<?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: Kanan</title>
    <description>The latest articles on DEV Community by Kanan (@kananpm07).</description>
    <link>https://dev.to/kananpm07</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%2F4180324%2F278a8d6f-8c59-4a06-9679-ef0c5cc4822f.png</url>
      <title>DEV Community: Kanan</title>
      <link>https://dev.to/kananpm07</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/kananpm07"/>
    <language>en</language>
    <item>
      <title>RepoSéance 👻 — Bringing Abandoned Code Back to Life with Local AI</title>
      <dc:creator>Kanan</dc:creator>
      <pubDate>Sat, 31 Oct 2026 18:19:35 +0000</pubDate>
      <link>https://dev.to/kananpm07/reposeance-bringing-abandoned-code-back-to-life-with-local-ai-3815</link>
      <guid>https://dev.to/kananpm07/reposeance-bringing-abandoned-code-back-to-life-with-local-ai-3815</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;Meet RepoSéance — an AI-powered digital archaeologist that brings abandoned software projects back to life.&lt;/p&gt;

&lt;p&gt;Every developer has projects they started with excitement but eventually abandoned. Behind every unfinished repository is a forgotten idea, incomplete code, experiments, and sometimes a story worth recovering.&lt;/p&gt;

&lt;p&gt;RepoSéance explores these forgotten repositories and gives them a digital afterlife.&lt;/p&gt;

&lt;p&gt;Users can upload a ZIP file of an old project or provide a local Git repository. The application analyzes its source code, README files, TODO comments, dependencies, and Git commit history to reconstruct the project's development journey.&lt;/p&gt;

&lt;p&gt;It then creates a digital ghost of the repository that users can interact with.&lt;/p&gt;

&lt;p&gt;Key features:&lt;/p&gt;

&lt;p&gt;Summon a Repository: Analyze an abandoned software project.&lt;/p&gt;

&lt;p&gt;Ghost Memory: Reconstruct the project's development timeline using Git history.&lt;/p&gt;

&lt;p&gt;Talk to the Ghost: Ask questions about the project's original purpose, implementation, and unfinished features.&lt;/p&gt;

&lt;p&gt;Resurrection Mode: Identify incomplete modules, unresolved TODOs, and potential improvements.&lt;/p&gt;

&lt;p&gt;Final Message: Generate a personalized farewell or resurrection message based on the project's history.&lt;/p&gt;

&lt;p&gt;The inspiration came from the idea that software projects, just like memories, shouldn't disappear simply because someone stopped working on them.&lt;/p&gt;

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

&lt;p&gt;RepoSéance is designed around open-source AI and local inference.&lt;/p&gt;

&lt;p&gt;Technology stack:&lt;/p&gt;

&lt;p&gt;Python — Core application logic.&lt;/p&gt;

&lt;p&gt;Streamlit — Interactive web interface.&lt;/p&gt;

&lt;p&gt;Ollama — Local AI model execution.&lt;/p&gt;

&lt;p&gt;Gemma 3 4B — Open-weight language model for repository analysis and storytelling.&lt;/p&gt;

&lt;p&gt;GitPython — Accessing Git repository history.&lt;/p&gt;

&lt;p&gt;JSON — Structured AI analysis and responses.&lt;/p&gt;

&lt;p&gt;The application follows a simple workflow:&lt;/p&gt;

&lt;p&gt;The user uploads an abandoned project.&lt;/p&gt;

&lt;p&gt;The backend extracts and analyzes repository files.&lt;/p&gt;

&lt;p&gt;Git history and project structure are examined.&lt;/p&gt;

&lt;p&gt;Relevant information is provided to the locally running AI model.&lt;/p&gt;

&lt;p&gt;The AI reconstructs the project's history and generates insights.&lt;/p&gt;

&lt;p&gt;Users can interact with the repository's digital ghost and explore ways to revive it.&lt;/p&gt;

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

&lt;p&gt;A developer's source code can contain personal projects, experimental ideas, credentials, and intellectual property.&lt;/p&gt;

