DEV Community: Rishi Yadav

GhostLore AI — The Haunted Story Keeper

Rishi Yadav — Wed, 28 Oct 2026 18:18:56 +0000

What I Built

Meet GhostLore AI, a spooky interactive storytelling website inspired by the idea that some stories never truly disappear.

For the Hacktoberfest Week 4 theme, Ghost in the Machine, I wanted to build something that combines the mystery of Halloween with the power of open-source AI.

GhostLore AI allows users to:

👻 Create spooky stories from their own ideas.

🌙 Choose different atmospheres, from eerie mysteries to folklore and remembrance.

💻 Explore a haunted terminal that responds to questions.

📚 Save stories in a personal archive directly in their browser.

📥 Copy stories or download them as text files.

🔒 Connect a local open-weight AI model to generate stories without sending prompts to a hosted AI service.

The goal is simple: make storytelling fun while giving users more control over their data and creative tools.

Live Demo:

Screenshot or Video: file:///C:/Users/Rishi/OneDrive/Documents/MY%20PPT%20AND%20NOTES/GhostLore%20AI/index.html

The project includes an offline demo mode. Actual open-weight AI generation requires Ollama and a compatible local model to be installed and running.

GitHub Repository:https://github.com/rishiyadavcreator/GhostLore-AI.git

Built with:

HTML5 for the website structure

CSS3 for the dark, atmospheric interface and responsive design

Vanilla JavaScript for story generation, the haunted terminal, and the local archive

Ollama for running a local language model

Qwen2.5 as the example open-weight model

How I Built It

I built GhostLore AI using HTML, CSS, and vanilla JavaScript, keeping the project lightweight and beginner-friendly.

The interface uses a dark background, glowing purple and pink accents, spooky typography, and responsive layouts to create a mysterious atmosphere.

The application has three main parts:

The Story Chamber

Users enter a story idea and select an atmosphere and story length. The website can connect to Ollama's local generation API to request an original story from a locally running model.

The Haunted Terminal

Users can enter questions and receive mysterious responses. When Ollama is available, the terminal can use the local model. Otherwise, it provides clearly labeled demo responses.

The Story Archive

Users can save stories in browser local storage, reopen them, copy their content, or download them as text files.

For the open-weight AI component, I chose Ollama with Qwen2.5 as an example configuration. This allows the model to run on the user's own computer rather than requiring a paid, hosted AI API.

The demo fallback makes it possible to explore the interface even before installing the model, but it is template-based and is not itself AI inference.

Why Does Open Innovation Matter?

For a project about mysterious stories and personal memories, privacy matters.

A closed, hosted AI API can be convenient, but it may require sending prompts to an external service. With a properly configured local model, GhostLore AI can generate stories on the user's own machine.

Open innovation makes several things possible:

Privacy: Prompts can remain on the local computer when inference is performed locally.

Control: Users can choose and replace compatible models.

Cost flexibility: Local inference avoids per-request hosted API fees, although running a model still uses computing resources.

Offline potential: After downloading the model and required files, local generation can work without an internet connection, provided the local runtime is available.

Freedom to experiment: Developers can explore different models and adapt the experience without depending entirely on one closed AI provider.

For me, open-source AI is not just about using a model for free. It is about learning how the technology works, having choices, and building tools that people can control.

My Agent Session

This section is optional.

If I record a development session using DevRelay, I can share the session link here so readers can see how GhostLore AI was built and improved.

The Spirit of It

Every family, community, and generation has stories worth preserving.

Some are spooky tales told at night. Others are memories of people, places, and traditions that shaped us.

GhostLore AI takes inspiration from that idea: technology can help us create new stories and keep the stories we care about close to home.

The ghost in the machine is not necessarily something to fear. Sometimes, it is simply a story waiting to be told.

Prize Categories

Add the partner prize categories that apply to your project according to the official challenge rules. Remove this section if you are not entering a specific partner category.

