DEV Community: Shubham Bawari

Sloth Summarizer: Read Less, Know More

Shubham Bawari — Wed, 21 Oct 2026 13:26:55 +0000

What I Built

Sloth Summarizer is a document summarizer. Paste text or upload a PDF, click one button, and get a short summary in seconds.

Long reports, study notes, and PDFs take time to read. A sloth would not read 20 pages. This tool reads them for you and gives you the main points, so students, job seekers, and busy readers can stay a little lazier.

Demo

Live app: https://ai-summarizer9.streamlit.app/

Step 1: Paste text or upload a PDF, then click Summarize

Step 2: Get a short summary

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 3: Slow and Steady

Code

GitHub repo: https://github.com/shubsolos19/Document-summarizer

How I Built It

Model: t5-small from Hugging Face. It is an open-weight model, small enough to run on a normal laptop CPU.
Framework: Hugging Face Transformers with PyTorch.
PDF reading: PyPDF2 pulls text out of PDF files.
Interface: Streamlit web app.

The flow is simple:

Get text from the input box or the PDF.
Split long text into small chunks, because t5-small can only read a limited amount at once.
Summarize each chunk with the model.
Join the chunk summaries into one final summary.

Summary length and chunk size are easy to change in the code (max_length, min_length, max_chars). The whole app is about 150 lines of Python.

Why Does Open Innovation Matter?

Works offline: After the first model download, the summarizer runs with no internet. A closed API needs a connection every time.
Your documents stay with you: Contracts, notes, and reports are summarized on your own machine. They are not sent to a server you do not control.
Free to run: No API key, no per-request bill, no usage limit. Set it up once and use it as much as you want.
Swappable: The model is open, so I can swap t5-small for another open model like BART with a small code change. With a closed API, the provider decides which model you get and when it changes.

The Lazy Payoff

Reading a 1000-word document can take several minutes. The app summarizes it in about 4 to 6 seconds on a normal machine. Fewer minutes reading, more time resting like a sloth.

Happy Sloth Day! 🦥

Menu Translator: A Private, Offline Menu Translation Tool Powered by Llama 3.2

Shubham Bawari — Tue, 13 Oct 2026 06:11:27 +0000

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

What I Built

Menu Translator is a small web app that translates restaurant menu items into another language and adds a short, customer-friendly description for each dish.

The target customer is a local restaurant owner or cafe manager who serves tourists, students, or foreign guests. Generic translators often mess up food names (for example, translating "Dal Makhani" word by word). This tool keeps the original dish name, gives a translated name, and explains what the dish is in one line.

There is also a "Restaurant notes" box. The owner can describe their restaurant and special items (for example, "Our Special Thali has 2 sabzi, dal, rice and 3 roti"), and the model uses this context while translating. This is the part a generic tool cannot do.

Supported languages: English, Hindi, French, Spanish, German, Japanese.

Demo

How to use it:

Paste dish names, one per line.
Optionally add restaurant notes.
Pick the target language.
Click Translate and get the result in a few seconds.

Code

shubsolos19 / Menu-Translator

Translates restaurant menu dishes using a local open-source model (Llama 3.2 via Ollama). Runs offline, no API cost, menu data stays private.

Menu Translator

Translates restaurant menu dishes using a local open-source model (Llama 3.2 via Ollama). Runs offline, no API cost, menu data stays private.

Screenshot

Setup

Install Ollama from ollama.com
ollama pull llama3.2
pip install -r requirements.txt
python app.py
Open http://localhost:5000

View on GitHub

The whole app is a single app.py file, so anyone can read it, run it, and change it in minutes.

How I Built It

Model: Llama 3.2, an open-weight model from Meta
Local inference: Ollama, which runs the model on my own laptop
Backend: Python and Flask, with one route that sends a prompt to Ollama's local API
Frontend: plain HTML, CSS, and JavaScript inside the same file, no build step

Flow: the user's dish list, restaurant notes, and chosen language go into a structured prompt. The prompt asks the model to keep the original name, give the translated name, and write a short description. Flask sends it to Ollama on localhost:11434 and shows the response on the page.