&lt;p&gt;Sending an entire repository to a third-party AI service just to understand it introduces privacy concerns and external dependencies.&lt;/p&gt;

&lt;p&gt;RepoSéance explores a different approach.&lt;/p&gt;

&lt;p&gt;By using open-weight models through Ollama, the project can run AI inference locally. This means repository contents can remain on the user's machine rather than being sent to a remote AI API.&lt;/p&gt;

&lt;p&gt;Open models also provide flexibility. Developers can experiment with different models, customize prompts, and adapt the application without being locked into a single proprietary AI provider.&lt;/p&gt;

&lt;p&gt;Another important advantage is accessibility: local inference avoids per-request API charges, making experimentation more affordable.&lt;/p&gt;

&lt;p&gt;I believe forgotten software deserves a second chance, and open-source AI makes it possible to explore that history without giving away ownership of the code.&lt;/p&gt;

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

&lt;p&gt;Every abandoned repository is like a digital graveyard of unfinished ideas.&lt;/p&gt;

&lt;p&gt;Some projects were abandoned because of deadlines. Others because their creators lost interest, ran out of resources, or simply moved on.&lt;/p&gt;

&lt;p&gt;But what if those projects could tell us their stories?&lt;/p&gt;

&lt;p&gt;RepoSéance gives these forgotten projects a voice, helping developers rediscover their original ideas and potentially bring them back to life.&lt;/p&gt;

&lt;p&gt;Because sometimes, the best way to build something new is to listen to something we left behind. 👻&lt;/p&gt;

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

&lt;p&gt;Main category.&lt;/p&gt;

</description>
      <category>devchallenge</category>
      <category>hf26challenge</category>
    </item>
    <item>
      <title>LifeAdmin-AI: Turning Documents into Actionable Tasks with Local AI</title>
      <dc:creator>Kanan</dc:creator>
      <pubDate>Sun, 25 Oct 2026 21:40:21 +0000</pubDate>
      <link>https://dev.to/kananpm07/lifeadmin-ai-turning-documents-into-actionable-tasks-with-local-ai-16nk</link>
      <guid>https://dev.to/kananpm07/lifeadmin-ai-turning-documents-into-actionable-tasks-with-local-ai-16nk</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;LifeAdmin-AI is a document assistant designed to turn notices and other text documents into actionable tasks.&lt;/p&gt;

&lt;p&gt;Instead of manually reading an entire notice and trying to remember every deadline, users can upload a PDF or TXT file, or paste text, and get a structured summary of the tasks, deadlines, priorities, and supporting evidence.&lt;/p&gt;

&lt;p&gt;The goal is simple: make everyday administrative work easier and reduce the chance of overlooking an important deadline.&lt;/p&gt;

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

&lt;p&gt;LifeAdmin-AI currently runs locally on my Windows computer using Streamlit and Ollama with the gemma3:4b model.&lt;/p&gt;

&lt;p&gt;To try the demo, the app and Ollama need to be running on the same computer. &lt;br&gt;
&lt;/p&gt;
&lt;div class="crayons-card c-embed text-styles text-styles--secondary"&gt;
    &lt;div class="c-embed__content"&gt;
      &lt;div class="c-embed__body"&gt;
        &lt;h2 class="fs-xl lh-tight"&gt;
          &lt;a href="https://drive.google.com/file/d/1JvEiQcm8QOGzJTsJAwhmpytTozxIoASg/view?usp=sharing" rel="noopener noreferrer" class="c-link"&gt;
            demo.mp4 - Google Drive
          &lt;/a&gt;
        &lt;/h2&gt;
        &lt;div class="color-secondary fs-s flex items-center"&gt;
            &lt;img alt="favicon" class="c-embed__favicon m-0 mr-2 radius-0" src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fssl.gstatic.com%2Fdocs%2Fdoclist%2Fimages%2Fdrive_favicon_2026_32dp.png" width="32" height="32"&gt;
          drive.google.com
        &lt;/div&gt;
      &lt;/div&gt;
    &lt;/div&gt;
&lt;/div&gt;


