DEV Community: Deepak Sharma

SecureNotes AI: The 100% Offline, Private Meeting Intake Tool for Solopreneurs

Deepak Sharma — Sun, 18 Oct 2026 19:04:29 +0000

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

What I Built

SecureNotes AI is a lightweight, distraction-free desktop web application designed to turn messy, stream-of-consciousness session notes into structured client files and polished follow-up emails in seconds.

The Problem & Who Pays For It:
Independent consultants, private tutors, life coaches, and freelance accountants deal with sensitive client details daily—financial figures, personal struggles, and proprietary business secrets. While tools like ChatGPT or Claude can format notes easily, professionals bound by NDAs, strict privacy laws, or professional ethics cannot safely paste confidential client data into proprietary cloud APIs.

The Solution:
SecureNotes AI provides an instant, private bridge. Users dump raw meeting scribbles into a clean, calming interface and receive structured action items, deliverables, and drafted follow-up emails instantly. Because the inference happens entirely locally on the user's laptop, sensitive data never leaves the machine. It solves the privacy headache permanently, making it a utility freelancers and solo consultants would happily pay a one-time license fee to own and run privately.

Demo

Here is a visual walkthrough of SecureNotes AI in action:

1. Distraction-Free Note Intake

The interface opens with a clean, centered workspace styled in calming, natural earthy tones designed to avoid screen fatigue.

2. Flexible Configuration & Model Controls

A collapsible sidebar lets the user pick which local model to run and toggle whether they need an internal structured summary, a drafted follow-up email, or both.

3. Structured Outputs in Real-Time

Raw brain-dumps are instantly transformed into clean, actionable cards without hitting the internet.

Code

You can find the full source code and setup instructions on GitHub:

Deepak-IITian07 / secure_notes_ai

How I Built It

SecureNotes AI was engineered for minimal setup and maximum privacy:

Inference Engine: Ollama runs local open-weight models (such as llama3.2, phi3, and mistral) completely offline.
Application Framework: Built using Python and Streamlit for a responsive, rapid-turnaround frontend.
Local Python Client: The official ollama Python SDK communicates directly with the local Ollama daemon via localhost, eliminating external HTTP requests.
Custom Natural Aesthetic: Configured via .streamlit/config.toml and targeted CSS injections to apply a calming sage green and warm-sand color palette with readable typography.

Why Does Open Innovation Matter?

Open innovation is not just an implementation detail for SecureNotes AI—it is the entire value proposition:

Zero-Leak Privacy Guarantee: Closed-source APIs process user text on remote cloud servers where telemetry and logging can happen outside the user's control. Open-weight models running on local hardware ensure that confidential client records never leave the device. You can unplug your Ethernet cable, turn off Wi-Fi, and the tool still functions flawlessly.
No Subscription Overhead: Solopreneurs and freelancers are overwhelmed by recurring SaaS fees. Building on open-weight models eliminates per-token API costs entirely. This unlocks a viable business model: selling a dependable local desktop utility with no recurring infrastructure bills eating into margins.
Model Independence: Closed APIs can deprecate endpoints, change pricing tiers, or modify model alignments overnight. Open-source foundations allow complete autonomy to swap between models without rewriting application architecture.

My First User

To test real-world practicality, I shared an early version of the app with a fellow student and peer tutor who runs private tutoring sessions covering introductory biology and STEM topics.

Their immediate reaction:

"Normally after tutoring, I spend 15–20 minutes re-reading messy scribbles to draft an email for the student's parents detailing what we covered and what needs practice. Having a tool that takes my disjointed notes and instantly drafts a clean summary—without me worrying about student privacy or paying a monthly tool subscription—is a lifesaver."

They tested it on unstructured notes detailing bacterial structures vs. viruses, and the local model immediately isolated the misconceptions, action items, and follow-up homework cleanly.

