TALA Auto-Layout Algorithm Goes Open-Source Under MPL-2.0

Terrastruct announced that TALA, its proprietary auto-layout algorithm for the D2 diagramming language, is now open-source under the MPL-2.0 license — the same license as D2 itself. TALA is an orthogonal layout engine designed specifically for software architecture diagrams, contrasting with DAG-based engines like Dagre and ELK that grow in a single direction. The algorithm blends techniques from published graph-drawing research with original methods, optimizing for aesthetics including symmetry, median distance, flow, and node clustering. A key differentiator is support for partial coordinate locking: users or AI agents can specify fixed positions for certain nodes while TALA handles routing for the rest — a capability the team highlights as especially useful for agentic AI workflows. Tradeoffs include non-determinism across different inputs (due to seed-based randomization), weaker performance on long DAG-style graphs, and nonlinear scaling for large diagrams. Runtime benchmarks are available at github.com/d2lang/d2-benchmarks. TALA ships bundled in D2 v0.9.0 and can be tested at play.d2lang.com, which runs fully client-side.