# How to Vibe Code Your Own Hypotenuse AI (and Stop Paying for It) > AI-first PXM Platform for Ecommerce - Site: https://hypotenuse.ai - Category: SaaS - Platforms: Web app - Verdict: **Serious undertaking** (35/100 vibecodeable) - Estimated effort: 3-4 months of dedicated evening and weekend work ## Verdict Build a personal subset for catalog management and AI copywriting, but the full multi-modal pipeline and ERP sync infrastructure require serious engineering depth. Hypotenuse AI combines an entire PIM database, an async batch processing pipeline, an AI orchestration layer juggling text, vision, and image-generation models, and bidirectional enterprise integrations. While you can easily spin up a Next.js UI that calls Claude and Flux via API, getting bulk catalog enrichment, multi-source web scraping via Firecrawl, and real-time schema mapping to sync cleanly with Shopify or Akeneo is a massive multi-month undertaking that will break constantly on rate limits, malformed payloads, and queue timeouts. ### What you can't replicate - Enterprise connector ecosystem for NetSuite, Salsify, and Salesforce Commerce Cloud - Fine-tuned brand voice models trained across millions of historical enterprise product listings - Native retailer compliance checking pipelines for Amazon, Walmart, and Target ## What it does Centralizes product data, attributes, digital assets, and catalogs into a single source of truth, automating enrichment, copywriting, and image generation. ### Core features - AI-first PIM database and hierarchy management - Multi-source product data enrichment (web scraping, image extraction, PDF parsing) - Automated product tagging and categorization against custom taxonomies - Bulk content generation (SEO product titles, descriptions, meta text) - Bespoke brand voice enforcement via RAG and prompt constraints - AI product photography and batch image editing (background removal, upscaling, lifestyle scenes) - Multi-language translation and localization (40+ languages) - Bidirectional PIM/ERP integrations (Shopify, Salsify, NetSuite, Akeneo) ## The business ### Pricing - Basic Tier: Custom - Ecommerce Enterprise Tier: Custom ### Funding $150K raised. - Seed Round (August 2020) Investors: Y Combinator, Smile Group, January Capital Founded 2020. Team size: ~15 to 20. ## The hard parts - Multi-modal AI pipeline orchestration chaining text LLMs, computer vision, and diffusion models in bulk - Reliable asynchronous queue architecture for processing thousands of SKUs simultaneously without rate-limit or timeout failures - Robust bi-directional webhook and schema mapping sync engines across diverse ecommerce APIs like Shopify and Akeneo - Low-latency brand voice retrieval and context injection across large catalogs ## How to vibe code Hypotenuse AI ### Prerequisites - Node.js (free): Runtime environment for the Next.js full-stack application and CLI tooling. - GitHub (free): Source control and automated deployment pipelines. - Anthropic / OpenAI API Keys (Pay-as-you-go): Powering the product description generators and attribute extraction models. ### Recommended AI tools - Claude Code: Best-in-class terminal coding agent for scaffolding complex full-stack schemas, worker pipelines, and multi-file Next.js components. - Cursor: Ideal for fine-tuning UI layouts, data tables, and batch management dashboards. ### Stack - Frontend: Next.js with Tailwind CSS and shadcn/ui components - Backend: Next.js Server Actions and API Routes with Trigger.dev for background batch jobs - Database: Neon (Serverless Postgres with pgvector for brand voice embeddings) - Auth: better-auth - Payments: None (Personal use clone) - Other: Firecrawl for web and spec sheet scraping, Fal.ai for AI product photography and image editing, Resend for email notifications ### Hosting - Vercel (Hosting the Next.js frontend and serverless API functions): $0/mo (Hobby Tier) - Neon (Serverless Postgres database with vector support): $0/mo (Free Tier) - Trigger.dev (Durable background task processing for bulk SKU enrichment and image generation pipelines): $0/mo (Free Developer Tier) ### Build guide 1. **Database Schema & PIM Core** — Initialize the Next.js project with Tailwind CSS, shadcn/ui, and better-auth connected to Neon Postgres. Design relational tables for catalogs, SKUs, dynamic attributes (key-value JSONB), and brand voice guidelines. ``` Create a complete Next.js 16 project structure using Tailwind CSS and TypeScript. Configure a Drizzle or Prisma schema for a Product Information Management (PIM) system with tables for 'catalogs', 'skus' (with a JSONB 'attributes' column for flexible specs), 'taxonomies', 'brand_voices', and 'media_assets'. Set up better-auth for secure single-user or team authentication. Build an initial