Skip to the work
SUNDAY, 2 AUGUST 2026

Zentise

“Building what's next.”

  • 4 applications live
  • No waitlists
  • No mock-ups
  • Source published under MIT
  • Shipped from India
  • Android
  • ChromaDB
  • FastAPI
  • Gemini Embeddings
  • Gemini Vision
  • Google Play
  • Groq
  • Groq · Llama 4 Scout
  • JWT
  • LangChain
  • lightweight-charts
  • OpenRouter
  • pandas
  • PyMuPDF
  • React
  • React 18
  • SQLAlchemy
  • SQLite
  • Tailwind
  • Vite
Now ShippingThe Front Page · Software

Everything here is already running.

Zentise is a one-person studio that builds working software and puts it online. There are no mock-ups on this page and nothing is waiting for a launch date. Each of the 4 titles below is deployed, publicly reachable and, wherever the source allows, open to inspection line by line. If you want to know whether something works, the fastest way to find out is to go and use it.

Read the workBrowse the source
Section B

The Work

Four applications, each solving a problem worth solving properly. Three share a spine — a Python service doing the thinking, a light front end doing the talking; the fourth ships to Android instead. Nothing here is a template of anything else, and every entry below links to the real thing.

4 titles4 live URLs3 open source
Research Desk

PaperBrain

Live now
P
Fig. 01 — PaperBrain, paperbrain.zentise.tech
Set in
  • FastAPI
  • LangChain
  • ChromaDB
  • Gemini Embeddings
  • OpenRouter
  • SQLite
  • PyMuPDF

A retrieval-augmented reading room

Every answer cited back to the page it came from — and to nothing else.

File your sources first, then start the conversation. PaperBrain ingests PDFs, web pages and YouTube transcripts into a private, per-chat index, then answers questions strictly from that material. When it cannot find the answer in your documents, it says so rather than inventing one.

What it does
  • Multi-source ingest: PDFs to 50 MB, live web pages, YouTube transcripts
  • Every answer footnoted back to its source passage
  • Conversational memory across the last three turns
  • Answer presets — Auto, Brief, Explain, Pointers, Detailed
  • Swappable models through a single OpenRouter gateway
  • Private accounts; your sources are never pooled with anyone else's
Open PaperBrainRead the source
Markets Desk

StockSage

Live now
S
Fig. 02 — StockSage, thestocksage.vercel.app
Set in
  • FastAPI
  • Groq · Llama 4 Scout
  • React 18
  • Vite
  • Tailwind
  • lightweight-charts
  • pandas

Five analysts, one verdict

A trading terminal where the reasoning is streamed live, not hidden behind a number.

StockSage runs a scanner, a technical analyst, a fundamental analyst and a sentiment analyst in sequence, then hands their findings to a signal generator that returns BUY, SELL or AVOID. The pipeline streams every intermediate step to the browser as it happens, so the call arrives with its own audit trail attached.

What it does
  • Five-agent pipeline: Scanner → Technical → Fundamental → Sentiment → Signal
  • Covers NSE Nifty 50 and the S&P 500
  • Five trade horizons: intraday, F&O, short-term, long-term, commodities
  • Reasoning streamed step-by-step over Server-Sent Events
  • Candlestick and volume charts with standard indicators
  • ATR-sized positions at a 1:2 risk-reward ratio
Open StockSageRead the source
Mobility Desk

EVidey

Live nowAndroid app
E
Fig. 03 — EVidey, evidey.vercel.app
Set in
  • Android
  • Google Play

Range anxiety, solved by arithmetic

Highway routing for electric vehicles on Indian roads, planned around chargers that actually work.

Generic range estimates assume conditions no highway ever provides. EVidey models the range you will really get — at highway speed, on your terrain, with the air conditioning running and a battery that has aged — then plots the trip around charging stops it has confirmed are live, with the food and washrooms nearby already accounted for. It is the one title here that ships as an Android app rather than a web page.

What it does
  • Range modelled per vehicle, adjusted for speed, terrain and battery age
  • Live charger availability across major Indian networks
  • Charging stops chosen for amenities — food, washrooms, hotels
  • Routes cached on-device for dead zones with no signal
  • Built for Indian highways specifically, not adapted from elsewhere
Get EVidey

Source not public

Style Desk

DripMate

Live now
D
Fig. 04 — DripMate, dripmate-red.vercel.app
Set in
  • FastAPI
  • SQLAlchemy
  • React
  • Vite
  • Tailwind
  • Gemini Vision
  • Groq
  • JWT

A stylist that only shops your closet

Outfit suggestions assembled entirely from clothes you already own.

Most style advice ends with something to buy. DripMate begins with a photograph of your wardrobe: upload what you own, let it read the garments, then ask for an outfit for the occasion. Every suggestion it returns is composed exclusively of items already hanging in your closet.

What it does
  • Chat with a stylist that knows your actual wardrobe
  • Upload garments as photos; vision reads colour, cut and category
  • Suggestions tuned to occasion, weather and personal style
  • Save the combinations that worked
  • Mobile-first — built to be used while standing at the wardrobe
Open DripMateRead the source
Section C

House
Rules

Four positions held consistently enough that they show up in the code. They are not a manifesto — they are what a one-person studio has to believe to keep shipping.

  1. Ship first, announce later

    Nothing is written about here until it has a URL you can open. A screenshot is not a product, and a roadmap is not a release. The list on this page is short because it only contains things that actually run.

  2. Show the reasoning

    Where a model is doing the work, its working is on display: PaperBrain footnotes every claim to a source passage, StockSage streams each analyst's findings as they arrive. A confident answer with no provenance is worth nothing.

  3. Small stacks, chosen on purpose

    A Python service doing the thinking, a light front end doing the talking, SQLite until it genuinely hurts. Every dependency has to earn its place, which is why there are so few of them.

  4. Open by default

    The source is published under MIT wherever it can be. Read it, fork it, tell me where it is wrong — that last one is the most useful thing anyone can do.

Section D

Colophon

Zentise is a GitHub organisation with exactly one member. It exists so that finished work has somewhere to live that is not a personal profile buried under half-started experiments — a place where a project can be handed to someone with a straight face. What gets published here has cleared a low but non-negotiable bar: it is deployed, it is reachable, and it does the thing it claims to do.

The recurring interest across the catalogue is the same one: taking something capable and constraining it hard enough to be trusted. Answer only from these documents. Recommend only from this wardrobe. Show every step of the analysis. Model the range you will actually get, not the one printed in the brochure. Constraint is what turns a demo into something a person will use twice.

Zentise on GitHubLetters to the editor

Staff box

Publisher
Zentise
Editor & sole compositor
ShrijithSM
Printed in
India
Titles in circulation
4
Distinct technologies
20
Licence
MIT, unless noted
Notice to readers

Every application listed in this edition is free to use and requires no invitation. If one of them is down when you visit, that is a bug and not a business model — do write in.