Software Graph

Appearance

Return to top
On this page

Architecture Overview

Software Graph combines three layers:

  1. Service graph modeling (sg-mesh.yaml) for nodes and dependency edges.
  2. Contract extraction + diffing for interface-level change detection.
  3. Operational workflows (CI, propagation, and pointer sync) that keep repos aligned.

Node Types

Common node kinds in the mesh include:

  • python-api
  • react-spa
  • python-sdk
  • node-sdk
  • infrastructure dependencies

Not every node exposes OpenAPI, but every node can participate in dependency traversal.

Edge Semantics

Edges are directional and typed.

If A -> B, then A consumes B. A change in B may require updates in A.

Edges can also carry tag-group information, which enables selective propagation when only specific API areas are breaking.

Branch-Scoped Mesh State

The mesh is represented by the meta-repo branch plus pinned submodule commits:

  • On dev, pointers should reference each submodule's dev line.
  • On main, pointers should reference each submodule's main line.

Submodule pointer updates are event-driven by GitHub Actions and recorded as normal commits in the meta-repo. The auto-sync lane is dev; main pointer updates are promoted from meta dev via PR.

Operational Architecture

null

For concrete command flows, see Operations / Workflow.