Test report - May 24, 2026

Green, useful, and honest.

The current suite is good regression coverage for the SDK, runtime primitives, hub manifest, and live demo UX. It is not yet a full model-quality eval harness, and the gaps are tracked below.

Total 136 passing tests

JavaScript 85 Node unit tests

Rust 27 native runtime tests

Browser 24 Chromium e2e tests

What is covered

Real contracts, not just smoke tests.

The tests cover backend selection, mock streaming, tokenizer behavior, compiled WASM loading, SmartField and SmartForm propagation, task registry behavior, hub manifest shape, OpenAI-compatible shim behavior, and four full browser demos.

Coverage

88.02% line coverage.

The Node unit coverage report shows 88.02% line coverage, 80.83% branch coverage, and 83.28% function coverage across imported JavaScript modules. The lower spots are browser-only backend paths and hub iframe behavior.

Important signal

The WASM path is exercised.

Four tests load the compiled dhamaka-runtime.wasm from disk, instantiate it through the JavaScript engine wrapper, stream tokens, check deterministic output, and verify abort handling.

Known gaps

Good tests, not complete tests.

Playwright e2e should be wired into CI.
window.ai and Transformers.js need stronger browser-backend coverage.
Formula transforms need deeper edge-case matrices.
LLM quality needs evals, not only deterministic unit tests.

Commands run

Reproducible from the repo.

npm test
cargo test --manifest-path crates/dhamaka-runtime/Cargo.toml
find packages -name '*.js' -not -path '*/node_modules/*' | xargs -n1 node --check
npx playwright test