Skip to content

Agent skill that turns natural language, code, Terraform/K8s, SQL, OpenAPI, AsyncAPI, Protobuf and GraphQL sources into editable, tested draw.io architecture diagrams: incremental sync, multi-view projection, drift diff, CI architecture tests, whiteboard derasterize, interactive HTML/PPTX/Mermaid exports.

2.0 KB
Raw Blame

Roadmap: Architecture Studio 3.1

This backlog is based on the 3.0 release gate, checked showcase, and dogfood run against this repository. It is intentionally ordered by observed usefulness.

P0 — semantic fidelity ✅ shipped in 3.2.0

  • Add source profiles so code modules use kinds such as module, library, adapter, and command instead of defaulting to service. ✅ (module/library/command)
  • Scope architecture contracts by kind/profile; ownership and production observability rules should not fire on ordinary source modules. ✅ (kinds + profile field)
  • Emit exact file-and-line provenance from language importers rather than only the scanned root path. ✅ (per-file path + Python edge line numbers)
  • Report when a projected view falls back to the entire model, including which metadata would make that view distinctive. ✅ (fallback/fallback_reason/hint)

P1 — richer analysis

  • Add policy parameters and per-rule severity overrides without introducing an expression evaluator that executes user content.
  • Model dependency direction and failure semantics explicitly: caller/callee, upstream/downstream, retry, circuit breaker, queue buffering, and redundancy.
  • Add a view-difference summary so CI can detect five nominal views that are effectively identical.
  • Support deterministic timestamps through SOURCE_DATE_EPOCH for reproducible checked artifacts.

P2 — release confidence and presentation

  • Add cross-platform visual golden tests for Graphviz and grid fallback layouts.
  • Export a small PNG/SVG gallery from the semantic showcase in release CI.
  • Track importer precision, policy signal-to-noise, manual-layout preservation, view differentiation, and Story accessibility as release metrics.

Exit criteria

3.1 is ready when a codebase dogfood run produces no architecture-only policy noise, provenance identifies source lines, fallback views explain themselves, and the showcase passes on macOS, Linux/Xvfb, and the pure-Python fallback.