<?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: Nikitha Bhandari</title>
    <description>The latest articles on DEV Community by Nikitha Bhandari (@nikitha_bhandari_2bd452ef).</description>
    <link>https://dev.to/nikitha_bhandari_2bd452ef</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%2F4187849%2F9f11f795-7cf7-4b15-a278-67ea2ba66840.png</url>
      <title>DEV Community: Nikitha Bhandari</title>
      <link>https://dev.to/nikitha_bhandari_2bd452ef</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/nikitha_bhandari_2bd452ef"/>
    <language>en</language>
    <item>
      <title>AfterHours — Some Signals Should Never Be Heard.</title>
      <dc:creator>Nikitha Bhandari</dc:creator>
      <pubDate>Sat, 31 Oct 2026 09:51:53 +0000</pubDate>
      <link>https://dev.to/nikitha_bhandari_2bd452ef/afterhours-some-signals-should-never-be-heard-1i19</link>
      <guid>https://dev.to/nikitha_bhandari_2bd452ef/afterhours-some-signals-should-never-be-heard-1i19</guid>
      <description>&lt;h1&gt;
  
  
  AfterHours — Some Signals Should Never Be Heard.
&lt;/h1&gt;

&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;AfterHours&lt;/strong&gt; is an AI-powered haunted shortwave radio mystery game inspired by the theme &lt;strong&gt;"Ghost in the Machine."&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;Imagine switching on an old radio late at night. Between the static and interference, you begin hearing strange broadcasts. At first, they sound like ordinary weather reports and emergency announcements. But as you tune into different frequencies, the transmissions begin contradicting one another.&lt;/p&gt;

&lt;p&gt;Something happened in the fictional town of Blackwood in 1988. And someone—or something—is trying to tell you the truth.&lt;/p&gt;

&lt;p&gt;In AfterHours, players become radio investigators. They manually tune across frequencies, detect signals, lock onto hidden transmissions, collect evidence and piece together the mystery behind the disappearance of Blackwood.&lt;/p&gt;

&lt;p&gt;The investigation features four radio stations:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;88.1 MHz — Blackwood Weather:&lt;/strong&gt; An ordinary weather report with an unsettling explanation for a mysterious incident.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;91.7 MHz — Civil Defense:&lt;/strong&gt; An emergency evacuation transmission that contradicts the first broadcast.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;103.4 MHz — Police Dispatch:&lt;/strong&gt; Evidence suggesting that the authorities knew more than they admitted.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;107.9 MHz — Station Zero:&lt;/strong&gt; A hidden transmission that reveals the truth behind Blackwood's disappearance.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Once all the clues are discovered, players must make a decision that determines one of two possible endings.&lt;/p&gt;

&lt;p&gt;I interpreted "Ghost in the Machine" as the idea that technology can become a vessel for something mysterious. In AfterHours, an ordinary radio becomes the medium through which a forgotten story tries to reach the living.&lt;/p&gt;

&lt;h2&gt;
  
  
  Screenshots
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;Screenshots:&lt;/strong&gt; &lt;br&gt;
&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Ff2ytxs1qw4k81vkmo5nj.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%2Ff2ytxs1qw4k81vkmo5nj.png" alt=" " width="800" height="375"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Ff6dqt5s9odd4oxnal5ak.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%2Ff6dqt5s9odd4oxnal5ak.png" alt=" " width="800" height="372"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Ftoy25a7ogrbcd0m74nly.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%2Ftoy25a7ogrbcd0m74nly.png" alt=" " width="800" height="346"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fmkwnw58y97nrjzptmjir.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%2Fmkwnw58y97nrjzptmjir.png" alt=" " width="800" height="354"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;The demo showcases:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;The cinematic opening screen.&lt;/li&gt;
&lt;li&gt;Frequency tuning and signal detection.&lt;/li&gt;
&lt;li&gt;Discovering and locking onto hidden transmissions.&lt;/li&gt;
&lt;li&gt;Collecting clues and uncovering the mystery.&lt;/li&gt;
&lt;li&gt;Experiencing the two different endings.&lt;/li&gt;
&lt;/ul&gt;

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

&lt;p&gt;&lt;strong&gt;GitHub Repository:&lt;/strong&gt;&lt;br&gt;&lt;br&gt;
&lt;a href="https://github.com/nikkisbhandari123-coder/AfterHours" rel="noopener noreferrer"&gt;https://github.com/nikkisbhandari123-coder/AfterHours&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;The project is open source. Contributions and suggestions are welcome!&lt;/p&gt;

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

&lt;p&gt;AfterHours is built using Python and Streamlit, with a focus on local AI inference and open-source technology.&lt;/p&gt;

