Feature request: Native enterprise architecture diagram generation (XML?) beyond Mermaid etc

I’m an enterprise architect and I use ChatGPT and Codex heavily for architecture work. Right now, diagram generation is a major gap

Mermaid or similar are useful for simple flows, but it is not suitable for real enterprise and solution architecture diagrams at scale. It struggles with layout control, complex notation, nested groupings, rich metadata, and producing artefacts that teams can actually edit and standardise.

Architects need diagrams as a primary deliverable. If ChatGPT can generate a diagram artefact that is immediately usable in standard tooling, it becomes materially more valuable for enterprise work. Without this, ChatGPT remains limited for architecture-heavy workflows

Request
Add first-class support for generating editable architecture diagrams in formats that teams already use for example draw-io. At the moment chatGPT can produce such diagrams but it’s hit and miss - they either are broken from the start or not fir for purpose and most of the time waste of time.

For me and I bet most of my peers what “good” looks like would be:

  • Generate valid files that can be opened directly in platforms like draw-io and edited
  • Support enterprise patterns such as C4, ArchiMate-style viewpoints, integration diagrams, sequence diagrams, deployment views, ERDs and high-level capability maps
  • Produce diagrams with sensible layout, alignment, spacing, and readable connector routing
  • Support containers, swimlanes, grouping, layers, and reusable templates
  • Support icons and shape libraries for common platforms and vendors (cloud, data platforms, identity, messaging, SaaS)
  • Allow embedding metadata on nodes and edges (IDs, ownership, interfaces, protocols, NFRs, risks)
  • Provide deterministic output options so the same prompt regenerates the same diagram structure when needed

It would be truly transformational for me if I was able to work on defining an architecture and strategy then describing to chatGPT for it to produce the visuals. This would literally save me days of work

Hey @FreddieDR, totally get why you would want that. Appreciate you taking the time to lay this out, your insights are genuinely helpful. I can’t share a timeline right now, but this is solid context on what would make things smoother. We’re looking at improvements with safety and reliability in mind, and I’ll pass this along internally.

I wanted to confirm that 5.4-Thinking is a major improvement for generating drawio diagrams. I was able to sketch capability maps and more complex context architecture diagrams with 5.4, generating diagrams that still needed work but were finally usable! Fixing the diagrams is now manageable and overall reduced the amount of time needed to create them saving me hours of work which I can now dedicate to actual strategic thinking! Looking forward to the next model iteration