-
-
Start with a real product dossier and governed 0% readiness intake.
-
Nutrient extracts five page-grounded claims and blocks the SPF conflict for human review.
-
100% readiness with Perfect Corp render, four live sponsor integrations, and seven audit receipts.
-
Approved SPF 30 plus three live SerpApi Shopping signals kept separate from product truth.
Inspiration
Independent beauty brands do not fail for lack of product ideas. They get stuck between a folder of specifications, certificates, images, and price sheets and the operational finish line: a trustworthy product page and a compelling shopper experience.
Most AI launch tools make content from prompts. BrandProof starts from evidence. It extracts product facts from the documents a brand already has, shows exactly where each fact came from, stops the workflow when material claims conflict, and publishes only after a human resolves uncertainty.
What it does
BrandProof orchestrates the path from product dossier to market:
- It ingests product specifications, ingredient sheets, certificates, wholesale pricing, and imagery.
- Nutrient DWS extracts a typed product record with source citations and confidence, redacts sensitive data, and creates a buyer-safe evidence packet.
- The review workspace blocks publication until contradictory or low-confidence claims are resolved.
- SerpApi supplies current structured market signals with their sources, kept separate from certified product facts and sent to human review.
- The consumer experience uses the verified shade record to configure Perfect Corp virtual try-on.
- A unified audit trail records the real provider calls and resource IDs behind the workflow.
How we built it
The client uses React, Vite, and TypeScript with a deterministic workflow state machine. Xano is the governed server boundary for workflow state, provider credentials, orchestration, and audit receipts. The client/backend contract rejects out-of-order transitions, keeps fixtures visibly labeled, and never silently falls back from live mode. AI and external data can propose or extract; policy gates and human review decide whether a material claim may proceed.
Xano build story
BrandProof replaces fragmented product-information spreadsheets, email approvals, document tools, market-research tabs, and disconnected launch handoffs. The old workflow made it easy to publish a claim without knowing which source supported it or who approved it.
OpenAI Codex supported implementation, debugging, test generation, browser QA, and submission packaging. Xano accelerated the governed runtime: translating four provider contracts into one stateful backend, validating XanoScript, tightening fail-closed response checks, and producing a traceable deployment without custom backend infrastructure. The project was built in approximately one week after the August 17 kickoff, with implementation beginning August 18.
Challenges we ran into
The hardest problem was not calling four APIs; it was proving that their outputs belonged in one trustworthy workflow. Nutrient returns structured page evidence inside a nested response wrapper, so we validated the actual three-page array and bound each product fact to the page that supports it. Our first SerpApi parser accepted the result envelope but produced empty evidence fields. We tightened it to reject empty titles, prices, or sources and re-ran the complete path. Perfect Corp is asynchronous, so Xano creates a task, polls with a hard timeout, and accepts only a successful HTTPS render.
Every transition fails closed. Search results cannot overwrite certified product facts, material conflicts cannot be auto-approved, provider failures create no success receipt, and live mode never silently falls back to fixtures.
Accomplishments that we're proud of
- A public end-to-end workflow that makes four sponsor integrations feel like one focused product.
- Five source-grounded claims extracted from a three-page dossier, including a deliberately conflicting SPF 50 marketing claim and SPF 30 laboratory result.
- A required human decision that approves SPF 30 before market research or try-on can proceed.
- Three populated live Google Shopping signals kept visibly separate from product truth.
- A completed Perfect Corp AI Look render using a fictional, permission-safe portrait.
- Seven visible non-secret receipts for every completed live workflow.
- Seven automated tests, ten validated XanoScript documents, a passing production build, and a clean public repository.
Sponsor integrations
- Xano: the operating backbone for product-launch workspaces, evidence-review state, provider workflows, and the audit record.
- Nutrient DWS: extracts page-grounded product claims from the three-page dossier while preserving the conflict that requires human review.
- Perfect Corp: turns the verified product record into a shopper-facing virtual try-on experience.
- SerpApi: provides structured, real-time Shopping results for current offers, retailer presence, and possible brand collisions.
What we learned
Human in the loop only matters when the human can actually stop the system. The strongest AI workflow is not the one with the most autonomous steps; it is the one that makes evidence, uncertainty, and authority legible. Integration depth is easier to judge when each provider owns a distinct responsibility and leaves a verifiable receipt.
What's next
Expand from beauty into wellness and consumer packaged goods, add reusable retailer requirement packs, support recurring certificate monitoring, and offer agencies a multi-brand workspace.
Log in or sign up for Devpost to join the conversation.