Prize Categories

Side Hustle

StitchCraft AI: Offline Tailor Job Ticket Generator Powered by Local Gemma

Deepak Sharma — Sun, 04 Oct 2026 19:57:35 +0000

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

What I Built

I built StitchCraft AI, an offline, privacy-first assistant for my sister, who runs a local tailoring boutique.

Every day, she receives dozens of messy, unstructured WhatsApp messages and audio notes from customers (e.g., "Bhaiya stitch an anarkali kurti, chest 38, waist 34, length 42, 3/4 sleeves, deep neck back with tassels. Needs by this Thursday urgent, fabric provided by customer. Advance paid 500 total 1200.").

This informal intake creates constant friction:

Critical body measurements get lost inside long message histories.

Hand-writing workshop slips wastes hours of productive stitching time.

Delivery deadlines and balance payments easily slip through the cracks.

StitchCraft AI allows her to paste raw conversational order text and instantly generate a standardized, print-ready "Tailor Job Ticket" featuring structured measurements, deadline indicators, design requirements, and balance calculations in seconds.

Demo

Here is the tool parsing a real-world chaotic WhatsApp message into a workshop slip:

Clicking the print button activates built-in @media print styling to isolate only the job card for paper/thermal printing:

Code

Deepak-IITian07 / stitchcraft-ai

🧵 StitchCraft AI - Tailor Job Slip Generator

Built for Hacktoberfest 2026 Weekend Challenge: "Build for a Friend"
An offline, private, open-source AI assistant designed for neighborhood tailors and boutique owners to turn chaotic WhatsApp messages and voice transcripts into clean, printable workshop slips.

💡 The Problem

Small boutique tailors receive dozens of unformatted, casual customer orders over WhatsApp every week:

"Bhaiya stitch an anarkali kurti, chest 38, waist 34, length 42, 3/4 sleeves, deep neck back with tassels. Needs by this Thursday urgent, fabric provided by customer. Advance paid 500 total 1200."

Key body measurements get lost in chat histories, delivery deadlines slip, and manually handwriting paper tickets takes away creative sewing time.

🛡️ Why Open-Source AI?

100% Data Privacy: Body measurements and personal customer contact numbers never leave the local workshop device.
Zero API Costs: Micro-businesses operate on tight margins and cannot afford monthly proprietary token subscriptions. Running…

View on GitHub

The project is built as a self-contained single-file Streamlit application (app.py) for lightweight local execution.

How I Built It

StitchCraft AI is powered entirely by an open-source AI core:

Open-Weight Core: The system runs Google's Gemma 2 (2B) locally via Ollama (http://localhost:11434/api/generate).

Structured JSON Parsing: The application feeds raw conversational inputs into a strict zero-shot extraction prompt, enforcing a typed schema covering customer details, 8+ body measurements, fabric sources, deadlines, and pricing.

Resilient Sanitation: A regex cleaning layer strips formatting artifacts or markdown fences to guarantee consistent JSON deserialization.

Single-Page UI & Print Pipeline: Built with Streamlit and customized CSS with @media print rules, allowing the user to print or save the ticket as a PDF directly via standard browser print dialogs (window.parent.print()) without external PDF libraries.

Why Does Open Innovation Matter?

Neighborhood artisans and small business owners face specific constraints where closed-source APIs fail:

Client Privacy: Customer body measurements and personal contact details never leave the boutique's machine, keeping personal customer data completely private.

Zero Operating Cost: Independent tailors operate on thin margins and cannot sustain monthly proprietary API bills or recurring subscriptions. Open-weight models offer zero ongoing software expenses.

True Offline Independence: Workshop spaces often encounter intermittent internet connectivity. An open-source model running on local consumer hardware guarantees uninterrupted daily work without an active internet connection.

My Agent Session

The project was scaffolded, debugged, and styled interactively directly inside the VS Code terminal using the Claude Code CLI paired with an Ollama cloud model runner (ollama launch claude --model gpt-oss:20b-cloud), allowing rapid end-to-end iteration from architecture prompt to production-ready script in under a day.

Prize Categories

Best Use of Gemma