoint and streams the reply as it is generated.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Any other model works too.&lt;/strong&gt; The Model field lists whatever is installed in Ollama and also accepts a typed name, so you can try Llama, Qwen, Mistral or a vision model like LLaVA.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Vision:&lt;/strong&gt; photos are base64-encoded in the browser and sent to Gemma as image input, so a receipt or product photo needs no separate OCR step.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Structured output:&lt;/strong&gt; Receipt Sorter asks Ollama for JSON, parses it into table rows and builds the CSV in the browser.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;No backend:&lt;/strong&gt; the whole app is one HTML file. You serve it with &lt;code&gt;python3 -m http.server 8000&lt;/code&gt; and open &lt;code&gt;localhost:8000&lt;/code&gt;, because a page opened straight from disk is blocked from calling Ollama.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Design:&lt;/strong&gt; a vertical navbar with a branching drop-right menu, and all the graphics (the swan with circuit-trace antlers, the icons, the diagram) are SVG. The palette is white, yellow, orange and gold.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;I checked the app end to end against a mock Ollama server, so the page, streaming and table parsing work. Real-world accuracy on your own receipts or menus depends on the model you pick, and you should review extracted rows before using them for accounting.&lt;/p&gt;

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

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Privacy:&lt;/strong&gt; receipts and product photos go only to &lt;code&gt;localhost&lt;/code&gt;. A closed API would mean sending a client's financial documents to someone else's server.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Cost:&lt;/strong&gt; a local model has no per-token bill, so a $9 a month product can still make money. The calculator in the app lets you check that with your own numbers.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Control:&lt;/strong&gt; the model is a dropdown and the prompts are plain text, so you can swap models, adjust behavior or add a restaurant's glossary without waiting on a vendor.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Offline:&lt;/strong&gt; after the model is pulled, the tools need no internet connection.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;No lock-in:&lt;/strong&gt; there is no pricing change or deprecation that can break a product built this way.&lt;/li&gt;
&lt;/ul&gt;

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

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Best Use of Gemma:&lt;/strong&gt; Gemma 3 is the default model for every tool, run locally through Ollama, including its image input.&lt;/li&gt;
&lt;/ul&gt;

</description>
      <category>devchallenge</category>
      <category>hf26challenge</category>
    </item>
    <item>
      <title>Sagittarius - Hacktoberfest '26</title>
      <dc:creator>Ashish Anand</dc:creator>
      <pubDate>Mon, 12 Oct 2026 02:26:04 +0000</pubDate>
      <link>https://dev.to/ashishanand/sagittarius-hacktoberfest-26-2g1m</link>
      <guid>https://dev.to/ashishanand/sagittarius-hacktoberfest-26-2g1m</guid>
      <description>&lt;p&gt;&lt;em&gt;This is a submission for the &lt;a href="https://dev.to/challenges/hacktoberfest-week1-2026-10-05"&gt;Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass&lt;/a&gt;&lt;/em&gt;&lt;/p&gt;

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

&lt;p&gt;&lt;strong&gt;Sagittarius&lt;/strong&gt; is a local-first outdoor companion that runs as a single HTML file. You open it, check the sky, get a plan, and then close the laptop.&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Outing planner:&lt;/strong&gt; pulls today's forecast (temperature, rain chance, wind, UV, sunrise and sunset) and suggests a plan for a hike, run, birding trip, foliage photo walk, gardening session or picnic, sized to the time you have.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Field guide:&lt;/strong&gt; a chat with an open-weight model running on your own machine. Describe a bird call, a plant or a trail question and get an answer, even with no signal.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Garden planner:&lt;/strong&gt; enter your frost dates (or pick a season if you have no frost) and see what to sow this week. It works fully offline.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Field journal:&lt;/strong&gt; notes and photos saved only in your browser.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Outdoor timer and screen clock:&lt;/strong&gt; the page shows how long you've been on screen and logs the minutes you spend outside, so the screen stays the shortest part of the day.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;It's for anyone who wants to get outside more: hikers, birders, gardeners and runners, and especially people who live somewhere with patchy signal or who don't want their location history sent to a server.&lt;/p&gt;

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

&lt;p&gt;&lt;a href="https://youtu.be/AMe2pqgkEkM" rel="noopener noreferrer"&gt;https://youtu.be/AMe2pqgkEkM&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;The model runs through Ollama on your own machine, so there's no hosted link. To try it, follow the quick start in the repo README.&lt;/p&gt;

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

&lt;p&gt;&lt;a href="https://github.com/ASAN-11/Sagittarius" rel="noopener noreferrer"&gt;https://github.com/ASAN-11/Sagittarius&lt;/a&gt;&lt;/p&gt;

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

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Open-weight model:&lt;/strong&gt; Gemma, served locally by &lt;a href="https://ollama.com" rel="noopener noreferrer"&gt;Ollama&lt;/a&gt;. The page streams responses from Ollama's &lt;code&gt;/api/chat&lt;/code&gt; endpoint, and the model menu lists whatever models you've pulled (Gemma, Llama, Qwen and others).&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Frontend:&lt;/strong&gt; plain HTML, CSS and JavaScript in one file, with no build step and no dependencies.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Weather:&lt;/strong&gt; &lt;a href="https://open-meteo.com" rel="noopener noreferrer"&gt;Open-Meteo&lt;/a&gt;, which is free and needs no key. It's the only online part, and it's optional.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Storage:&lt;/strong&gt; browser &lt;code&gt;localStorage&lt;/code&gt; for the journal, frost dates and outdoor-time log.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Graceful fallback:&lt;/strong&gt; if Ollama isn't running, the planner still works from built-in rules. When the model is running, it writes a fuller plan on top of those rules, using the real forecast numbers.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Design:&lt;/strong&gt; a light-mode interface with a collapsible sidebar, and background media holders so you can add your own video or images.&lt;/li&gt;
&lt;/ul&gt;