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

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Python:&lt;/strong&gt; Core game logic, mystery progression, signal detection and ending management.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Streamlit:&lt;/strong&gt; Interactive radio interface and game experience.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Gemma 3 4B:&lt;/strong&gt; Open-weight language model used through Ollama for generating atmospheric radio fragments, mysterious whispers and ambient transmissions.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Ollama:&lt;/strong&gt; Runs the language model locally, allowing AI inference without depending on a hosted language-model API.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Custom CSS:&lt;/strong&gt; Vintage shortwave radio aesthetics, CRT effects, scanlines, static-inspired visuals and cinematic backgrounds.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Procedural Audio:&lt;/strong&gt; Pure Python audio generation for radio static, interference, tuning sounds and carrier-lock effects.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;ElevenLabs:&lt;/strong&gt; Optional text-to-speech narration for radio broadcasts.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;One important design decision was separating AI-generated atmosphere from the actual mystery.&lt;/p&gt;

&lt;p&gt;The canonical broadcasts, clues, investigation progression and endings are controlled by Python. Gemma generates atmospheric content, but it cannot invent canonical evidence, change story facts or determine the player's ending.&lt;/p&gt;

&lt;p&gt;This keeps the mystery consistent while still allowing AI to make the radio experience feel unpredictable.&lt;/p&gt;

&lt;p&gt;The application also includes fallback mechanisms so the core experience remains available when Ollama or ElevenLabs is unavailable.&lt;/p&gt;

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

&lt;p&gt;Open innovation made it possible to experiment with AI without making the entire project dependent on a closed API.&lt;/p&gt;

&lt;p&gt;Using Gemma through Ollama allowed me to explore local inference and integrate an open-weight model directly into the game.&lt;/p&gt;

&lt;p&gt;It also gave me more control over how AI interacts with the application. Rather than allowing the model to control the entire story, I could use it for creative atmospheric generation while keeping the important game mechanics deterministic and reliable.&lt;/p&gt;

&lt;p&gt;Open-source tools made experimentation more accessible, gave me control over the development process and helped me understand how different components of an AI-powered application work together.&lt;/p&gt;

&lt;p&gt;For a project built around mysterious signals and hidden communication, the ability to run AI locally felt especially fitting.&lt;/p&gt;

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

&lt;p&gt;[Add your DevRelay agent session link here, if available.]&lt;/p&gt;

&lt;p&gt;I used AI coding assistance during development to help implement and test different parts of the application, including frequency detection, story progression, AI integration and procedural audio.&lt;/p&gt;

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

&lt;p&gt;The most interesting part of AfterHours is the moment when the player realizes that the broadcasts are not independent.&lt;/p&gt;

&lt;p&gt;A weather report mentions a mudslide. A civil defense transmission describes an armed evacuation. A police dispatch reveals that the evacuation was planned before the supposed disaster.&lt;/p&gt;

&lt;p&gt;The radio slowly stops feeling like a machine receiving random signals.&lt;/p&gt;

&lt;p&gt;It starts feeling like something is trying to tell you what really happened.&lt;/p&gt;

&lt;p&gt;And when Station Zero finally appears, the player has to decide whether the truth should be broadcast to the world or buried forever.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Some signals should never be heard. But some secrets refuse to stay silent.&lt;/strong&gt;&lt;/p&gt;

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

&lt;p&gt;[Select the applicable partner prize categories from the official challenge page. Potentially relevant areas include open-weight AI and local inference, depending on the published eligibility criteria.]&lt;/p&gt;

</description>
      <category>devchallenge</category>
      <category>hf26challenge</category>
    </item>
    <item>
      <title>SlothFix: A Local AI Debugger That Thinks Beyond the Error</title>
      <dc:creator>Nikitha Bhandari</dc:creator>
      <pubDate>Sun, 25 Oct 2026 06:38:54 +0000</pubDate>
      <link>https://dev.to/nikitha_bhandari_2bd452ef/slothfix-a-local-ai-debugger-that-thinks-beyond-the-error-2pfj</link>
      <guid>https://dev.to/nikitha_bhandari_2bd452ef/slothfix-a-local-ai-debugger-that-thinks-beyond-the-error-2pfj</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;Debugging can take longer than fixing the actual bug. First you have to understand the error, figure out why it happened, work out the safest change, and then consider what that change might break elsewhere.&lt;/p&gt;

&lt;p&gt;I built &lt;strong&gt;SlothFix&lt;/strong&gt;, a local-first AI debugging assistant that brings those steps together in one place.&lt;/p&gt;

&lt;p&gt;You provide a code snippet and an error or unexpected behavior. SlothFix uses a locally running Gemma model to help explain the root cause, suggest a minimal patch, identify possible downstream effects, and generate a regression test suggestion.&lt;/p&gt;

