ADR-0001: Stack choice — NestJS + Next.js + Postgres + Prisma

• Status: Accepted
• Date: 2026-04-17
• Deciders: Vitor Rodovalho, maintainers

Context

Panorama is a greenfield rewrite of two existing PHP systems (Snipe-IT, AGPL-3.0, Laravel; and SnipeScheduler-FleetManager, GPL-3.0, raw PHP). We need a stack that:

• Attracts a large contributor pool (open-source viability)
• Supports horizontal scaling for SaaS hosting and plausible enterprise on-prem
• Has a mature, typed ORM that we can trust for row-level multi-tenancy
• Lets us share typed DTOs between backend and frontend
• Has first-class i18n primitives (we ship trilingual at launch)
• Has good mobile paths (React Native / Expo pulls from the same package graph)
• Is not locked into any single cloud vendor

We considered three realistic options. Details below.

Decision

TypeScript monorepo driven by pnpm workspaces + Turborepo:

• Backend: NestJS 10 on Node 20, with BullMQ workers on a separate process
• Frontend web (end-user + admin): Next.js 14 App Router, React 18, Tailwind, shadcn/ui
• Mobile: React Native via Expo (deferred — ships in 1.1)
• Database: PostgreSQL 16, accessed via Prisma 5 (row-level tenancy enforced in Prisma middleware)
• Cache / queues: Redis 7 (+ BullMQ for deferred jobs and email queue)
• Object store: S3-compatible (MinIO in dev, AWS S3 / GCS / Azure Blob in prod)
• Event bus: NATS JetStream, optional in Community, recommended in Enterprise
• Search: OpenSearch, optional; Postgres FTS is the default

Shared code lives in packages/* (plugin SDK, UI kit, i18n bundles, shared domain types). All packages are typed end-to-end; no any without an ESLint justification comment.

Alternatives considered

Option A — Stay on PHP (Laravel 11)

• ✅ Cheapest migration path (SnipeScheduler is PHP; Snipe-IT is Laravel already)
• ✅ Laravel ecosystem has Passport, SCIM, SAML, Nova, etc.
• ❌ Contributor pool for modern OSS products skews JS/TS
• ❌ Weak mobile story (we'd need a second stack for React Native anyway)
• ❌ Type sharing between server and browser still awkward

Rejected because the long-term contributor-pool concern outweighs the short-term migration cost, and because we're rewriting anyway (new schema, new UX), so the "just fork Snipe-IT" shortcut is not actually available.

Option B — Go (backend) + Next.js (frontend)

• ✅ Excellent runtime performance and static binaries for on-prem
• ✅ Tight concurrency primitives for event-heavy workloads
• ❌ Smaller OSS contributor pool for fleet/asset-domain SaaS apps
• ❌ No meaningful type sharing between server and browser without codegen yak-shaving
• ❌ ORMs for Go (ent, gorm, sqlc) are each a compromise; none gives us Prisma's DX + middleware story

Rejected for contributor-pool and DX reasons. Reconsidered if a perf bottleneck appears that Node genuinely can't solve; mitigation is to rewrite that one service in Go later (e.g. the event bus relay).

Option C — Python (FastAPI + SQLAlchemy) + Next.js

• ✅ Huge contributor pool; strong data/ML adjacency (useful for predictive maintenance)
• ❌ Python's typing story is still weaker than TypeScript's for API contract sharing
• ❌ Two runtimes in the monorepo; harder onboarding
• ❌ SCIM/SAML server libraries are less battle-tested than Laravel-JS equivalents

Rejected for DX and single-language consistency.

Consequences

Positive

• Single language across web, admin, core API, plugin SDK, mobile → one toolchain, one CI
• Prisma + Zod + tRPC-or-OpenAPI gives us typed API contracts end-to-end
• The hiring market for TS/Nest is large and global
• Vercel / Cloudflare / Railway-style deploys all "just work" for the web tier

Negative

• Node's CPU-bound performance is weaker than Go's; PDF generation and report aggregation may need workers with tuned concurrency
• node_modules size is a recurring ops cost (CI cache, Docker image size); we mitigate with pnpm + standalone Next.js builds
• Prisma has occasional migration-UX footguns; we will codify a migration review checklist

Neutral

• NestJS is opinionated (modules, DI). It's a bet on structure over flexibility. For a 3+ year product, that trade favours us.

Related

• ADR-0003 Multi-tenancy (Prisma middleware choice is downstream of this)
• ADR-0006 Plugin SDK (SDK exposes typed hooks to NestJS modules)