Canvas Capabilities & Current Limitations
This section describes what can and cannot currently be represented on the PinPole canvas. Gaps noted here are being addressed across Phases 1–3 of the product roadmap — see Upcoming Features for timeline.
| Structural pattern | Support | Known gap |
|---|---|---|
Linear traffic flow (e.g. Route 53 → API Gateway → Lambda → DynamoDB) | ✅ Full | None |
Fan-out (SNS → multiple SQS) | ✅ Full | No visual distinction between synchronous call edges and async publish edges — both render as identical arrows |
| Async / event-driven (SQS, SNS, EventBridge, Kinesis, Step Functions) | 🔶 Partial | No visual semantic difference between a synchronous API call and a queued message on the canvas |
| Caching layers (CloudFront, ElastiCache) | 🔶 Partial | No representation of cache hit vs. miss traffic paths — the canvas shows a single arrow through a cache node |
| Resilience patterns (circuit breaker, DLQ, active-passive pairs) | 🔶 Partial | AI recommendations correctly generate these node sets, but standby/passive paths look identical to active paths visually |
| Containment hierarchy (VPC, AZ, Subnet, Security Group) | 🔴 Minimal | Nodes are in the catalogue but are not spatial envelopes — a Lambda node has no positional relationship to the subnet it would occupy in a real deployment |
| Network security zones (Internet → WAF → Public subnet → Private subnet → Data tier) | 🔴 Minimal | No zone envelopes; no enforcement preventing a private-subnet service from being wired directly to an internet-facing endpoint without the correct intermediary |
| Cross-boundary connectivity (VPC peering, Transit Gateway, PrivateLink) | 🔴 Minimal | Entirely dependent on containment hierarchy, which is currently minimal |
| CQRS, CDC, event sourcing, sharding | 🔴 Minimal | No purpose-built connection semantics for these patterns |
Practical guidance for current users
When designing architectures that include resilience pairs (active-passive), async decoupling, or multi-tier network zones, trust the AI Recommendations output over the visual representation. The AI engine correctly generates the node sets and connections for these patterns — the visual language to distinguish them is the part that is still being built.