Pipeline (Dev + Build Share One Core)

Source files → Resolver → Persistent dependency graph → Transform engine (OXC first, SWC fallback) → CAS (`.ionify/cas///...`) → Dev server / Bundler

Rust-Powered Core

Ionify's performance-critical operations (graph management, parsing, bundling) are implemented in Rust and exposed via native bindings. The TypeScript SDK orchestrates the workflow while delegating heavy lifting to the native core.

Two Hashes, One Mental Model

Ionify intentionally version-isolates artifacts so cached outputs remain correct:

  • `versionHash`: derived from config inputs (parser/minifier/define/resolve/etc). It partitions the graph + CAS.
  • `depsHash`: derived from `versionHash` + lockfile + optimizeDeps options. It partitions optimized deps under `.ionify/deps//`.

This is why Ionify can restart fast without “mystery stale cache” behavior.

Deps Optimizer (`/@deps/*`)

Bare imports are rewritten to deterministic stable URLs, for example:

On request, Ionify optimizes the dependency (Rust), writes it into `.ionify/deps//`, and persists a manifest so future runs can serve it immediately.

Hybrid Transformation Engine

Ionify uses a hybrid transformation strategy by design.

OXC is used as the primary engine for parsing and transformation, optimized for performance and modern JavaScript syntax. SWC acts as a fallback layer to ensure robustness and compatibility across edge cases and evolving ecosystems.

This approach allows Ionify to remain framework-agnostic while balancing speed, correctness, and long-term maintainability.

Storage

  • Graph persistence (native Rust implementation)
  • Transformed outputs stored in version-isolated CAS
  • Automatic invalidation via configuration hash

Why this architecture enables intelligence

Because Ionify persists the dependency graph and transformed outputs, the engine can observe patterns over time:

  • Which modules change frequently
  • Which transformations are expensive
  • How dependency structure affects rebuild cost

This data is the basis for future analyzer tooling and AI-assisted optimization.