Runtime & App
Bun + Next.js 16 + TypeScript
High throughput with modern App Router primitives and strict typing across boundaries.
Architecture Ledger
A production-ready stack chosen for speed, control, and AI.
Every package was picked for a reason. This page breaks down each layer — runtime, data, AI, billing, and observability — along with the tradeoffs you should know before you ship.
LayerChoicesWhy this matters
Data & State
PostgreSQL 17 + Drizzle ORM + Redis
Durable relational core with type-safe access and low-latency caching/rate limiting.
AI Fabric
Vercel AI SDK + Groq/OpenAI
Provider flexibility, stream orchestration, and fast fallback options for reliability.
Memory & Retrieval
Qdrant + Neo4j
Vector similarity plus relationship modeling for richer context and long-horizon retrieval.
Ops & Async
BullMQ + Docker + OpenTelemetry
Background execution, reproducible environments, and traceable production behavior.
Monetization
Stripe + Better Auth
Secure subscription lifecycle and identity controls for usage-gated products.
Tradeoff
Vector + graph instead of vector only
Higher setup complexity, but stronger retrieval quality for connected knowledge domains.
Tradeoff
Bun runtime with standard tooling
Faster local loops while preserving compatibility with mature TypeScript workflows.
Tradeoff
Provider-agnostic AI routing
Slight orchestration overhead for better latency/cost control and failover safety.
Practical result
The stack is engineered so teams can keep momentum: reliable persistence, flexible model routing, and production instrumentation are solved together instead of patched in as separate projects.