MorphicCloudflareSupabase

Morphic

An AI answer engine with cited sources.

morphic

Vercel's Morphic template (miurla/morphic) — an AI answer engine with a generative UI: it searches the web and answers with its sources cited, quick and adaptive search modes, a model picker over OpenAI, Anthropic, Google and any OpenAI-compatible API, chat history and sharing, a library of notes and files, multi-user accounts with Better Auth — running on Cloudflare Workers + Static Assets through @opennextjs/cloudflare, its chat history and accounts a Supabase project's database reached through Cloudflare Hyperdrive, both provisioned by Terraform through Orun and upstream's schema applied by Orun's db-migrate — the way the Lumen 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 Cloudflare account must be on Workers Paid: the Worker is about 7.7 MiB gzipped, over Workers Free's 3 MiB.

The template's code is the code. What the port changes is what the Workers runtime and a fresh, unattended deployment force: the database client is built per request over the Hyperdrive binding behind upstream's own db (lib/db/index.ts); Better Auth's signing key is generated in the database by a migration instead of typed; the instrumentation loads OpenTelemetry only when Langfuse tracing is on; the fetch guard's undici dispatcher steps aside inside a Worker, which has no host network to guard; SearXNG's crawling route, which needs jsdom and TCP Redis, is not carried; and a final migration shuts Supabase's Data API out of every table. Models and search are the owner's keys, set as Orun secrets after the build — the platform brokers neither — so a fresh product serves with "No enabled model available" until they are. 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/morphic 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/morphic
cd morphic
orun new --blueprint repo-blueprint.yaml --out ~/products/acme-search --run-hooks --resume \
  --set reponame=acme-search --set productname="Acme Answers" \
  --set githuborg=acme-inc --set subdomain=<workers-dev-subdomain>

Or, once the platform's registry lists it, from the Orun console (Baselines → morphic → 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 model or search key is asked for: the owner sets them as Orun secrets afterwards, with the address of the account that may sign up first (the product's README, "First steps").

What makes it a baselineWhere
The console card (inputs, secrets, programme)blueprint.yaml
The build document (modules, phases, hooks)repo-blueprint.yaml
This tree's identity, the product's README and the bootstrap's task contracts.baseline/
The rename engine (product-name seam: apps/web/lib/site.ts)tooling/rebrand/rebrand.mjs
The bootstrap's helpers (deferrals, redeploy marker, work-tree strip)tooling/bootstrap/
The deployment-record rendererhooks/render-deployment-docs.sh
The product's context packai/context/
The baseline's tests: contract, leak, offline bootstrap and plans, end to end (with an OpenAI-compatible stand-in)testing/, .github/workflows/baseline.yml
Cutting a baseline-vN tag.github/workflows/tag.yml

What is here (the product)

SurfaceWhere
The product's nameapps/web/lib/site.ts
The answer engine: the page, the chat route, the research agent and its toolsapps/web/app, apps/web/lib/agents, apps/web/lib/tools
Models and search providersapps/web/lib/utils/registry.ts, apps/web/lib/models, apps/web/lib/tools/search/providers
Accounts (Better Auth), the generated signing keyapps/web/lib/auth, apps/web/lib/auth/better-auth/config.ts
Drizzle schema and the per-request clientapps/web/lib/db (index.ts)
The database check the deploy smoke-testsapps/web/app/api/health/route.ts, apps/web/lib/health.ts
Workers build and deploy contractapps/web/next.config.mjs, open-next.config.ts, wrangler.template.jsonc → wrangler.jsonc (committed), component.yaml
The owner's keys, handed to the Worker at deployapps/web/component.yaml (optionalSecretEnv), tooling/wire/worker-secrets.mjs
The schema as migrations, and Lumen's runnerpackages/db (@saas/db); apps/web/drizzle is the same SQL for drizzle-kit
The Supabase project, the Hyperdrive config, the migration stepinfra/terraform/supabase, infra/terraform/cloudflare-hyperdrive, infra/db-migrate
Teststhe template's own (apps/web, vitest, 150 files), tests/web, tests/db

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              # next build --webpack + opennextjs-cloudflare build → apps/web/.open-next
pnpm --filter @saas/db run manifest:check

The build needs no database, no model, no key and no Cloudflare credential. Running it locally — a Postgres on localhost:5432, pnpm --filter @app/web migrate, a model key and a search key in apps/web/.dev.vars, next dev or wrangler dev — is in the product's README (.baseline/product/README.md), which is what a product built from here carries.

