QuickFix

Three Roles, One Workflow, Enforced at the Database

Section 1 of 4

QuickFix coordinates rental property maintenance between three parties who each see a different slice of the same job. A customer reports a problem with photos and an address. A manager triages it into the right trade pool and controls what the tenant is ultimately quoted. An agent — a tradesperson in one category — claims the job, quotes it, travels to it, does the work and evidences completion.

That shape is what makes it an interesting build rather than a CRUD app. Three roles with genuinely different permissions over the same documents means authorization cannot live in the interface, because the interface is not what protects the data. A job moving between parties means a state machine with real transitions rather than a status field someone sets arbitrarily. And money entering the flow — an agent's quote, a manager's adjustment, a customer's acceptance — means the record of who agreed to what, and when, has to survive.

This is a side project of mine and the repository is public, so unlike the Strollr case study everything described here is readable. The published copy was deliberately hardened for release from the original private version, and that hardening pass is itself covered in the next section, since it is the most transferable engineering in the project.

3

roles with distinct permissions

8

states in the complaint lifecycle

7.8K

lines of TypeScript

2

platforms from one codebase

0

credentials in the public repo

System Architecture

Three role-specific dashboards over one Next.js application, talking directly to Firebase from the client — which is only safe because the security rules, not the client, are the enforcement boundary.

Customer Dashboardreport, track, accept quotesAgent Dashboardclaim, quote, work, evidenceManager Dashboardtriage, categories, quote controlNext.js 15 App RouterReact + TypeScript + Tailwind + Radix primitives · one server route for team passwordsSecurity Rules — the enforcement boundaryFirestore and Storage rules evaluated server-side on every read and writeFirebaseAuth (phone + Google) · Firestore · Cloud Storage · Cloud Messaging · App HostingAndroid via Capacitorsame codebase, static-exported and wrapped as a native appGoogle Maps Places supplies address capture on the customer side; Google Maps renders it for manager and agent

What Made It Worth Building

This is a client-heavy application: the browser holds a Firebase session and reads and writes Firestore directly, with no application server in between mediating queries. That is fast to build and genuinely pleasant to work with, and it has one consequence that dominates every other design decision — anyone holding a valid session can attempt any query they like, regardless of what the interface offers them.

Hiding a button does not prevent the write behind it. So every access decision in this project lives in the security rules, which are evaluated server-side on every single operation and cannot be bypassed by a modified client. The interface is treated as a convenience layer over rules that would hold even if the interface did not exist.

The Roles and Access Control section walks through those rules in detail, including the one that took the most thought: letting an agent claim an unassigned job from their own trade pool without letting them assign it to anybody else.

A note on scope

This is a side project, not production software with users. It is described here as what it is: a complete, working three-role application with its authorization model taken seriously, built to be read. Where something is scaffolded rather than shipped — an AI toolkit is wired up but no flows are implemented, and the final two terminal states are modelled but not yet reachable — these sections say so rather than rounding up.