DEV Community: Spliot S

Find an Open-Source Issue Nobody Has Claimed Yet

Spliot S — Tue, 06 Oct 2026 04:00:39 +0000

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass

Find Good First Issues That Are Actually Available

What I Built

Finding a good first issue on GitHub is easy. Finding one that is actually unassigned, relevant to your skills, and worth contributing to is much harder.

I built Hacktoberfest Starter to help developers find open-source issues that are actually available to work on.

You add your skills and proficiency levels, and the app searches for relevant issues, filters out issues that appear to be already claimed or unsuitable, and helps you identify the issues that are a good match.

The goal is simple: spend less time scrolling through GitHub and more time actually contributing to open source.

Demo

https://hacktoberfest-starter-xyz.streamlit.app/

Code

https://github.com/Adii0906/hacktoberfest-starter

How I Built It

I built the project around GitHub issue discovery and AI-assisted analysis.

The workflow takes a developer's skills and proficiency levels, finds potentially relevant open-source issues, analyzes the results, and ranks them so the developer can focus on issues that actually match their experience.

The main idea was to make issue discovery more intelligent than simply searching GitHub for good first issue.

The project was built with open-source tools and AI-assisted development, with the AI layer helping analyze and organize issue data into useful recommendations.

Why Does Open Innovation Matter?

Open source is one of the easiest ways to learn by building with real projects, but getting started can be intimidating.

There are thousands of issues across GitHub, and beginners often don't know which ones are genuinely available or suitable for their skill level.

This project itself is open source because I want other developers to be able to improve the way issues are discovered, ranked, and matched.

The same community that contributes to open-source projects can also improve the tool that helps people find those projects.

My Agent Session

Not included.

Prize Categories

Open-Source AI Challenge — Week 1: Touch Grass

A Cute Cat that Lives on Your Screen

Spliot S — Fri, 02 Oct 2026 05:08:30 +0000

I Built a Tiny Desk Cat That Knows When I’m Coding 🐱

What I Built

I built Mewly, a tiny pixel-art desktop companion for a friend who spends way too much time coding.

Instead of another productivity dashboard or notification-heavy app, Mewly just lives on the desktop and reacts to what you're doing.

When you're actively coding, Mewly moves around your screen. When you stop for a while, it notices the inactivity and eventually goes to sleep. You can let it move automatically or take control yourself.

The idea was simple: make long coding sessions feel a little less lonely without adding another distracting app to the workflow.

📦 Download Mewly:

https://github.com/Adii0906/Mewly/releases/download/v1.0.0/Mewly.exe

Code

The entire project is open source:

🔗 GitHub:

https://github.com/Adii0906/Mewly

Feel free to explore the code, try Mewly yourself, or contribute.

How I Built It

Mewly is built as a lightweight Windows desktop application designed to quietly run in the background without getting in the way.

I designed its behavior around simple activity states:

Coding → Active → Idle → Sleeping

It includes:

🐱 Automatic movement by default
🖱️ Manual movement when you want control
⏱️ Inactivity detection
😴 Automatic sleeping after being idle
🖥️ Desktop interaction
⚡ Lightweight background execution

I wanted the implementation to stay lightweight instead of turning a tiny desktop companion into another resource-heavy application.

Why Does Open Innovation Matter?

Mewly is open source because the fun part of a desktop companion is being able to change it.

Someone can modify the animations, add new behaviors, connect it to their own workflow, create new characters, or completely change what happens when they're away from their computer.

Because the source is open, people can experiment with the idea instead of treating the application as a finished product.

I Built a Meeting Copilot That Remembers What Everyone Said

Spliot S — Fri, 02 Oct 2026 05:03:22 +0000

This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend

What I Built

I built Fathom Rebuild, an AI meeting copilot for a friend who spends a lot of time in meetings and often needs to go back through transcripts to find specific decisions, action items, or things someone mentioned.

Instead of manually searching through long meeting transcripts, the app lets you ask questions across your meetings and get answers with references back to the relevant transcript.

It has two main workflows:

Meeting Analysis — processes a meeting transcript and extracts useful information.
Ask Across Meetings — lets you ask questions across multiple meetings and find relevant information.

The goal was simple: make meeting history something you can actually talk to instead of something you have to search through.

Demo

🎥 Walkthrough:

Walkthrough.mp4 - Google Drive

drive.google.com

🚀 Deployed App:

vercel.com

Code

The project is completely open source:

🔗 GitHub:

Adii0906 / fathom-rebuild

Cadence: meeting intelligence

A Fathom-style meeting assistant built for a 24-hour assignment: record → transcript → AI summary, decisions, action items and highlights, plus search and "ask across all meetings" with cited answers.

Stack: Next.js 15 (App Router) · TypeScript · Tailwind CSS + shadcn/ui · LangChain + LangGraph · Groq · optional Supabase/Postgres.

The real meeting-capture layer (Zoom/Meet/Teams bot) is intentionally stubbed. See Capture layer.

Run it locally

Prerequisites: Node 20.9+ (Node 22 recommended) and a free Groq API key.

git clone https://github.com/Adii0906/fathom-rebuild && cd fathom-rebuild
git checkout claude/happy-hopper-jytoyu      # until this is merged to main

npm install

cp .env.example .env.local                   # then edit it
#   GROQ_API_KEY=gsk_...                     (required for AI generation, never commit this)

npm run dev                                  # http://localhost:3000

The app is fully populated on first load, because the seed data needs no API key or database. GROQ_API_KEY is only used when you:

switch a meeting…

View on GitHub

The repository contains the application code as well as the agent development logs.