How it maps to Vercel

VercelHere
Next.js hosting (Turbopack build, Node.js proxy, streamed chat route)One Cloudflare Worker (.open-next/worker.js) + Static Assets through OpenNext, Next 16.3.8 (the adapter's floor; upstream locks 16.2.6), built with webpack: Turbopack's server chunks duplicate modules per route, and the Worker came out at 11.7 MiB gzipped, over Workers Paid's 10 MiB
DATABASE_URL (Neon or any Postgres), one postgres client at module scopeThe Supabase project's Postgres through a Hyperdrive config; db is a proxy onto a client built per request from the HYPERDRIVE binding's connection string
bun run migrate (drizzle) or the Docker entrypointpackages/db + infra/db-migrate: upstream's twenty-one migrations byte for byte, then the generated Better Auth key and the Data API shut out, applied through Supabase's Management API before the Worker deploys; drizzle-kit stays for local work (apps/web/drizzle, the same files)
BETTER_AUTH_SECRET and BOOTSTRAP_ADMIN_EMAIL typed into .env.localThe key generated in app_private.secrets by migration 200 (a Worker secret still wins); the bootstrap address an Orun secret; sign-up invite-only from the first deploy
Model and search keys in the Vercel project's environmentOrun secrets per environment, handed to the Worker as Worker secrets by the deploy (optionalSecretEnv, worker-secrets.mjs), never seen by the build
SearXNG (Docker Compose) and its /api/advanced-search crawl (jsdom, Redis over TCP)Tavily by default, as upstream defaults; Exa, Firecrawl, Brave, or your own SearXNG at its basic and advanced depths — the crawling route is not carried
Langfuse/OpenTelemetry registered at startLoaded only when ENABLE_LANGFUSE_TRACING=true
undici connect-time DNS guard on every fetch the agent makesThe same off Workers; inside a Worker (no host network) literal private addresses are refused on every hop and global_fetch_strictly_public keeps fetches on the public internet
A Postgres database created in the dashboardinfra/terraform/supabase (one project per environment) + infra/terraform/cloudflare-hyperdrive (one config per environment), ids and credentials as orun secrets
"Deploy with Vercel" (clone the template)orun new --blueprint repo-blueprint.yaml — a new repository, branded, provisioned, deployed

CI

WorkflowWhat it runsShips to a product
ci.ymlthe Orun pipeline, and the only workflow a product carries: orun plan --changed, then one job per lane. A pull request runs the verify lanes (the app built and dry-deployed per environment, both Terraform roots checked and planned, the migrations planned); no lane runs the typecheck, lint or test scripts — baseline.yml's e2e tier runs them on the bootstrapped product before every tag, and a product's developers run them before they push. The bootstrap's own landings (orun/BASE-*) run plan alone and nothing waits for it — the landing was checked in the build box; a merge applies, migrates and deploys stage → prod. It plans nothing while intent.yaml names the placeholder workspace ws_CHANGE_ME, as this repository does. A manual run (Orunbase's Deploy button, or Run workflow on main) plans the environment and components it names under the push-to-main trigger, so they deploy as a merge deploys themyes
baseline.ymlthe baseline's own tiers: contract tests; the blueprint read by the pinned orun, the leak gate, the offline bootstrap and the product's own plans; the product built (the template's own test suite with it), migrated into a local Postgres with Supabase's roles, dry-deployed, served as it deploys, then given a stand-in model, asked a question and the chat kept; the Data API shut outno
tag.ymlcuts a baseline-vN tag (manual)no