&lt;p&gt;The part I wanted to focus on was what happens &lt;em&gt;after&lt;/em&gt; the immediate fix. A change might resolve an exception but still affect a caller, an API response, or a data-handling path. SlothFix's “Butterfly Effect” analysis is meant to help developers think about those possibilities before applying a change.&lt;/p&gt;

&lt;p&gt;The goal isn't to replace debugging skills or code review. It's to reduce the repetitive work involved in understanding a bug and give developers a useful starting point.&lt;/p&gt;

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

&lt;p&gt;SlothFix currently runs locally using Streamlit and Ollama.&lt;/p&gt;

&lt;p&gt;🎥 Watch the demo: &lt;a href="https://screenapp.io/app/share/psRvllgVdgdIQdla12LRKOeY" rel="noopener noreferrer"&gt;https://screenapp.io/app/share/psRvllgVdgdIQdla12LRKOeY&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;📸 Screenshots: &lt;br&gt;
&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fiov6g42elsgrwgv59l82.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%2Fiov6g42elsgrwgv59l82.png" alt=" " width="799" height="418"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Note: The app currently runs locally, so viewers need to follow the setup instructions to run it themselves.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Project repository:&lt;/strong&gt; &lt;a href="https://github.com/nikkisbhandari123-coder/slothfix" rel="noopener noreferrer"&gt;SlothFix on GitHub&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;A screenshot or short recording of the app can be added here to show the debugging workflow in action.&lt;/p&gt;

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

&lt;p&gt;&lt;strong&gt;GitHub repository:&lt;/strong&gt;&lt;br&gt;&lt;br&gt;
&lt;a href="https://github.com/nikkisbhandari123-coder/slothfix" rel="noopener noreferrer"&gt;https://github.com/nikkisbhandari123-coder/slothfix&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;The repository includes the application code, setup instructions, and automated tests.&lt;/p&gt;

&lt;p&gt;Setup guide: &lt;a href="https://github.com/nikkisbhandari123-coder/slothfix#readme" rel="noopener noreferrer"&gt;https://github.com/nikkisbhandari123-coder/slothfix#readme&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;The project is written in Python and organized into separate modules for the interface, model communication, debugging logic, prompts, and response parsing.&lt;/p&gt;

&lt;p&gt;The automated test suite currently passes 13 tests covering the debugging service and response parser.&lt;/p&gt;

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

&lt;p&gt;I built SlothFix with Python, Streamlit, Ollama, and Google's Gemma model.&lt;/p&gt;

&lt;p&gt;Streamlit provides the interface. The debugging service prepares the request, the Ollama client communicates with the local model, and the response parser organizes the generated result into sections that are easier to review.&lt;/p&gt;

&lt;p&gt;I also separated the interface into reusable components instead of keeping everything in one large application file. This makes the project easier to maintain and gives the individual pieces their own tests.&lt;/p&gt;

&lt;p&gt;One design decision was to keep the debugging workflow focused on generating suggestions. SlothFix does not automatically execute submitted code or generated tests on the host machine. The proposed patch and regression tests still need to be reviewed and tested separately.&lt;/p&gt;

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

&lt;p&gt;For a debugging tool, the code being analyzed can be the most sensitive part of the request.&lt;/p&gt;

&lt;p&gt;Sending source code and stack traces to a hosted AI service may not be appropriate for every developer or organization. Running Gemma through Ollama gives SlothFix a local inference path, so the code sent to the model stays within the local application and Ollama setup rather than being sent to a cloud AI API by SlothFix.&lt;/p&gt;

&lt;p&gt;Open-weight models also give developers more control over the model they run and the environment in which they run it. Once the model and dependencies are installed, the core workflow can operate without cloud inference or an internet connection.&lt;/p&gt;

&lt;p&gt;That doesn't automatically make every local setup secure, and the quality of the results still depends on the model. But it makes local AI-assisted debugging a practical option for people who want more control over their code and tools.&lt;/p&gt;

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

&lt;p&gt;The payoff I'm aiming for is less time spent switching between an error message, a debugging explanation, a proposed fix, and a test-writing step.&lt;/p&gt;

&lt;p&gt;Instead of starting from scratch for each of those tasks, a developer gets a structured first pass in one place. They can then spend their time checking the reasoning and deciding whether the change is safe.&lt;/p&gt;

&lt;p&gt;I haven't measured a reliable time-saving figure yet, so I won't claim a specific number. The next step is to try it against a set of real debugging cases and compare how useful its explanations, patches, and test suggestions are.&lt;/p&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; — SlothFix uses Google's Gemma model through Ollama for local debugging assistance.&lt;/li&gt;
&lt;/ul&gt;

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