Immutable. This exact content is served forever at /api/v1/blob/4022d39d72a4d9ab.
---
name: "architecture-and-stack"
description: "Cloudflare-first platform selection. Decision trees for Workers, D1, R2, KV, DO, Queues, Vectorize, Containers, Sandboxes, Flagship, Agent Memory, Workflows v2. Default stack, override conditions, auth, data patterns, reliability."
metadata:
version: "2.1.0"
updated: "2026-05-03"
effort: "high"
model: "opus"
license: "Rutgers"
compatibility:
claude-code: ">=2.0.0"
agentskills: ">=1.0.0"
submodules:
- cloudflare-primitives.md
- data-patterns.md
- reliability.md
- auth-patterns.md
priority: 2
pack: "backend"
triggers:
- "architecture"
- "stack"
- "cloudflare"
- "d1"
- "workers"
paths:
- "concern:cloudflare-workers"
---
# 05 — Architecture and Stack
Default stack: `_kernel/standards.md#stack`. Override conditions below.
## Cloudflare-first decision tree
### Compute
- **Workers** (default) — every HTTP-serving surface, every cron, every queue consumer
- **Pages** — static-only marketing (rare; Workers + Assets binding usually better)
- **Containers** — non-JS runtimes (Playwright headful, ffmpeg, Python ML), browser rendering, build orchestration
- **Sandbox SDK** — generated/risky code execution before promotion to live runtime
### State
- **D1** (default) — relational, ≤10GB per database, Sessions API for read-replicas, Time Travel 30-day PIT
- **KV** — eventually-consistent k/v, cache, sessions, feature-flag state
- **R2** — object storage (images, PDFs, exports, backups), lifecycle Standard→IA after 30d
- **Durable Objects** — coordination + strongly-consistent storage (chat rooms, builder sessions, rate limiting per-user). SQLite-backed since Apr 2025.
- **Hyperdrive** — front external Postgres/MySQL (Neon, PlanetScale)
- **Vectorize** — semantic search, RAG embeddings (5M dim/index, topK 100, 10 metadata indexes)
### Async
- **Queues** — best-effort delivery, 5000 msg/sec, R2 event notifications
- **Workflows v2** — deterministic step-based, 50K concurrent, 300 creates/sec, 2M queued/workflow, `step.do` + `step.sleep` + `step.waitForEvent`
- **Inngest** — event-driven w/ better DX/observability (where it fits)
### AI
- **Workers AI** — Llama 3.3 70B FP8 free, Llama 3.1 8B FP8, Llama 4 Scout 17B vision
- **AI Gateway** — caching + rate-limit + fallback + logging for every LLM call
- **Vectorize** — embeddings + ANN search
## Override conditions (when CF isn't enough)
| Need | Fallback | Adapter pattern |
|---|---|---|
| Advanced SQL (RLS, OLAP, partial indexes) | Neon Postgres via Hyperdrive | `SqlPort` |
| Redis primitives at scale (sorted sets, streams) | Upstash Redis | `KvPort` |
| Sub-millisecond global state | Upstash QStash | `QueuePort` |
| Specific provider (OpenAI assistants, Anthropic batch) | Direct API via AI Gateway | `AiPort` |
| Vector + SQL co-located | Neon pgvector | `VectorPort` |
Adapters live in `libs/core/ports/`. Product code imports port, never vendor SDK directly. See `rules/cloudflare-hostable-supervisor.md`.
## Auth (default Clerk M2M JWT)
- **Clerk** — M2M JWT (free, networkless verification), passkeys, OAuth, magic links
- **Better Auth** — when Clerk pricing doesn't fit (rare)
- Hash API keys at rest. Audit log every sensitive action.
- Tenant isolation: every table carries `org_id`, every query filters by it (404 on mismatch, never 403)
## Data patterns
### D1 typed bindings
```toml
# wrangler.jsonc
[[d1_databases]]
binding = "DB"
database_name = "myapp"
```
- `wrangler types` against `compatibility_date` + bindings (preferred over hand-maintained Env interface)
- Drizzle v1 RQBv2 + Zod for query + validation
- Batch via `db.batch([...])` (no transactions in D1)
- Sessions API for read-replicas: `db.withSession(bookmark)`
- Time Travel 30-day PIT: `wrangler d1 time-travel restore`
### R2 patterns
- Per-extension content-type on upload
- Lifecycle Standard → IA after 30d for backups/exports
- R2 event notifications → Queues at 5000 msg/sec for thumbnailing/indexing
- Versioning for asset rollback
### DO patterns
- One DO per stateful entity (chat room, builder session, rate-limited user)
- SQLite-backed by default since Apr 2025 — 10GB per DO
- Direct stub: `env.MY_DO.getByName(name)` (replaces `idFromName` → `get` two-step)
- Alarm misfires → idempotent handler
## Reliability
- Workers CPU 10ms free / 50ms paid default (configurable 5min)
- Wall time 30s paid
- `ctx.waitUntil()` for async post-response work
- `ctx.passThroughOnException()` for graceful degradation
- WebSocket payload up to 32 MiB (CF Workers + DO)
- JSRPC payload up to 32 MiB
## Cost discipline
- Workers free tier: 100k requests/day. Sufficient for most solo projects.
- Workers Paid: $5/mo + $0.30/M requests + $12.50/M CPU-ms
- D1: 5GB free, $0.75/GB-month, $0.001/M reads, $1/M writes
- R2: 10GB free, $0.015/GB-month, $0/egress (huge win vs S3)
- Workers AI: Llama 3.3 70B FP8 FREE
- AI Gateway: free
- Reference: solo SaaS at <$100k/mo MRR stays 10-100× cheaper than AWS-equivalent on CF
## Default config (`wrangler.jsonc`)
```jsonc
{
"name": "myapp",
"main": "src/worker/index.ts",
"compatibility_date": "2026-04-15",
"compatibility_flags": ["nodejs_compat"],
"observability": { "enabled": true },
"secrets_required": ["CLERK_SECRET_KEY", "RESEND_API_KEY"],
"d1_databases": [{ "binding": "DB", "database_name": "myapp" }],
"kv_namespaces": [{ "binding": "CACHE", "id": "..." }],
"r2_buckets": [{ "binding": "BUCKET", "bucket_name": "myapp-assets" }],
"ai": { "binding": "AI" }
}
```
## Decision template (use for every architecture call)
1. Can CF primitive do this? → Use it.
2. Does this need adapter for portability? → Adapter only if real business case.
3. Cost projection at 10× current scale → Still affordable?
4. Failure mode → Graceful degradation defined?
5. Migration path → If we have to leave CF, what does it cost?
## See submodules: cloudflare-primitives.md, data-patterns.md, reliability.md, auth-patterns.md.