0.13.1 - 2026-09-21

Added

• vertigo serve and vertigo watch now compress responses (brotli/gzip/zstd, by content
  negotiation). The wasm binary and wasm_run.js are then about 3 times smaller.
  Pass --disable-compression behind a proxy or CDN that already compresses.
  The fullstack template (vertigo new -t fullstack) does the same through
  App::wrap(Compress::default()).

Performance

• Reduced WASM sizes by around 10% by de-genericization.
• Reduces wasm_run.js by 9% - dead code removed, and different functions were optimized.
• SSR: Per-request WASM instantiation got about 24% cheaper (see fix below).

Fixed

• Hydration fail when the app subscribes to a value before the root dom! block.
• --proxy forwarded the upstream's Content-Encoding and Content-Length even though awc
  had already decoded the body, so proxying to an upstream that compressed produced a response
  no client could decode.
• SSR: WASM instantiation types order problem. Imports are now resolved by name, once at startup.
  Per-request instantiation got about 24% cheaper. A missing or mistyped import is now reported
  at startup, naming the import, instead of per request.

Internals

• Added tests for the JsJson codec (jsjson.ts).

• Added benchmarks:

  • task ssr-bench renders a set of pages and reports where the time went,
  • task hydration-bench loads a set of server-rendered pages in a real browser and reports
    how long hydration took and how many DOM mutations it performed,
  • task reactive-bench runs the reactive-graph benchmark suite,
  • task dom-bench runs the end-to-end benchmark suite.

  task bench runs all four benchmarks. task bench-compare runs all four benchmarks and
  compares them against the baseline.

  See benchmarks

• vertigo-cli gained an off-by-default ssr-timings feature, which is what exposes the
  per-phase breakdown through ServerState::request_timed. It compiles to nothing when off