Built with curiosity, a little mystery, and an appreciation for open-source AI. 👻

Hacktoberfest #OpenSource #AI #JavaScript #Ollama #GhostInTheMachine

SlothDesk AI — Do Less, Gently

Rishi Yadav — Tue, 20 Oct 2026 14:24:41 +0000

What I Built

I built SlothDesk AI, a productivity assistant designed for people who want to get things done without feeling overwhelmed.

We all have tasks we keep postponing, long notes we don't want to read, and big goals that feel difficult to start. SlothDesk AI makes these things easier by breaking tasks into smaller steps and simplifying everyday work.

✨ Features include:

Lazy Plan Generator: Breaks big tasks into small, manageable steps.
Note Simplifier: Turns long notes into short summaries.
Focus Mode: Helps users concentrate on one task at a time.
Task Manager: Add, complete, and delete tasks.
Persistent Storage: Saves tasks in the browser using local storage.
Local AI Integration: Supports Ollama and an open-weight language model for AI-powered assistance.

The idea is simple: spend less energy organizing your work and more time doing what matters.

Demo

🔗 Live Demo: [ file:///C:/Users/Rishi/OneDrive/Documents/Slow%20and%20Steady/index.html]

The demo lets users organize their tasks, generate simple plans, summarize notes, and use Focus Mode.

Code

💻 GitHub Repository: [ [https://github.com/rishiyadavcreator]]

Built with HTML, CSS, and vanilla JavaScript, with optional integration for Ollama.

How I Built It

I built SlothDesk AI using HTML, CSS, and JavaScript.

HTML provides the structure, CSS creates a clean, calming, responsive interface, and JavaScript powers the task manager, Focus Mode, browser storage, and productivity tools.

For the AI component, I designed an optional integration with Ollama, which can run an open-weight language model locally. The app can use a model such as Qwen2.5 3B to generate task plans and summarize notes.

When the model is unavailable, basic fallback functionality keeps the productivity tools usable.

My goal was to create something lightweight and beginner-friendly without depending entirely on paid cloud APIs.

Why Does Open Innovation Matter?

Open innovation matters because AI should be accessible, customizable, and under the user's control.

With a locally running open-weight model, SlothDesk AI can offer several advantages:

Privacy: Prompts can be processed on the user's own computer rather than sent to an external AI provider.
Cost control: Local inference avoids per-request API charges, although hardware and electricity still have costs.
Flexibility: Users can experiment with different compatible models instead of being locked into one provider.
Offline potential: Once a model is downloaded and configured, inference can work without an internet connection.
Accessibility: Basic productivity features remain available even when the AI service is unavailable.

A closed API can be convenient, but open-weight models give developers more freedom to experiment and users more control over how their tools work.

For me, open-source AI makes it possible to build a productivity assistant that can grow with its users instead of depending entirely on a paid service.

My Agent Session

I developed SlothDesk AI with the goal of making everyday productivity simpler and less stressful.

🔗 Agent Session: Add your DevRelay session link here if you create and save one.

The Lazy Payoff

SlothDesk AI is built around one question:

What if getting started took less effort?

Instead of staring at a large task, users can break it down into smaller actions. Instead of manually organizing every task, they can keep everything in one place. Focus Mode helps them work on one thing at a time.

The intended payoff is less planning overhead, fewer distractions, and an easier way to begin.

I haven't measured actual time savings yet, so the next step is to ask users to try it and collect feedback on how much effort it saves them.

Prize Categories

Primary theme: Slow and Steady — open-source AI for easier everyday life.

Technology focus: Open-weight models and local inference.

The project is intended to demonstrate how open-source AI can make everyday productivity tools more accessible and customizable.

Built with 🌿, JavaScript, and a little sloth energy.

SlothDesk AI — Do Less, Gently. 🦥

DukaanAI — Open-Source AI Product Listing Assistant

Rishi Yadav — Thu, 15 Oct 2026 15:13:50 +0000

🚀 DukaanAI: An Open-Source AI Assistant for Small Sellers

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle.

What I Built

I built DukaanAI, an AI-powered product listing assistant designed to help small business owners, shopkeepers, and online sellers create attractive product listings quickly.

Many small sellers struggle to write professional product descriptions, catchy titles, and social media captions for their products. They may not have the time, writing skills, or budget to hire a marketing professional.

