Genesis AI

A marketplace of installable AI skills and prompt packs.

2
Catalogues: skills and prompts
1 click
Install
Curated
Listings

What they came with

A team's working prompts live in a Notion page, a Slack pin and somebody's screenshot. Each one gets copied, edited and quietly forked until nobody can say which version is current or what changed. The search that matters is by the job in hand — write the discovery brief, clean this export — not by the name whoever wrote it happened to choose. Genesis AI is the catalogue underneath that: skills published as versioned entries, installed into a workspace, and found by the task they do.

What the engagement covered

  • Searchable catalogue of skills and prompt packs by category
  • One-click install into the team's assistant
  • Publishing flow with versioning and moderation

Technical detail

Skills versioned, installs pinned

Every publish writes an immutable version row, and an install points at a version rather than at the skill. An author shipping a change cannot alter what a team is already running; upgrading is an explicit action with the two versions shown side by side.

Catalogue searched by task

The index covers the task description and worked examples, not just titles, using Postgres full-text with a pg_trgm fallback so a near-miss spelling still returns something. Keeping search in Postgres means one fewer service to run, at the cost of ranking that is simpler than a dedicated engine's.

Install copies rather than links

Installing writes the skill body into the workspace's own row, so a withdrawn or delisted skill keeps working for teams that already have it. The origin reference is kept only to surface update notices.

Model access proxied per workspace

Calls to OpenAI go through a server-side proxy keyed per workspace, so provider keys never reach the browser and usage is counted per install for rate limits. A workspace over its limit gets refused at the proxy rather than at the provider, which keeps the failure legible.

The stack

Interface

Next.jsReactTypeScriptTailwind CSS

Catalogue and data

FastAPIPostgreSQLpg_trgmAlembic

Skills and generation

OpenAICeleryRedis

Billing and hosting

StripeDockerDocker ComposeTraefik
Sector
AI marketplace
Audience
Teams adopting AI
Shape
Built, shipped and run by us
Stack
Next.js, PostgreSQL
Next case studyAn AI agent that answers, reschedules and confirms installs in two languagesA Phoenix solar and HVAC installer was losing rescheduling calls to voicemail while crews were on roofs. We put an inbound voice agent on their main line that verifies the caller, detects Spanish and translates live, offers real crew openings from the dispatch calendar, rebooks the job, texts a confirmation and writes every action back to the CRM, while dispatchers watch the call unfold in a live console.

Something like this to build?

Tell us what runs today and where it hurts. An engineer reads it and replies.

Start a projectAll case studies