HomeSpark Sentinel: Multi-Tenant BI Platform Development by Toolshed (Data, Automation, AI Agents, Framer, Retool)HomeSpark Sentinel: Multi-Tenant BI Platform Development by Toolshed (Data, Automation, AI Agents, Framer, Retool)
Built with Retool

HomeSpark Sentinel: Multi-Tenant BI Platform Development

Toolshed (Data, Automation, AI Agents, Framer, Retool)

Toolshed (Data, Automation, AI Agents, Framer, Retool)

Verified

HomeSpark Sentinel - Multi-Tenant Insurance BI Platform

Challenge

HomeSpark possessed rich consumer app data—ranging from 3D LiDAR home scans and maintenance patterns to AI chat sessions—but lacked a B2B analytics surface to package and sell to insurance carriers. Furthermore, development had to proceed in parallel with an actively evolving backend, creating significant technical hurdles around changing API contracts, expiring auth tokens, inconsistent multi-tenant filtering, strict investor demo deadlines, and SPA-routing conflicts within Cloudflare Pages.

Goal

To convert HomeSpark's consumer data into "Sentinel," a high-fidelity, multi-tenant B2B business intelligence and workflow platform for home insurance carriers. The objective was to build a demo-ready Retool application alongside the evolving backend, ensuring complete tenant data isolation, dynamic AI workflows, and zero broken views during investor and partner presentations.

Our Contribution

Toolshed executed a SOW-phased engagement (MVP v1.0–v1.2) to design, build, and deploy the platform inside HomeSpark's frontend architecture:
Resilient Architecture & Data Fallback Logic: Engineered transformer-level fallback mechanisms across all Retool query transformers. This ensured that whenever parallel backend endpoints (Tinybird API and FastAPI staging services) returned empty, changing, or placeholder data, the UI seamlessly rendered complete demo data without breaking.
Multi-Tenant Row-Level Security (RLS): Enforced strict multi-tenant data isolation by passing brand_id (and org_id where available) as primary RLS keys across all queries, allowing each insurance client to view only their portfolio analytics.
Dynamic AI-Driven Workflows: Leveraged Retool AI to generate real-time, dynamic output rather than static mock content. This included automated outreach drafting ("Contact Homeowner", "Escalate to Agent") using real alert parameters, as well as live trend insight text generation.
Custom React Components & 3D Integration: Standardized pixel-exact UI components matching Figma designs, including custom React tables for Key Item Alerts and Inventory, an image comparison slider, and an interactive 3D Model Viewer iframe with "Frame on Tap" functionality.
Cloudflare Pages Embedding & SPA-Routing: Integrated the Retool application into HomeSpark's React wrapper on Cloudflare Pages, updating routing components to properly resolve SPA deep links while preserving proxy functions for Retool assets.

Overview of Project Success

Toolshed successfully delivered a unified, four-surface platform comprising Dashboard/Platform Analytics, Policyholder Profile, Analytics, and AI Insights Settings. The built-in RLS and graceful fallback logic allowed HomeSpark to run live, flawless demos for investors and insurance partners with real-time AI capabilities and strict tenant isolation.
Like this project

Posted Jul 21, 2026

Developed a high-fidelity B2B BI and workflow platform for insurance carriers using Retool, ensuring strict tenant data isolation and AI-driven insights.

Likes

0

Views

0

Timeline

Apr 22, 2026 - May 19, 2026

Clients

HomeSpark