DukaanAI helps solve this problem by generating product titles, descriptions, and promotional captions from simple product details. Users can choose their preferred language, including English, Hindi, and Hinglish, and select a writing tone that suits their business.

The goal is to make AI-powered marketing assistance more accessible to small sellers and people starting their own side hustles.

Demo

https://rishiyadavcreator.github.io/DukaanAI-Open-Source-AI-Product-Listing-Assistant/

The website provides the user interface for the project. The AI-powered generation feature is designed to run locally with an open-weight model.

Code

💻 GitHub Repository: https://github.com/rishiyadavcreator/DukaanAI

The project is built with HTML, CSS, and JavaScript, with a Node.js backend for AI integration.

How I Built It

I built DukaanAI using:

HTML to structure the user interface.
CSS to create a responsive and user-friendly design.
JavaScript to handle user input, display generated content, copy results, download listings, and save recent listings in the browser.
Node.js to provide the backend API.
Ollama to run an open-weight language model locally.
Qwen2.5 3B as the planned local model for generating product titles, descriptions, and captions.

The intended workflow is simple: the user enters product details, the frontend sends them to the Node.js backend, and the backend requests generated content from the local model through Ollama.

I chose local inference to explore how AI features can work without depending entirely on a paid, closed-source AI API.

Why Does Open Innovation Matter?

Open innovation matters because powerful AI tools should be accessible to more than just large companies with big budgets.

By using an open-weight model and local inference, DukaanAI explores a way to build AI-powered tools with more control over the model, fewer dependencies on paid API services, and opportunities for developers to customize the experience.

This approach could help small sellers create better product listings without needing advanced writing skills or expensive marketing tools.

It also helps me learn how open-weight AI models, frontend development, and backend APIs work together in a practical project.

My goal is to keep improving DukaanAI and make AI-powered business assistance more accessible to people starting small businesses and side hustles.

Prize Categories

Open-Source AI / Open-Weight Models — for the use of an open-weight model in the project.

DukaanAI — AI Product Listing Assistant

Rishi Yadav — Thu, 15 Oct 2026 09:24:02 +0000

🛍️ DukaanAI: Turn Products into Sales-Ready Stories | Hacktoberfest Week 2

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle.

What I Built

I built DukaanAI — an AI Product Listing Assistant, a web application designed to help small business owners, local shopkeepers, freelancers, and online sellers create product marketing content more easily.

Many small sellers struggle to write attractive product descriptions, product titles, and social media captions. Creating this content manually takes time and can be difficult for people without marketing or writing experience.

DukaanAI aims to simplify this process by allowing users to enter product details and generate content they can use for their online stores and social media.

Key Features

📝 Product title generation
📄 Product description creation
📱 Social media captions
🌐 Language and writing-style options
📋 Copy generated content
💾 Download product listings
📱 Responsive, user-friendly interface

Who would pay for it?

Small online sellers, local shops, handmade-product businesses, and freelancers could benefit from a tool like DukaanAI. In the future, it could offer a free basic version and paid advanced features for businesses that create product listings regularly.

Demo

![ ](https://dev-to-uploads.s3.us-east-2.amazonaws.com/uploads/articles/eouxkdde5ngxlcbxfr2h.png)

🚀 Try DukaanAI:

https://rishiyadavcreator.github.io/DukaanAI/#

Explore the interface and try creating a product listing.

Code

💻 GitHub Repository:

https://github.com/rishiyadavcreator/DukaanAI

The project is built with HTML, CSS, and JavaScript, making the frontend simple to understand and extend.

Contributions and suggestions are welcome!

How I Built It

I built DukaanAI using basic web technologies:

HTML5 for the structure and forms.
CSS3 for a clean, modern, responsive design.
JavaScript for handling user input, generating content in the current prototype, and providing listing-related interactions.

My goal was to create a practical tool that small businesses could use without needing complicated software.

Open-Source AI Integration

The current prototype uses JavaScript-based templates for content generation. It does not yet run a real AI model.