dashboard layout with a sidebar navigation, dark mode support, and a clean data table component using shadcn/ui that displays SKUs with pagination, filtering, and sorting. ``` 2. **CSV Import & Multi-Source Ingestion Pipeline** — Build file upload endpoints and a CSV parser to ingest raw product feeds, mapping incoming column headers to internal PIM fields. ``` Implement a robust CSV/XLSX file upload and parsing utility in Next.js using 'papaparse'. Create an import wizard UI where users upload a supplier spreadsheet, map incoming columns (e.g., 'Item Name', 'Cost', 'Desc') to internal PIM schema fields (title, price, description, sku), preview the parsed rows with validation error highlighting, and batch-insert them into the Neon Postgres database inside a transaction. ``` 3. **AI Attribute Enrichment & Web Scraping Engine** — Integrate Firecrawl and OpenAI/Anthropic APIs to parse external URLs, spec sheets, and images to fill missing product attributes automatically. ``` Build an AI data enrichment module that takes an SKU and optional supplier URL or image asset. Use the Firecrawl API to scrape product details from the URL into clean markdown, and send the text along with product imagery to the Anthropic API using structured JSON mode. Extract missing attributes (material, dimensions, color, weight) according to the catalog's taxonomy schema, handle rate limits gracefully, and present the AI-suggested attribute diffs in a review modal for user approval before writing to the database. ``` 4. **Bulk Content Generation & Brand Voice RAG** — Create bulk generation workflows for SEO titles, descriptions, and meta tags constrained by stored brand voice guidelines and vector embeddings. ``` Create a bulk content generation workspace that allows users to select multiple SKUs and generate SEO-optimized product titles, descriptions, bullet points, and meta tags simultaneously. Implement a Brand Voice engine using Neon's pgvector extension to store style guide chunks and successful copywriting examples. When generating content, retrieve the most relevant brand guidelines via vector similarity search, inject them into the system prompt, and stream the generated outputs back to the UI in real time. ``` 5. **AI Photography & Batch Image Editor** — Integrate fal.ai to support background removal, upscaling, and lifestyle scene generation for product imagery. ``` Integrate the fal.ai API to handle product image editing and generation workflows. Build a visual asset management gallery where users can select a raw product photo and execute actions: background removal, high-res upscaling, and placing the product into a custom AI-generated lifestyle background or on-model scene. Store processed assets in Cloudflare R2 or Neon BLOB storage and link them back to the respective SKU records. ``` 6. **Asynchronous Batch Processing via Trigger.dev** — Set up Trigger.dev durable background jobs to handle massive multi-SKU enrichment and translation runs without serverless timeout failures. ``` Configure Trigger.dev in the Next.js application to handle long-running asynchronous batch workflows. Create a background task that processes a 1,000-SKU catalog enrichment and translation job (across 40+ languages). Implement step functions with built-in retries, exponential backoff for AI API rate limits, real-time progress logging stored in the database, and a live progress bar UI component in the dashboard reflecting job completion status. ``` 7. **Shopify Export & API Integration Sync** — Build bidirectional export endpoints and Shopify API integration to push enriched product details directly to ecommerce stores. ``` Build a secure integrations settings page and an export module that connects to the Shopify Admin REST/GraphQL API using OAuth access tokens. Implement a sync service that packages enriched PIM SKUs, attributes, tags, and media assets into compliant Shopify product payloads, pushes them upstream, logs sync success or failure states in an audit table, and allows users to retry failed pushes. ``` ### Cost vs paying **Starting costs (one-time):** - AI API Credits (Anthropic/OpenAI/Fal.ai): $50 - Total: ~$50 one-time **Ongoing costs (monthly):** - Vercel / Neon / Trigger.dev Free Tiers: $0 - API Usage (LLMs & Image Gen): $15 - $30/mo - Total: ~$20/mo - Paying for the SaaS instead: $99 - $500+/mo (Custom Enterprise) - Build time: 60-80 hours - AI tool credits: $40 (Cursor/Claude Pro subscriptions) - Break-even: N/A (Built for personal catalog control and learning) ## Sources - [Hypotenuse AI Official Website & Feature Pages](https://hypotenuse.ai) - [Y Combinator Directory - Hypotenuse AI (S20)](https://www.ycombinator.com/companies/hypotenuse-ai) - [PitchBook Company Profile - Hypotenuse AI](https://pitchbook.com)