Forem: vyomanshi gohel

I Built Schedule Sensei & Pushed It to GitHub – Here's What's Inside (And I Need Your Help 👀)

vyomanshi gohel — Mon, 25 May 2026 03:49:46 +0000

Okay, so in my last post I talked about the idea behind Schedule Sensei – an Android app that auto-adds calendar events from your WhatsApp and Gmail notifications using on-device NLP.
Well… the code is live on GitHub now. 🎉
🔗 https://github.com/vyomanshi27/Schedule-Sensei

What's actually in the repo right now
Here's a quick rundown of what's been pushed:
🔔 NotificationListenerService – The core of the whole thing. Hooks into Android's notification system (with your explicit permission) and intercepts messages from WhatsApp, Gmail, and SMS before they disappear.
🧠 On-device NLP Pipeline – This was the fun part to build. Uses regex + a lightweight rule-based parser to pull out:

Dates ("tomorrow", "next Friday", "25th June")
Times ("at 3 PM", "10 AM")
Event titles (best-guess extraction from message context)
Locations when mentioned

📅 Calendar Integration – High-confidence events get added directly via the CalendarContract API. No third-party dependencies, no permissions you don't expect.
⚠️ Confidence Scoring System – Not every message is "Meeting at 3 PM tomorrow." Some are "maybe Thursday? idk lol". The app gives every parsed event a confidence score – high = auto-add, low = plays a custom sound and queues it for your review.
⚙️ Settings Screen (Jetpack Compose) – Toggle per-source (WhatsApp only, Gmail only, SMS only), toggle auto-add, toggle custom sounds. It's minimal right now but functional.

What's NOT done yet (being honest here)
A few things are still being polished before I'd call this v1.0:

Better NLP for recurring events ("every Monday at 9")
Handling multiple dates in one message ("available Mon or Wed?")
Edge case: messages with time zones
UI polish on the pending events review screen
Testing on more Android versions

Why I'm posting before it's 100% done
Because I've been building this mostly solo and I've hit the point where I genuinely can't see my own blind spots anymore.
I need people to:

⭐ Star the repo if the idea excites you (even just to follow the journey)
🐛 Open issues for edge cases I haven't thought of
💬 Drop comments here – what messages would break my parser?
🛠 Contribute if you're into Android/NLP/Jetpack Compose

What would feel right to you? Auto-add everything and let me undo? Or always confirm first?
Drop your thoughts below 👇
🔗 Check out the repo – contributions, issues, and stars all massively appreciated. Let's build this thing properly. 🚀

I'm Building Schedule Sensei: An Android App That Auto‑Adds Calendar Events from WhatsApp & Gmail

vyomanshi gohel — Wed, 20 May 2026 02:42:09 +0000

The Problem We All Face
Let me paint a picture.

You're in a WhatsApp group. A message pops up:

"Tomorrow's class is shifted to 10 AM instead of 9 AM."

You see it. You think "okay." Then tomorrow comes… and you completely forgot. Sound familiar?

Or maybe you get an email: "Meeting rescheduled to Friday at 3 PM in Conference Room B." You manually open your calendar, type it in… and accidentally put 4 PM.

We receive dozens of time‑sensitive messages every day. Yet we still rely on our fallible human memory or tedious manual entry. There has to be a better way.

💡 The Idea: Schedule Sensei
I'm building an Android app that automatically:

Listens to notifications from WhatsApp, Gmail, and SMS (with your permission, of course)

Extracts date, time, title, and location using on‑device NLP

Adds high‑confidence events directly to your calendar

For uncertain messages, it reminds you with a custom sound to check the app

Works entirely offline – your data never leaves your phone

No more missed classes. No more double‑booking. No more privacy worries.

How It Will Work (Demo Flow)
User grants notification access (one‑time setup)

A WhatsApp message arrives: "Team sync tomorrow at 2 PM"

Schedule Sensei detects it → extracts "tomorrow at 2 PM" → confidence high → auto‑adds to calendar

User receives a silent confirmation (or can disable auto‑add in settings)

If the message is vague: "Maybe the schedule changed?" → low confidence → app plays a custom sound and shows a pending event for user review.

🔐** Privacy First**
I know what you're thinking: "An app that reads my messages? No thanks."

Here's the deal:

All processing happens on your device – nothing is uploaded anywhere.

The app only sees notification content, not your full chat history.

You can turn it off anytime, or disable per source (WhatsApp only, Gmail only, etc.).

The code will be open source – you can audit it yourself.

🙌 Why I'm Posting Here
I'm sharing this early because I believe good ideas get better with community feedback.

I'd love your input on:

What messaging apps would you want supported? (Telegram? Signal? Instagram?)

Would you trust an on‑device notification reader? What would make you feel safer?

What edge cases should I handle? (Multiple dates in one message? Recurring events?)

Should I build a manual forward fallback (user shares a message to the app)?

LearnCurator - I built a YouTube tutorial search engine that filters videos, ranks by AI‑analysed comments

vyomanshi gohel — Mon, 18 May 2026 05:20:32 +0000

LearnCurator – Stop wasting time on bad YouTube tutorials

I love learning from YouTube, but I kept running into the same problems:

Shorts mixed into search results (too short to teach anything)
Outdated tutorials (still showing 4‑year‑old videos even if better ones exist)
No way to tell quality until I’ve watched 5 minutes

So I built LearnCurator – a smart search engine for YouTube tutorials.

🔍 How it works

Initial engagement score The app first computes a base score for each video using: -engagementRatio = likes / views (capped at 1.0) -recencyFactor based on publish age

In youtubeService.js: -score = engagementRatio * 0.7 + recencyFactor * 0.3 -So videos with high like/view ratio and recent publish date rank higher.

Sentiment and final score After getting the top candidate videos, it enriches them with sentiment analysis and a popularity boost.

Final score formula: finalScore = engagementRatio * 0.3
recencyFactor * 0.1
sentimentScore * 0.3
viewBoost * 0.3

🤖 AI‑powered comment analysis

I integrated Gemini 1.5 Flash to analyse YouTube comments. The AI counts positive vs negative feedback and adds a sentiment score to each video. This helps surface content that the community actually loved – not just clicked on.

👍👎 User feedback

Every video card has 😍 😐 😞 buttons. Users can rate a tutorial and leave optional comments. I built a private admin dashboard where I can see real‑time feedback, average ratings, and trending complaints.

🌐 Live demo

👉 [https://learncurator.netlify.app/]

GitHub repo. Link - [https://github.com/vyomanshi27/LearnCurator]

Star the repo if you find it useful – it helps others discover the project ⭐

Try searching for “JavaScript promises” or “React hooks” – you’ll only get videos, no shorts, all longer than 5 minutes, ranked by quality.

🛠️ Tech stack

Frontend: HTML, CSS, vanilla JS (responsive, works on mobile)
Backend: Netlify Functions (Node.js)
APIs: YouTube Data API v3, Gemini 1.5 Flash
Database: Supabase (feedback storage + caching)
Deployment: Netlify (frontend + functions)

📊 What I learned

How to handle YouTube API quotas and caching
Using Gemini for sentiment analysis on real‑world data
Building a secure admin dashboard without a traditional backend
Deploying a full‑stack app entirely on Netlify’s free tier

🚀 What’s next

Add transcript relevance scoring (search inside videos)
Personalised recommendations based on past feedback
Dark mode (because every app needs it 😄)

I’d love your feedback. Try it out and let me know what you think!

If you're a learner tired of low‑quality tutorials, or a developer curious about AI + YouTube, give LearnCurator a spin.