Domoranda

Everything your home needs, remembered.

Comment

Domoranda

Everything your home needs, remembered.

Domoranda is the operating guide your home never came with. It turns the systems, appliances, filters, outdoor spaces, climate conditions, documents, and maintenance history of one specific home into a practical care plan a homeowner can understand and use.

Live application: https://domoranda.com Primary judge route: assigned account at https://domoranda.com/sign-in No-login fallback: https://domoranda.com/demo Category: Apps for Your Life

Assigned judge accounts use independent synthetic households. Credentials are provided only in Devpost's private testing instructions. No payment method, API key, email verification, or real customer data is required.

Inspiration

Homes contain hundreds of quiet obligations, but the facts are scattered across manuals, labels, receipts, warranties, inspection notes, seasons, climate guidance, and memory. A generic calendar cannot tell whether this house has three air-filter locations, a cartridge pool filter, a salt system, a well-treatment stage, a window balance spring, or a finish that needs a specific touch-up record.

Unexpected repairs often begin as forgotten ordinary care. Domoranda exists to answer a simple question:

What is this house forgetting?

The project began as a GPT-5.6 product-brainstorming conversation in ChatGPT. During Build Week, it moved into Codex and became a deployed consumer application rather than a developer-facing evidence tool.

What Domoranda does

Builds a home-specific map

Guided onboarding records ownership context, home type, climate, heating and cooling, water systems, appliances, alarms, outdoor areas, pools and spas, grills, irrigation, regional hazards, and other relevant features.

Homeowners can add multiple items of the same type as separate records:

every heating and cooling filter position; refrigerator, under-sink, whole-house, reverse-osmosis, softener, and well filters; multiple appliances or equipment units; pool bodies, sanitizers, filters, and readings; windows, finishes, outdoor equipment, and custom items.

Unknown facts remain unknown. Domoranda does not invent a brand, model, serial number, service date, interval, or local requirement.

Produces an applicable care plan

The application matches saved home facts to 93 reviewed Domoranda Home Guides. Each applicable task can include:

why the work matters; a homeowner-friendly explanation; estimated first and recurring dates; tools, supplies, cleaners, and replacement details; personal protective equipment and shutoffs; preparation and safe steps; stop conditions; when qualified professional help is appropriate; and reviewed source context.

A homeowner can record a task as complete, mark it not applicable, change an estimated interval, add notes, and see the next due date update.

Keeps home care usable

Consumer 1.0 includes:

Home Overview; separate Home Items with exact locations and replacement details; Today and Coming Up care views; maintenance-session planning; completion history and recurring schedules; pool and spa readings and equipment records; receipts and warranties; household access; records and incident history; cost outlook; quote organization and priority support; weekly email preferences; and Free, Home, and Home Plus plan entitlements.

The public no-login demo uses a deterministic synthetic household so evaluation remains available even if authentication or an external provider is temporarily unavailable.

GPT-5.6 Home Coach

The signed-in Home Coach answers homeowner questions in everyday language using only the applicable reviewed guides and confirmed facts supplied by Domoranda.

The production route is deliberately bounded:

GPT-5.6 Luna with low reasoning handles an ordinary answer. One GPT-5.6 Terra medium retry is allowed only when a completed Luna result fails one of six machine-checkable integrity or safety validations. Customers cannot select a model, reasoning effort, tools, or token limit. Deterministic application code retains authority over applicability, dates, safety classifications, entitlements, and billing. Compatible reviewed answers can be reused without another provider call. Failed or incomplete answers release the customer's allowance. Delivered questions and answers are saved server-side, searchable, filterable by home area, and recoverable after a browser closes or a response is dropped.

During final live verification, Codex found that the original 400-token structured-answer ceiling could stop an otherwise valid Luna response before its JSON completed. The release raised the bounded ceiling, preserved the cost and escalation rules, passed CI, and proved a completed Luna answer in production.

Plans and billing

Domoranda has three customer plans:

Free: one property, basic home profile and schedule, manual tasks, reminders, up to ten active home items, and three newly generated Home Coach answers per calendar month. Home: $9.99/month or $89/year, with full home mapping, expanded records and care planning, weekly summaries, and ten newly generated Home Coach answers per monthly service period. Home Plus: $14.99/month or $129/year, with two properties, household collaboration, deeper records, cost forecasting, quote organization, priority support, and ten newly generated Home Coach answers per monthly service period.

Compatible saved guidance does not use a newly generated answer.

Stripe provides embedded checkout, recurring billing, receipts, cancellation, and the customer portal. Prices and entitlements are server-authoritative. Cloudflare Turnstile protects higher-risk checkout attempts without placing a permanent challenge in front of ordinary customers.

How it works

A person joins and saves a home. Domoranda separates confirmed systems, excluded systems, and questions that still need clarification. Saved items receive applicable Home Guides and estimated care dates. The homeowner opens a task to see its reason, supplies, tools, safety limits, and source. Completion updates history and the next due estimate. The Home Coach can explain an applicable home-care question using confirmed facts and reviewed guidance. Manuals, receipts, warranties, item details, and history stay attached to the household.

Technical implementation

Domoranda is a Next.js 16 and strict TypeScript modular monolith deployed on Netlify.