To run it:

Install Ollama and run ollama pull llama3.2
pip install -r requirements.txt
python app.py
Open http://localhost:5000

Why Does Open Innovation Matter?

Cost: There are no API fees and no per-request charges. A small restaurant can translate its menu as many times as needed for free, which makes this a realistic product for a tiny business.
Privacy: Everything runs on the local machine. Menus, recipes, and notes about specials never leave the owner's computer.
Works offline: After the model is downloaded, no internet is needed.
Customizable: The restaurant notes box adapts the output to each business. With an open model I can also swap in a different one by changing a single word in the code (MODEL = "llama3.2"), or fine-tune for a specific cuisine later.
No lock-in: I am not tied to any provider's pricing, rate limits, or policy changes.

Touch Grass Planner: A Local AI That Sends You Outside

Shubham Bawari — Wed, 07 Oct 2026 07:36:43 +0000

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

What I Built

Touch Grass Planner is a small web app that helps you leave your screen and go outside.

You type your city, how much free time you have, and your mood. A local AI model gives you a short 3 step outdoor plan, a list of things to bring, and one tip for this time of year. Then you press one button and a countdown timer starts. The app tells you to close the tab and go.

The screen part is short on purpose. Plan in under a minute, then walk out the door.

It is for anyone who keeps saying "I will go for a walk later" and never does. Students, developers, and anyone who sits too long in front of a laptop.

Demo

The app runs fully on a local machine, so there is no hosted link. Here is how it looks and how to run it.

Run it yourself:

Install Ollama
ollama pull llama3.2:3b
pip install -r requirements.txt
python app.py
Open http://localhost:5000

Code

shubsolos19 / touch-grass

Touch Grass Planner

Small app that makes a short outdoor plan using a local open-weight AI model. You type your city, free time and mood. The model gives a 3 step plan. Then a timer starts and you go outside.

Why open AI

Runs fully on your laptop with Ollama
Works with no internet after setup
Your city and mood stay on your machine
Free to run
You can swap the model with one setting

Setup

Install Ollama: https://ollama.com
Pull a small model: ollama pull llama3.2:3b
Install packages: pip install -r requirements.txt
Run: python app.py
Open http://localhost:5000

Change model

Windows PowerShell: $env:MODEL="gemma2:2b"; python app.py

Mac or Linux: MODEL=gemma2:2b python app.py

Stack

Python, Flask, Ollama, open-weight Llama 3.2

View on GitHub

How I Built It

Open model: Llama 3.2 3B, an open-weight model
Local inference: Ollama runs the model on my own laptop
Backend: Python and Flask. One route sends the prompt to Ollama and returns the answer.
Frontend: Plain HTML, CSS and JavaScript. No heavy framework.

The prompt tells the model to keep the answer short, follow a fixed format, and avoid made up place names. It suggests types of places like a park or a riverside instead. This keeps the plan safe and useful.

The model name is one environment variable, so I can swap to another open model without changing any code.

Why Does Open Innovation Matter?

Works offline. After the model is downloaded, no internet is needed. Good for trails and parks with weak signal.
Private. My city and mood never leave my laptop. No server I do not control.
Free. No API key, no bill, no usage limit.
Swappable. I can change the model in seconds, and use a smaller one on a slow machine.

A closed API would need internet, an account and money. For an app about going outside, that felt wrong.

How It Went Outside

WRITE 2 OR 3 TRUE LINES HERE: where you went, what plan you got, how long, how it felt.

Meal Planner for a Friend: Allergy-Safe Weekly Plans with Local Gemma

Shubham Bawari — Fri, 02 Oct 2026 06:46:44 +0000

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

What I Built

I built Meal Planner for a Friend, a small allergy-aware meal planner for my friend Shrey.