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

&lt;p&gt;GitHub repository: &lt;a href="https://github.com/kanan-pm07/LifeAdmin-AI" rel="noopener noreferrer"&gt;https://github.com/kanan-pm07/LifeAdmin-AI&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;The repository contains the application code and setup instructions.&lt;/p&gt;

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

&lt;p&gt;I built LifeAdmin-AI using:&lt;/p&gt;

&lt;p&gt;Python for application logic&lt;/p&gt;

&lt;p&gt;Streamlit for the user interface&lt;/p&gt;

&lt;p&gt;Ollama and Gemma 3 (gemma3:4b) for local language-model inference&lt;/p&gt;

&lt;p&gt;pypdf for extracting text from PDF documents&lt;/p&gt;

&lt;p&gt;The application asks the model to return structured JSON containing a summary, tasks, and uncertainties. Python validation checks the response structure before the results are displayed.&lt;/p&gt;

&lt;p&gt;I also focused on grounding extracted tasks in evidence from the original document and flagging uncertain details instead of silently treating them as facts.&lt;/p&gt;

&lt;p&gt;Running the model locally means the application does not need a hosted model API for its current demo setup.&lt;/p&gt;

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

&lt;p&gt;Open-source tools make it possible to experiment with AI without depending entirely on closed, hosted APIs.&lt;/p&gt;

&lt;p&gt;Using Ollama and an open-weight model gave me a practical way to explore local inference, structured outputs, and validation while keeping the application accessible for further development.&lt;/p&gt;

&lt;p&gt;It also helped me understand that building an AI application is not just about generating a response. The application needs to validate that response and make uncertainty visible to the user.&lt;/p&gt;

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

&lt;p&gt;One of the biggest challenges was getting consistent structured JSON from a language model.&lt;/p&gt;

&lt;p&gt;I learned to combine prompt instructions with Python-side validation instead of relying on the model's output alone. I also learned the importance of testing deadlines, evidence excerpts, and ambiguous instructions with sample documents.&lt;/p&gt;

&lt;p&gt;LifeAdmin AI is still a work in progress, and I plan to keep improving its reliability and usability.&lt;/p&gt;

&lt;h2&gt;
  
  
  What's Next?
&lt;/h2&gt;

&lt;p&gt;Test with more types of notices and documents&lt;br&gt;
Improve handling of ambiguous deadlines and conditional tasks&lt;br&gt;
Make the demo easier for others to try&lt;br&gt;
Continue refining validation and error handling&lt;br&gt;
Thanks for checking out LifeAdmin AI!&lt;/p&gt;

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

&lt;p&gt;LifeAdmin AI is designed to reduce the effort of reading long notices and manually extracting tasks and deadlines.&lt;/p&gt;

&lt;p&gt;Instead of scanning a document, identifying action items, and rewriting them into a checklist, users can upload a PDF or TXT file and receive a structured summary with tasks, deadlines, priorities, and supporting evidence.&lt;/p&gt;

&lt;p&gt;The biggest benefit is reducing repetitive reading and organizing. The user still needs to verify important dates and ambiguous instructions, but the first pass becomes much easier.&lt;/p&gt;

&lt;p&gt;I'm currently testing the app locally and plan to measure its accuracy and time savings as I test it with more documents.&lt;/p&gt;

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

&lt;h3&gt;
  
  
  Best Use of Gemma
&lt;/h3&gt;

&lt;p&gt;LifeAdmin AI uses Google's open-weight Gemma 3 (&lt;code&gt;gemma3:4b&lt;/code&gt;) model through Ollama for local document analysis. The model extracts tasks, deadlines, priorities, and supporting evidence into structured JSON, while Python validation checks the response before displaying it.&lt;/p&gt;

&lt;p&gt;Using Gemma locally makes it possible to experiment with document automation without relying on a hosted model API for inference in the current demo setup.&lt;/p&gt;

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