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

&lt;p&gt;&lt;strong&gt;Best Use of Gemma:&lt;/strong&gt; Sagittarius runs Gemma locally through Ollama as its default field-guide model. It powers the streamed chat answers and the written outing plans, all on the user's own machine with no API key and no internet needed for the model.&lt;/p&gt;

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

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;It works with no signal.&lt;/strong&gt; The model lives on your laptop, so a bird-call question at the trailhead still gets an answer. A closed API would fail the moment you lose signal.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Your places stay private.&lt;/strong&gt; Where you hike, bird and garden is personal. Prompts go to &lt;code&gt;localhost&lt;/code&gt;, and journal entries never leave your device.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;You can change the model.&lt;/strong&gt; Swap Gemma for another model in one dropdown, rewrite the system prompt, or fine-tune on your own region's species list. No vendor decides what the app can do.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;It costs nothing to run.&lt;/strong&gt; No API keys, accounts or per-token fees, so someone can use it every day without thinking about a bill.&lt;/li&gt;
&lt;/ul&gt;

</description>
      <category>devchallenge</category>
      <category>hf26challenge</category>
    </item>
    <item>
      <title>Recipe Keeper - Hacktoberfest '26</title>
      <dc:creator>Ashish Anand</dc:creator>
      <pubDate>Mon, 05 Oct 2026 03:32:50 +0000</pubDate>
      <link>https://dev.to/ashishanand/recipe-keeper-hacktoberfest-26-42om</link>
      <guid>https://dev.to/ashishanand/recipe-keeper-hacktoberfest-26-42om</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;
  
  
  What I Built
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;Recipe Keeper&lt;/strong&gt; is a small web app that turns [Grandpa's / Name's] rambling voice memos into a printable family cookbook.&lt;/p&gt;

&lt;p&gt;I built it for my grandfather. He cooks entirely by feel: "a fistful of rice, no, more, and onions until it smells right." None of it is written down, and the recipes will be lost if nobody captures them. A normal recipe app would force exact measurements he never uses, and it would throw away the stories that make the food his.&lt;/p&gt;

&lt;p&gt;You paste in a transcript of what he said. A local AI model turns it into a clean recipe card with a title, ingredients, steps and notes. It follows two rules:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;It never invents quantities.&lt;/strong&gt; "A fistful" stays "a fistful".&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Unclear parts become questions to ask him&lt;/strong&gt;, such as "Ask Grandpa: how long is 'until it smells right'?"&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Every card also keeps a quote of his original words underneath. Each recipe has a photo slot for the dish, and the whole cookbook can be printed or saved as a PDF to give back to the family.&lt;/p&gt;

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

&lt;p&gt;&lt;a href="https://youtu.be/6sBEymtBEBc" rel="noopener noreferrer"&gt;https://youtu.be/6sBEymtBEBc&lt;/a&gt;&lt;/p&gt;

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

&lt;p&gt;&lt;a href="https://github.com/ASAN-11/Recipe-Keeper" rel="noopener noreferrer"&gt;https://github.com/ASAN-11/Recipe-Keeper&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;It's a single file, &lt;code&gt;recipe-keeper.html&lt;/code&gt;, with no build step and no dependencies.&lt;/p&gt;

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

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Model:&lt;/strong&gt; an open-weight model run locally with &lt;strong&gt;Ollama&lt;/strong&gt; (I used &lt;code&gt;llama3.2&lt;/code&gt;, and the app lets you pick any model you have installed).&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Inference:&lt;/strong&gt; the page calls Ollama's local &lt;code&gt;/api/chat&lt;/code&gt; endpoint. Responses are streamed so you see the model working, and JSON mode (&lt;code&gt;format: "json"&lt;/code&gt;) with a low temperature keeps the output a structured recipe.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Prompting:&lt;/strong&gt; the system prompt tells the model to use only what was said and never guess amounts. It also tells it to flag unclear steps as questions for the cook.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;App:&lt;/strong&gt; plain HTML, CSS and JavaScript. The cookbook is saved in the browser's local storage, and it can be exported or imported as a JSON backup. The hero section takes a background photo or video, and a print stylesheet makes a clean PDF.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Setup:&lt;/strong&gt; &lt;code&gt;ollama pull llama3.2&lt;/code&gt;, then serve the page from localhost (&lt;code&gt;python -m http.server&lt;/code&gt;) or start Ollama with &lt;code&gt;OLLAMA_ORIGINS="*"&lt;/code&gt;.&lt;/li&gt;
&lt;/ul&gt;

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

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Privacy:&lt;/strong&gt; these are family recipes and a grandparent's own voice and stories. With a local model, none of it goes to a server I don't control, and there's no account to make.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Works offline:&lt;/strong&gt; after the one-time model download, it runs with no internet. That matters in a kitchen or at a grandparent's house.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Free to run:&lt;/strong&gt; there are no API keys and no per-token cost, so I can process as many memos as I want.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Swappable and tunable:&lt;/strong&gt; I can change models from a dropdown, and later I could fine-tune one on the finished cookbook so the cards sound like him.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;I control the behavior:&lt;/strong&gt; the "never invent quantities" rule matters a lot for heirloom recipes. With an open model I own the prompt and settings, and no provider can change them under me.&lt;/li&gt;
&lt;/ul&gt;




&lt;p&gt;&lt;strong&gt;What my father said when I showed him:&lt;/strong&gt; "Good, Boy!!" He asked me to do with his scrambled eggs recipe and shahi panner and it worked.&lt;/p&gt;

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