You enter a name, allergies, dislikes, diet type and budget. The app gives you:

7 day plan: breakfast, lunch and dinner for the week, plus a short grocery list.
Suggestions: when Shrey tried it, he said we can add a suggestion box too. So there is a second tab. He types what he needs right now (for example "quick dinner under 20 minutes") and gets 3 simple options that follow his allergies and dislikes.

Everything runs on my laptop. No cloud, no API key, no cost.

Demo

I did not deploy it, because the point is that it runs locally. You can run it in a few minutes:

ollama pull gemma3:1b
git clone https://github.com/shubsolos19/friend-meal-planner.git
cd friend-meal-planner
pip install -r requirements.txt
streamlit run app.py

Sample input: name Shrey, allergies peanuts and milk, dislikes karela, diet Vegetarian, budget Low. Click Make 7 day plan for the full week, or open the Suggestions tab for quick ideas.

What Shrey said

"we can add a suggestion box too bro."

That one line turned into the Suggestions tab.

Code

shubsolos19 / friend-meal-planner

Allergy-aware 7 day meal planner, local Gemma via Ollama

🍛 Safe food for someone you care about. Zero cloud. Zero cost.

✨ Overview · 🔒 Why Open Source AI · 🚀 Quick Start · 🛠 Troubleshooting

✨ Overview

Planning weekly meals is hard when someone has food allergies, a tight budget, and foods they simply dislike.

This app takes those inputs and returns a full 7 day Indian meal plan with a grocery list, in seconds, running entirely on your own laptop.

💛 Built for Shrey for the Hacktoberfest 2026 Weekend Challenge: Build for a Friend on DEV.

🧾 You enter	🤖 Gemma thinks	🍽 You get
Name, allergies, dislikes, diet, budget	Local model, no internet needed	Breakfast, lunch, dinner for 7 days + grocery list

🔒 Why Open Source AI

🔐 Private

Allergy and health info never leaves the laptop. No server. No logs. No third party.

💸 Free

No API key, no subscription…

View on GitHub

How I Built It

Gemma 3 (1B): open-weight model from Google
Ollama: runs the model locally and exposes a simple API on localhost
Streamlit: clean web form and tabs in a few lines of Python
Python and requests: glue code

The whole app is one file. The form inputs go into a prompt that tells Gemma to never use the listed allergens or anything made from them, to skip dislikes, and to keep dishes simple and cheap. Ollama runs the model on my machine and Streamlit shows the result.

The core of the plan prompt:

prompt = f"""Make a 7 day Indian meal plan (breakfast, lunch, dinner) for {name}.
Diet: {diet}. Budget: {budget}.
NEVER use these allergens or anything made from them: {allergies}.
Do not include: {dislikes}.
Keep dishes simple and cheap. Then give a short grocery list."""

A prompt alone is not a safety net, so I added a second layer. After the model answers, the app scans the text for the listed allergens and related foods. For example, a milk allergy also flags paneer, curd, ghee, butter and cheese. If anything shows up, the app shows a red warning and asks you to generate again. The app also reminds you to check ingredient labels, because AI can make mistakes.

Why Does Open Innovation Matter?

Allergies are personal health information, so where the data goes matters.

Private: with a local open model, allergy details never leave the laptop. A hosted API would send them to a server I do not control.
Free: no API key, no per-token bill, so a plan costs nothing to generate.
Offline: after the one-time model download, it works with no internet.
Swappable: I used the small 1B model so it runs on an ordinary laptop. Changing one line in app.py switches to a bigger Gemma, and I am not tied to one provider.

I did not benchmark this against a closed model, so I cannot claim it writes better plans. What I can say is that for a small personal tool that handles health details, running open and local fit the job.

What I Learned

Building for one real person kept the scope small. Shrey's one line of feedback became a feature, and the allergen scan came from asking what happens if the small model slips. Next I want to add a download button for the plan and a cuisine option, since the prompt is currently Indian only.

