Insight Wellness Guide

An AI-powered life-coaching platform that unifies the website, WhatsApp, Instagram, payments, and bookings—helping Dr. Tanya support clients while keeping every action human-approved.

Comment

Insight Wellness Guide

Inspiration

Insight Wellness Guide began during Build Week with a conversation between me and my sister, Dr. Tanya Soni.

She established her life-coaching business in late December 2025 and had already started working with clients. When I asked to see her website, I found a very basic site created with a no-code tool. It communicated the essential information, but it did not reflect the emotional, personal nature of her work.

I wanted to help her build a website that felt trustworthy, memorable, and worthy of the business she was creating.

She also uses WhatsApp and Instagram to communicate with potential clients, share ideas, and promote her work. This led to a larger question: could I create a system that connected her website, conversations, content, and consultation requests so she could focus more of her time on coaching?

That became the central idea behind the project:

Three channels. One customer story.

What it does

Insight Wellness Guide combines a public life-coaching website with the foundation for an AI-assisted business operations platform.

The project includes:

  • a responsive public landing page;
  • an INR 500 initial-consultation offer;
  • a Customer Pulse operations-dashboard prototype;
  • a customer-concierge agent scaffold;
  • a content agent for future Instagram workflows;
  • an Owner Liaison Agent for business changes and booking decisions;
  • scheduling and follow-up workflows;
  • official WhatsApp and Instagram integration boundaries;
  • human approval, escalation, takeover, and audit mechanisms.

The goal is to create one coherent customer journey across the website, WhatsApp, and Instagram while keeping Dr. Tanya in control of business facts, sensitive conversations, bookings, and public content.

How I built it

The landing page was built with semantic HTML, responsive CSS, JavaScript, custom branding, and generated visual assets.

The broader system is structured as a TypeScript monorepo managed with pnpm. It includes a React and Vite dashboard, shared Zod domain contracts, service boundaries, OpenAI Agents SDK scaffolding, synthetic adapters, and Vitest contract and journey tests.

I used product requirements documents and feature-specification files to define the dashboard, agents, lifecycle states, responsibilities, safety rules, and approval boundaries before coordinating the implementation.

How I used Codex and GPT-5.6

Codex was the primary development environment, with GPT-5.6 Medium acting as the main model throughout the build.

I used it for design exploration, implementation, architecture, research, documentation, testing, and project coordination.

One of the most useful workflows was asking Codex to launch separate tasks from the main project task. I used focused tasks to research Meta APIs, investigate free deployment options, compare agent frameworks, and develop the dashboard PRD without overwhelming the main website-building context.

For the platform implementation, Codex used the PRD and feature specifications to delegate bounded pieces of work to parallel sub-agents. Those agents worked on shared contracts, services, tests, and dashboard features before reporting their results to the coordinating task.

This made it possible to explore and build several connected parts of the system in parallel while maintaining a shared repository structure.

Challenges

The largest product challenge was balancing ambition with honesty. Connecting a website, WhatsApp, Instagram, agents, bookings, content operations, and a dashboard is much larger than building a landing page.

I needed to distinguish clearly between:

  • what is working;
  • what is simulated;
  • what is scaffolded;
  • what remains blocked for production.

The most memorable technical challenge happened during the parallel implementation workflow. After the sub-agents had written the dashboard and service code, the coordinating model interpreted the work as exceeding the initial request and began removing it, believing that only feature-specification documents were required.

Fortunately, the work had been committed with Git. Codex was able to recover the implementation without rebuilding everything from scratch.

The experience taught me to define deliverables more explicitly, create checkpoints before major automated changes, inspect diffs carefully, and commit working milestones early.

Accomplishments that I am proud of

I am proud that the project started as a conversation with my sister and became both a polished public website and a credible foundation for a larger operating platform.

I am also proud that the system does not pretend to be more complete than it is. The current implementation uses synthetic data and provider-neutral adapters. Live Meta credentials, production authentication, payments, hosted PostgreSQL, and autonomous publishing remain disabled until the necessary policies and safeguards are established.

What I learned

I learned that working with Codex is an iterative process. I could explain what I wanted in natural language, review the result, and keep refining what “great” should look like.

I also learned that powerful AI agents increase the importance of clear scope, product judgment, source control, verification, and human approval.

I still want to improve my prompting, architecture, and production engineering skills. However, Build Week allowed me to try new development workflows, create something meaningful for my sister, and build a foundation that I can continue developing.

What's next

After Build Week, I plan to:

  • connect the official WhatsApp Cloud API and Instagram Graph API;
  • add production authentication and persistent data storage;
  • integrate consultation scheduling and payments;
  • connect the dashboard to real operational data;
  • implement approval-based content publishing;
  • test the platform with real business workflows;
  • continue improving the website based on customer feedback.

The landing page is working today. The dashboard and agents provide the foundation for what comes next.

Built With

Share this project:

Updates