Doc Search

doc-search

Supabase's Next.js OpenAI Doc Search Starter — a ⌘K dialog that answers a question about your docs: the MDX pages under pages/docs split into sections by heading, each section embedded and stored in Postgres with pgvector, the question embedded and matched against them, and an answer streamed from the closest sections — running on Cloudflare Workers + Static Assets through @opennextjs/cloudflare, embedding and answering with Workers AI through the Worker's AI binding, its index a Supabase project's database reached through Cloudflare Hyperdrive, both provisioned by Terraform through Orun and the schema applied by Orun's db-migrate, the way the rag-chatbot baseline does it, and packaged as an Orun baseline: one command takes a new product from nothing to live on stage and prod, in its own repository, under its own name — in ONE landing.

The template's code is the code. What the port changes is what the Workers runtime and a fresh, unattended deployment force: the database is reached with SQL over the Hyperdrive binding (lib/db/index.ts) instead of through Supabase's Data API with a service-role key; the models are Workers AI's (lib/ai/models.ts) — the platform brokers no AI vendor, and the Cloudflare account the app deploys into already is one — with upstream's OpenAI model behind an optional key for the answers; and the docs are embedded by the deployed Worker, not by the build. Upstream's pnpm build embedded every changed page with OpenAI and wrote it through the Data API, with both keys in the build; here the build only splits the docs into sections (lib/docs/index.generated.json), and the deploy's smoke step POSTs /api/embeddings, where the Worker embeds the pages whose checksum the database does not hold through its own AI binding. The column and the search function are sized to Workers AI's embedding model (768 dimensions, migration 020_workers_ai_embeddings). What the baseline adds is everything needed to make a copy of it somebody else's — built the way cirrus-baseline is. See specs/epics/doc-search for the design, the milestones and the as-built status, including every departure.

Instantiating a product

git clone --depth 1 --branch <baseline-tag> https://github.com/sourceplane/doc-search
cd doc-search
orun new --blueprint repo-blueprint.yaml --out ~/products/acme-docs --run-hooks --resume \
  --set reponame=acme-docs --set productname="Acme Docs" \
  --set githuborg=acme-inc --set subdomain=<workers-dev-subdomain>

Or, once the platform's registry lists it, from the Orun console (Baselines → doc-search → Build). Two phases: 01-product creates the repository, mints the five brokered credentials (Cloudflare and Supabase) and lands the whole product as one pull request, whose convergence on main provisions and deploys it per environment in a fixed order —

supabase ──▶ cloudflare-hyperdrive ──▶ db-migrate ──▶ web

— each edge mode: enforced on a push to main and advisory on a pull request; then 02-docs records the live deployment from probed facts. BOOTSTRAP.md is the operator guide; docs/phases/ has a page per phase.

It creates two Supabase projects, <reponame>-stage and <reponame>-prod. Supabase's free plan allows two active projects per organization, so the connected organization needs room for both (BOOTSTRAP.md, §0).

The new repository receives the product only — the app, the Terraform roots, the migrations, the tests, its CI, its tooling, a README of its own and its operating contract — renamed throughout to the product's name, repository, Worker, Supabase project and Hyperdrive config. None of this repository's machinery ships. No AI vendor key is asked for: Workers AI needs none, and OpenAI for the answers is the owner's to add afterwards (the product's README, "Use OpenAI for the answers").

What is here (the product)

Getting started

pnpm install
pnpm run wire:fixture   # re-render apps/web/wrangler.jsonc from its template (committed; this changes nothing)
pnpm typecheck
pnpm lint
pnpm test
pnpm build              # the docs index + next build + opennextjs-cloudflare build → apps/web/.open-next
pnpm --filter @saas/db run manifest:check

The build needs no database, no model and no Cloudflare credential. Running it locally — a Postgres with pgvector on localhost:5432 with the migrations applied, wrangler login for Workers AI, next dev or wrangler dev, then POST /api/embeddings to index the docs — is in the product's README (.baseline/product/README.md), which is what a product built from here carries.

How it maps to Vercel

CI