<?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: Rupak Chatterjee</title>
    <description>The latest articles on DEV Community by Rupak Chatterjee (@rupak_chatterjee9799).</description>
    <link>https://dev.to/rupak_chatterjee9799</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%2F3781067%2Fb3ac4692-bcbd-4dab-a575-214a8331eab5.jpg</url>
      <title>DEV Community: Rupak Chatterjee</title>
      <link>https://dev.to/rupak_chatterjee9799</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/rupak_chatterjee9799"/>
    <language>en</language>
    <item>
      <title>Echoes: Bringing Memories Back to Life with Open-Source AI</title>
      <dc:creator>Rupak Chatterjee</dc:creator>
      <pubDate>Fri, 30 Oct 2026 19:06:46 +0000</pubDate>
      <link>https://dev.to/rupak_chatterjee9799/echoes-bringing-memories-back-to-life-with-open-source-ai-2n3p</link>
      <guid>https://dev.to/rupak_chatterjee9799/echoes-bringing-memories-back-to-life-with-open-source-ai-2n3p</guid>
      <description>&lt;h2&gt;
  
  
  What I Built
&lt;/h2&gt;

&lt;p&gt;I built &lt;strong&gt;Echoes&lt;/strong&gt;, an AI-powered memory preservation and storytelling application inspired by the theme "Ghost in the Machine."&lt;/p&gt;

&lt;p&gt;The idea is simple: memories and personal stories deserve to be preserved.&lt;/p&gt;

&lt;p&gt;Echoes aims to help people turn personal memories and stories into meaningful written narratives using open-source AI.&lt;/p&gt;

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

&lt;p&gt;The application uses Python and Streamlit for its interface and an open-weight Gemma model through Ollama for local AI inference.&lt;/p&gt;

&lt;p&gt;The AI workflow is designed to transform user-provided memories into structured stories while preserving the original information.&lt;/p&gt;

&lt;p&gt;The application is built around local inference rather than depending on a proprietary hosted AI API.&lt;/p&gt;

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

&lt;p&gt;Open innovation gives developers greater control over how AI applications are built and operated.&lt;/p&gt;

&lt;p&gt;Using open-weight models and local inference makes it possible to process personal stories on a user's own machine without sending their content to a third-party hosted AI service.&lt;/p&gt;

&lt;p&gt;It also makes model experimentation, customization, and independent development more accessible.&lt;/p&gt;

&lt;p&gt;For a memory-preservation application, privacy and control over personal stories are especially important.&lt;/p&gt;

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

&lt;p&gt;Best Use of Gemma — include only if Gemma is meaningfully integrated and the category requirements are satisfied.&lt;/p&gt;

&lt;h2&gt;
  
  
  Closing Thoughts
&lt;/h2&gt;

&lt;p&gt;The past should not disappear just because time moves forward. With open-source AI, we can build tools that help preserve stories and memories while keeping users in control of their data.&lt;/p&gt;

</description>
      <category>devchallenge</category>
      <category>hf26challenge</category>
    </item>
    <item>
      <title>Lazybrief</title>
      <dc:creator>Rupak Chatterjee</dc:creator>
      <pubDate>Sun, 25 Oct 2026 05:55:08 +0000</pubDate>
      <link>https://dev.to/rupak_chatterjee9799/lazybrief-2eek</link>
      <guid>https://dev.to/rupak_chatterjee9799/lazybrief-2eek</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;LazyBrief turns meeting notes, study material, and technical text into a summary, key insights, and a checklist. It runs Gemma 3:1B through the Ollama API on the configured endpoint, which defaults to localhost. There is no hosted AI provider or paid key in the inference path.&lt;/p&gt;

&lt;p&gt;I created this project during the Hacktoberfest Open-Source AI Challenge Week 3 window. The goal is straightforward: spend less time rereading notes before getting back to the work they describe.&lt;/p&gt;

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

&lt;p&gt;&lt;a href="https://dev-to-uploads.s3.us-east-2.amazonaws.com/uploads/articles/gnijxjpwrec5wayug1gl.png" rel="noopener noreferrer"&gt;Image description&lt;/a&gt;&lt;/p&gt;

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

&lt;p&gt;(&lt;a href="https://github.com/Rupakrc9776/lazybrief" rel="noopener noreferrer"&gt;https://github.com/Rupakrc9776/lazybrief&lt;/a&gt;.)&lt;/p&gt;

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

&lt;p&gt;The interface is a Streamlit app. The Python modules validate text and UTF-8 TXT/Markdown uploads, build a prompt that treats document contents as untrusted data, send one structured request to Ollama, and validate the returned JSON.&lt;/p&gt;

&lt;p&gt;Source-based insights are shown as exact sentences selected by the model's source-segment IDs; Python retrieves the original sentence rather than displaying a model-written paraphrase as fact. Action items use a separate conservative rule: Python shows only source sentences containing explicit task or commitment wording, unchanged. That avoids inventing a new owner, deadline, or commitment, although implicit tasks can be missed. AI interpretations are labelled separately, and the model can still misunderstand the source.&lt;/p&gt;

&lt;p&gt;The app checks that the configured model is installed, gives the default localhost endpoint, and requires a typed confirmation before sending text to a non-local endpoint. Unit tests mock HTTP calls. I also ran the sample through a real local gemma3:1b model; that smoke test demonstrates the request path, not general summary accuracy.&lt;/p&gt;

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

&lt;p&gt;An open-weight model that runs through a local service gives people more control over where their notes go and which model they use. LazyBrief's default inference endpoint is on the user's own computer, and the source code can be inspected or adapted. A user who changes the endpoint should understand that the chosen server will receive the text. I have not verified that every aspect of an Ollama installation runs offline, so I do not describe LazyBrief as fully offline.&lt;/p&gt;

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

&lt;p&gt;Potential fit: Best Use of Gemma. Gemma 3:1B is the actual model used for the generated summaries and insights, not an incidental dependency. I will check the current category-specific requirements before entering.&lt;/p&gt;

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