React and Next.js: responsive consumer product and server routes Supabase/PostgreSQL: authentication, row-level household isolation, entitlements, immutable policy and AI accounting, schedules, records, and judge fixtures OpenAI Responses API: bounded GPT-5.6 Home Coach generation Stripe: live Home and Home Plus subscriptions, webhooks, portal, renewals, and cancellation Resend: transactional and household email delivery Cloudflare Turnstile: server-verified abuse protection Zod: strict public and provider boundary validation GitHub Actions: application, database, security, secret-boundary, and four-shard responsive-browser checks Netlify release workflow: committed-files-only staging, immutable verification, and exact production provenance

The repository includes setup instructions, synthetic sample data, database migrations, test suites, architecture and AI contracts, release evidence, and a 90-second evaluation path.

Built during OpenAI Build Week

During Build Week, Domoranda was designed and implemented as a new consumer home-care product, including:

household onboarding and evidence confirmation; applicability-aware maintenance planning; 93 source-backed Home Guides; exact consumable and equipment tracking; completion history and recurrence; pool and spa care records; Free, Home, and Home Plus entitlement architecture; the GPT-5.6 Home Coach; saved-answer reuse and durable answer history; Stripe, Supabase, Resend, Netlify, and Turnstile boundaries; isolated synthetic judge households; and a deterministic no-login demonstration.

How Codex accelerated the workflow

Codex was not used as a one-shot code generator. It acted as the engineering workspace and release partner across the product.

Codex accelerated:

translating the product and technical specification into working vertical slices; parallel implementation of consumer UX, maintenance coverage, database contracts, billing, email, security, accessibility, and release tooling; migration design and PostgreSQL contract tests; responsive review at mobile, tablet, desktop, and enlarged-text scopes; homeowner-language and accessibility reviews; adversarial checks for account isolation, billing authority, API spending, and secret exposure; GitHub PR, CI, clean-build, and exact Netlify deployment workflows; and live browser verification of AI delivery, persistence, search, filtering, plan visibility, and judge reset behavior.

Concrete defects found through the Codex workflow included a first-use policy-origin failure behind the Netlify proxy, a support-queue Boolean defect, stale consumer error messages, incorrect pool setup behavior, dropped Home Coach responses, and the structured-output truncation boundary described above.

That combination of implementation, independent review, and live verification compressed work that would normally span multiple specialized engineering roles into one Build Week.

Key decisions

The human owner retained the consequential decisions:

keep Domoranda consumer-facing rather than expose a developer tool; make “Everything your home needs, remembered.” the primary promise; preserve unknown facts instead of generating false certainty; require tools, supplies, cleaner limits, PPE, shutoffs, and stopping points; support repeated items and exact locations rather than one generic appliance per category; keep the public demo deterministic and synthetic; protect customer AI spending with allowances, reuse, validation, and fail-closed accounting; keep recommendations neutral and the optional professional marketplace closed; use Dravara, LLC as operator; and enable sales and AI only after their independent production gates passed.

GPT-5.6 assisted early product ideation and powers the bounded Home Coach. Codex accelerated implementation, reviews, debugging, documentation, CI, and production release. The owner set the product, safety, privacy, financial, and launch boundaries.

Challenges

Exhaustive coverage without pretending certainty

Home maintenance is highly conditional. A useful system must distinguish a pool from no pool, a return grille with a filter from one without, and an applicable freeze task from irrelevant advice.

Plain language with precise safety limits

Consumer language cannot erase chemistry, electrical, pressure, access, or contamination boundaries. Domoranda makes stopping points visible instead of burying them in technical prose.

Durable AI delivery

An answer should not disappear because a browser closed or a network response was interrupted. Domoranda commits completed answers server-side, returns allowance only after delivery, and exposes searchable household history.

Commercial correctness

Prices, renewals, cancellation, tax, webhooks, entitlements, AI allowances, and customer data must remain correct across retries and external-provider events. These boundaries are enforced server-side and tested independently.

Accomplishments

A deployed, coherent Consumer 1.0 rather than a technical proof of concept A specific household mapped to applicable source-backed care 93 homeowner-readable Home Guides Working completion, recurrence, notes, records, and care sessions Live GPT-5.6 Home Coach with validated delivery and durable history Live Free, Home, and Home Plus entitlement and billing architecture Isolated synthetic judge accounts plus a no-login fallback Strict tenant, secret, billing, AI-cost, and provider boundaries Production release proven against the exact reviewed Git commit

What we learned

The best maintenance system is not the longest checklist. It is the right obligation for this home, explained clearly, with the supplies, recurrence, uncertainty, and stopping point still attached.

We also learned that AI cost control and recovery are product features. Reusing compatible guidance, releasing interrupted work, persisting delivered answers, and refusing invented home facts make the experience more trustworthy as well as less expensive.

Codex was most valuable as a coordinated engineering workflow: it connected implementation, testing, independent review, and release evidence while the owner retained final authority.

Safety and limitations

Domoranda is informational software, not an inspector, contractor, emergency service, insurer, warranty provider, or substitute for a qualified professional.

Unknown home facts stay unknown. Domoranda does not invent models, dates, equipment, or local requirements. Hazardous or uncertain work routes to the appropriate professional. Current manufacturer instructions and applicable local requirements prevail. Judge households and the public demo contain synthetic data only.

Built With

Share this project:

Updates