Phase 1: Prisma Postgres Setup

Goal

Create a clean Prisma Postgres environment for TendSocial with split connection settings for runtime traffic versus migrations/admin tooling. This phase creates and seeds the Prisma database. It does not migrate production data from Neon.

Why this matters

Prisma Postgres provides pooled and direct connection paths. Application traffic should use the pooled connection. Migrations, schema tools, Studio, dump/restore, and long-running/session-dependent work should use the direct connection. Prisma documents that the pooler uses PgBouncer in transaction mode, where session state does not persist between transactions; it also warns that using the direct hostname for application traffic can silently exhaust connections under load. Source: Prisma Postgres connection pooling.

Environments to create

Create at least:

tendsocial-prisma-dev
tendsocial-prisma-staging
tendsocial-prisma-benchmark

Recommended purpose:

Environment	Purpose
dev	Local and developer smoke testing.
staging	Production-like pre-release app testing.
benchmark	Repeatable destructive benchmark/seeding runs.

Required env vars

Use split URLs everywhere:

# Runtime traffic: API handlers, workers, serverless functions
DATABASE_URL="prisma-postgres-pooled-url"

# Admin traffic: Prisma CLI, migrations, Studio, dump/restore
DIRECT_DATABASE_URL="prisma-postgres-direct-url"

In Prisma 7, keep schema.prisma datasource free of deprecated url / directUrl fields:

datasource db {
  provider = "postgresql"
}

Configure Prisma CLI/admin datasource routing in apps/backend/prisma.config.ts:

import "dotenv/config";
import { defineConfig } from "@prisma/config";

export default defineConfig({
  datasource: {
    url: process.env.DIRECT_DATABASE_URL,
  },
});

Runtime application code still passes process.env.DATABASE_URL to new PrismaPg({ connectionString }).

Runtime placement

Update:

backend/.env
backend/.env.local
backend/.env.test
backend/.env.benchmark
Cloud Run environment variables
CI/CD secrets
Vercel environment variables if any DB-connected route exists there

Rules:

Consumer	URL
Backend API runtime	DATABASE_URL pooled URL
Background workers	DATABASE_URL pooled URL
Serverless functions	DATABASE_URL pooled URL
prisma migrate deploy	DIRECT_DATABASE_URL through prisma.config.ts
prisma db push	DIRECT_DATABASE_URL through prisma.config.ts
prisma db pull	DIRECT_DATABASE_URL through prisma.config.ts
Prisma Studio	DIRECT_DATABASE_URL through prisma.config.ts
pg_dump / pg_restore	DIRECT_DATABASE_URL direct URL
Long-running maintenance queries	DIRECT_DATABASE_URL direct URL

Schema creation

Preferred production-like path:

npx prisma migrate deploy

Disposable benchmark path only:

npx prisma db push

Then generate the client:

npx prisma generate

Initial seed

Create or update:

prisma/seed.ts

Seed enough data to reproduce TendSocial behavior:

Data group	Minimum seed
Users	3-5
Workspaces	2-3
Workspace members	3-10
Brands	2 per workspace
Social accounts	Placeholder Google, Instagram, LinkedIn, TikTok, Facebook
Campaigns	5-10
Short-form content records	100-500
Blog posts	25-50
Calendar/scheduled posts	60-180
AI draft records	100+
Publishing jobs	50+
Activity/audit rows	Enough to test current behavior

Run:

npx prisma db seed

Suggested seed profile

Create two profiles:

seed:small      realistic solo/dev
seed:benchmark 50-user equivalent content volume

Example scripts:

{
  "scripts": {
    "db:setup:prisma": "prisma migrate deploy && prisma generate && prisma db seed",
    "db:seed:small": "SEED_PROFILE=small prisma db seed",
    "db:seed:benchmark": "SEED_PROFILE=benchmark prisma db seed"
  }
}

Smoke tests

Run these after setup:

pnpm test:smoke
pnpm test:db
pnpm dev

Manual smoke flow:

1. Start backend against Prisma Postgres.
2. Start frontend.
3. Log in with Supabase auth.
4. Load workspace.
5. Open dashboard.
6. Open content list.
7. Open editor.
8. Save a draft.
9. Open calendar.
10. Schedule a post.

Acceptance criteria

Item	Target
Prisma Postgres DBs created	Dev, staging, benchmark
Runtime traffic uses pooled URL	Yes
CLI/admin traffic uses direct URL	Yes
Schema can be created from repo	Yes
Seed is repeatable	Yes
App boots against Prisma Postgres	Yes
Smoke tests pass	Yes
No production data migration required	Confirmed

Sources

• Prisma Postgres connection pooling
• Prisma pricing