Prize Categories

Best Use of Gemma: the whole project runs on Gemma 3, served locally through Ollama.

SnapNote

Shubham Bawari — Sat, 11 Apr 2026 11:10:58 +0000

This is a submission for the DEV April Fools Challenge

What I Built

SnapNote is a lightweight Chrome Extension that lets you save any ChatGPT or Claude response as a study note in one click. Instead of losing valuable AI-generated answers in an endless chat history, SnapNote injects a "Save as Note" button directly into the AI chat interface. You can tag notes by subject, search through them instantly, and access everything from a clean popup dashboard — all without leaving your browser.

Demo

📌 Pin the SnapNote extension to your Chrome toolbar, head to ChatGPT or Claude, ask any question, and click the "Save as Note" button that appears below the response. Open the popup to browse, search, and manage all your saved notes.

shubsolos19 / snapnote

📝 SnapNote — AI Study Notes

Save any ChatGPT or Claude answer as a study note in one click. Search, tag, and export your notes.

✨ Features

🤖 One-Click Save: Instantly save any response from ChatGPT or Claude directly from the chat interface.
🏷️ Smart Tagging: Categorize your notes by subject (e.g., English, Coding, or Custom tags).
🔍 Search & Filter: Quickly find exactly what you're looking for through the intuitive extension popup.
💾 Local Storage: Notes are safely and securely saved in your browser's local storage.
⚡ Seamless UI: Clean, lightweight pop-up and minimally intrusive on-page buttons.

🛠️ Project Structure

The project follows a standard Manifest V3 Chrome Extension architecture:

SnapNote/
├── 📄 manifest.json   # Extension configuration
├── 📜 content.js      # Injects "Save" button into ChatGPT/Claude
├── 🎨 content.css     # Styles for the Save button and tag picker
├── 🖼️ popup.html      # Extension popup UI structure

…

View on GitHub

Code

The project is a vanilla JavaScript Chrome Extension built on Manifest V3 — no frameworks, no dependencies, just clean and minimal code.

shubsolos19 / snapnote

📝 SnapNote — AI Study Notes

Save any ChatGPT or Claude answer as a study note in one click. Search, tag, and export your notes.

✨ Features

🤖 One-Click Save: Instantly save any response from ChatGPT or Claude directly from the chat interface.
🏷️ Smart Tagging: Categorize your notes by subject (e.g., English, Coding, or Custom tags).
🔍 Search & Filter: Quickly find exactly what you're looking for through the intuitive extension popup.
💾 Local Storage: Notes are safely and securely saved in your browser's local storage.
⚡ Seamless UI: Clean, lightweight pop-up and minimally intrusive on-page buttons.

🛠️ Project Structure

The project follows a standard Manifest V3 Chrome Extension architecture:

SnapNote/
├── 📄 manifest.json   # Extension configuration
├── 📜 content.js      # Injects "Save" button into ChatGPT/Claude
├── 🎨 content.css     # Styles for the Save button and tag picker
├── 🖼️ popup.html      # Extension popup UI structure

…

View on GitHub

How I Built It

SnapNote is built entirely with vanilla JavaScript, HTML, and CSS following the Chrome Extension Manifest V3 architecture. A content script (content.js) observes the DOM of ChatGPT and Claude using a MutationObserver, detects when a new AI response is rendered, and injects a Save button below it. Clicking the button captures the question and answer, attaches a user-defined tag, and stores everything using the Chrome Storage API. The popup (popup.html + popup.js) then reads from storage and renders all saved notes with live search and delete functionality.

Prize Category

Submitting for Community Favorite — SnapNote solves a real everyday problem for students, developers, and researchers who use AI tools to learn. Instead of re-asking the same questions or digging through old chats, your best AI answers are saved, tagged, and searchable forever. It's simple, it's useful, and it fits naturally into how people already use ChatGPT and Claude daily.

Team Size: Individual
DEV USERNAME: @hello_shubham