My next step is to integrate an open-weight language model through Ollama, with a backend API connecting the website to the model. This would allow DukaanAI to generate more flexible, context-aware product descriptions and marketing captions.

I want to build toward an approach where developers can inspect, customize, and run the model according to their needs, rather than depending entirely on a closed AI API.

Why Does Open Innovation Matter?

Open innovation makes technology more accessible to students, independent developers, and small businesses.

For a project like DukaanAI, open-weight AI models could make it possible to experiment with local inference, customize content-generation workflows, and reduce dependence on paid third-party AI APIs.

It also gives developers more flexibility to explore privacy-conscious implementations and build tools for people whose needs may not be served by expensive commercial products.

As a student learning software development and AI, I see open innovation as an opportunity to learn by building, share what I create, and improve my work through community contributions.

My long-term goal is to make DukaanAI a genuinely useful tool for small sellers by combining accessible web technologies with open-source AI.

Prize Categories

Primary theme: Side Hustle

Potential fit: Open-source AI and small-business productivity.

The current project is a frontend prototype, and the real open-weight AI integration is planned. I would not claim eligibility for a model-specific partner category until that integration is implemented and meets the category requirements.

Thanks for checking out DukaanAI! 🚀

If you have ideas for improving product listings or making AI more accessible to small businesses, I'd love to hear your feedback.

DukaanAI — Write. Sell. Grow.

NatureMate AI: Touch Grass, Explore More — Bringing Open-Source AI into the Real World

Rishi Yadav — Fri, 09 Oct 2026 11:04:17 +0000

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

What I Built

🌿 NatureMate AI — Touch Grass, Explore More

I built NatureMate AI, an AI-powered outdoor companion that encourages people to spend less time in front of screens and more time exploring the real world.

The idea behind this project is simple: technology should help us experience nature, not keep us glued to our devices.

NatureMate AI helps users discover outdoor activities, generate personalized nature challenges, plan beginner-friendly gardening activities, and maintain a digital journal of their outdoor observations.

Key Features

🌱 AI Adventure Generator: Creates personalized outdoor missions based on users' interests, such as bird watching, nature exploration, gardening, and photography.
🪴 Smart Garden Planner: Provides gardening suggestions based on the user's region, available space, and experience.
🍃 Outdoor Challenges: Encourages users to observe leaves, watch birds, explore clouds, and discover their surroundings.
📔 Nature Journal: Allows users to save observations and memories from their outdoor experiences.
✅ Challenge Tracker: Tracks completed outdoor activities.
🔒 Privacy-Focused AI: Uses a locally running AI model so prompts can be processed on the user's own computer.

NatureMate AI is designed for students, beginners, nature lovers, and anyone looking for simple ways to reconnect with the outdoors.

Demo

🌐 Live Website: https://rishiyadavcreator.github.io/NATURE-MATE/

The website demonstrates the NatureMate interface and its outdoor exploration features.

Note: The GitHub Pages version is a static frontend. Features that require the locally running AI backend need the project to be set up on a computer with Ollama and the selected model installed.

Code

💻 GitHub Repository: https://github.com/rishiyadavcreator/NATURE-MATE

The project is built using HTML, CSS, and JavaScript, with a Node.js backend for connecting to local AI.

The repository contains the website code and project documentation.

How I Built It

I built NatureMate AI using familiar web technologies and an open-weight language model.

Technology stack:

HTML5: Structures the website and its sections.
CSS3: Creates the responsive, nature-inspired user interface.
JavaScript: Handles user interactions, outdoor challenge tracking, journal entries, and AI requests.
Node.js: Provides the backend for serving the application and forwarding AI requests.
Ollama: Runs the language model locally.
Qwen2.5 1.5B: Generates outdoor mission ideas and gardening suggestions.
Browser LocalStorage: Saves journal entries and challenge progress on the user's device.

How the AI works

The user selects an outdoor activity or enters gardening details.
JavaScript prepares a prompt based on the user's input.
The Node.js backend sends the prompt to Ollama's local API.
The locally running Qwen2.5 model generates a response.
The response is displayed in the website.

The project is designed so that AI requests can run locally without an external AI API, once the model and required software have been installed.