AGNIYA · build progress · updated 19 November 2026
The apps are now clickable, end to end, on sample data
Last week we walked through the 120 approved screen designs, in light and dark, and the
landing page. Since then the team has turned the first set into working apps you can open
in a browser, and built the real backend behind them. Everything here runs on realistic
sample data; no customer data is used.
Assigned visits, Visit detail, Check-in with the distance from the site, Areas,
Checkpoint capture with the AI's suggestion, Result picker, Observation, Other
observation, Stop the visit early, Audit summary, Ask the customer a question,
Sign-off with the expert's declaration, the signed report and its corrected version
2.Open auditor mode (best on a
phone)
Staff sign-in (AGNIYA's team only) and the app's frame and navigation are in place;
its other screens come after the customer-facing apps.
PDF reports8 / 8 templates
Self-audit and expert-audit report pages are built as templates, printing the
expert's declaration exactly as signed. Generating them from a finished audit comes
with the report step.
Website: agniya-landing.pages.dev. It now
opens in light mode, with a sun and moon switch in the top bar for dark mode.
Behind the screens
Real backend working
Sign-in with a phone code, sessions and role-based access
Premises, self-audits and evidence photo uploads
Enterprise overview, sites, findings and fix reminders
Audit requests, and questions with answers and attachments
A site manager sees only their own sites
AI and data foundations done
Past expert audits read in; 178 checkpoints catalogued, each mapped to NBC 2016 Part
4
Personal details and faces removed before anything is used for training
First training dataset prepared and versioned
Security and hosting next: AWS Mumbai
Data stays in India (AWS Mumbai), encrypted at rest
Uploads checked for size and type; strict browser security headers
Limits on how often sign-in codes can be requested
Every change is tested automatically before it's merged
Next week · 20 to 27 November
Consumer app
The self-audit report
Home and the premise profile: type, details, audit plan
The voice-guided self-audit: the guide, camera and quick questions
Enterprise web app
Findings and fixes: who owns each fix, due dates, reminders
Users: invite colleagues, roles and sites
Add a site
Auditor mode
Checking proof that a fix was done
Running on the real API instead of sample data
Backend and AI
Auditor mode on the real API: visits, decisions and sign-off
The self-audit end to end: photo in, AI result out, scored with a versioned scoring
table
Importing sites from a spreadsheet
Admin web app
Staff sign-in, then the review queue for AI findings the model isn't sure about
Hosting
A staging environment on AWS Mumbai for AGNIYA's team to try with test numbers
Automated end-to-end checks of every enterprise screen on every change
What would help from AGNIYA
The Terms and Privacy wording for the consent screen (it shows placeholders today)
The words for score bands and fix statuses, so the apps match how AGNIYA talks to
customers
The operations email address customers should write to
How many working days an expert takes to check proof of a fix
Pricing for the business safety check
See it yourself
These links run the real apps on sample data. Open the mobile links on a phone for the
true size.
Enterprise web app · on a laptop
Overview: the sample company,
Northwind Foods, with 48 sites and its Fix now findings.