How I Built It

I built the application around an agentic workflow using LangGraph and LangChain, with Groq for LLM inference.

The architecture is split into separate workflows for different tasks instead of putting everything into one large prompt.

The main flow looks roughly like:

Transcript → Processing → Analysis → Structured Information

and for cross-meeting questions:

Question → Retrieval → Relevant Meeting Context → Agent Reasoning → Cited Answer

I also added references to the source transcript so the user can trace an answer back to where the information came from.

The goal was to use the AI agent for actual reasoning and workflow orchestration rather than simply sending a transcript to an LLM and asking it to summarize.

Why Does Open Innovation Matter?

Open innovation made it possible for me to build the entire agent workflow around tools and frameworks that I could inspect, modify, and experiment with.

Using open-source frameworks like LangGraph and LangChain meant I could control how the different AI steps were connected instead of treating the intelligence as a black box.

It also makes the project easier for someone else to extend — whether that's adding new meeting analysis nodes, changing the retrieval strategy, adding another model provider, or building completely new workflows on top of the existing system.

Prize Categories

Open Source / Hacktoberfest
Build for a Friend

🚀 Getting Started with Machine Learning (Made Super Simple)

Spliot S — Mon, 25 Aug 2025 18:34:16 +0000

Let’s learn the 3 main types of Machine Learning:

Supervised, Unsupervised, and Reinforcement Learning.

We’ll explain them in a fun way you’ll never forget! 🧠✨

📘 Supervised Learning

Think back to 1st standard.

Your teacher gave you a book with pictures: 🐱 🐶 🦈 🐢 🪑

You looked at each picture, and the teacher told you the answer:

“This is a 🐶 dog.”
“This is a 🐱 cat.”

Over time, your brain learned the patterns.

That’s called Supervised Learning.

👉 You see the picture.

👉 You get the answer.

👉 You learn from the answers.

🌍 Unsupervised Learning

Now imagine you go to a new city.

No teacher. 🚫👩‍🏫

No map. 🗺️

No translator. 🌐

You walk around and notice things yourself:

👔 Some people wear suits and go into tall buildings.
🍳 Some wear aprons and stand near food stalls.
🚓 Some wear uniforms and direct traffic.

Nobody tells you who they are.

But your brain starts grouping them: office workers, chefs, police officers.

That’s called Unsupervised Learning.

👉 You learn by looking for patterns on your own.

👉 No teacher. No labels. Just observing.

🎮 Reinforcement Learning

Now think about playing a video game 🎮.

At first, you don’t know what to do.

You try moving, jumping, exploring.

If you do something good ✅ (like collecting a coin), the game gives you points or a reward.
If you do something bad ❌ (like falling in a hole), the game takes away points or you lose a life.

Over time, you learn what actions give rewards and which ones cause trouble.

That’s Reinforcement Learning.

👉 Learn by trial and error.

👉 Get rewards for good moves.

👉 Avoid punishments for bad moves.

🏆 Easy Way to Remember

Learning Type	Example Story 📖	How You Learn 🧠
📘 Supervised	Teacher shows 🐱🐶 pictures	Learn from answers/labels ✅
🌍 Unsupervised	New city, no guide 🚶	Find patterns on your own 🔍
🎮 Reinforcement	Video game with points & lives 🎮	Learn by trial & error 🎯

✨ That’s it!

Now you know the 3 big types of Machine Learning:

Supervised = Teacher 👩‍🏫
Unsupervised = Explorer 🕵️
Reinforcement = Gamer 🎮

And you’ll remember them for life! 🚀

🧠 DocMind_KB: Ask Your Documents Anything Using MindsDB, Hugging Face (No OpenAI Needed)

Spliot S — Sun, 15 Jun 2025 09:11:35 +0000

What is DocMind_KB?

🧠 Lightweight Local Knowledge Base

A 100% local knowledge base that lets you upload CSV or PDF files and ask questions in natural language. Your documents are processed into searchable text chunks, stored in PostgreSQL, and indexed by MindsDB for semantic search. All AI models run locally—no API keys, no cloud, just fast, private answers on your machine.

🔍 How It Works

Simple Explanation:

Upload Documents → Your files (CSV/PDF) get converted to text chunks
Store in Database → Text chunks saved in PostgreSQL (your local database)
Create AI Index → MindsDB reads your data and creates "smart embeddings" (mathematical understanding of meaning)
Ask Questions → When you ask something, the system finds relevant chunks based on meaning (not just keywords)
Generate Answers → Local AI model combines relevant information to create natural language answers

Why This Works:

PostgreSQL = Your filing cabinet (stores the actual documents)
MindsDB = Your smart librarian (understands what documents mean and finds relevant ones)
Local AI = Your assistant (reads relevant documents and answers your questions)

🤖 AI Models (340MB Total)

Embeddings: multi-qa-MiniLM-L6-cos-v1 (QA-optimized, 80MB) – Finds relevant information for your questions
Question Answering: distilbert-base-cased-distilled-squad (260MB) – Extracts precise answers from your documents
Lightweight & Fast – Runs on CPU, no GPU required

Link to https://github.com/Adii0906/DocMind_KB/tree/mainthe Project

Research Buddy An Websearch Agent

Spliot S — Mon, 14 Apr 2025 11:43:49 +0000

A streamlined web research agent powered by Mistral AI that automatically finds, retrieves, and summarizes information from relevant sources based on your research queries.