npm package discovery and stats viewer.

Discover Tips

  • General search

    [free text search, go nuts!]

  • Package details

    pkg:[package-name]

  • User packages

    @[username]

Sponsor

Optimize Toolset

I’ve always been into building performant and accessible sites, but lately I’ve been taking it extremely seriously. So much so that I’ve been building a tool to help me optimize and monitor the sites that I build to make sure that I’m making an attempt to offer the best experience to those who visit them. If you’re into performant, accessible and SEO friendly sites, you might like it too! You can check it out at Optimize Toolset.

About

Hi, 👋, I’m Ryan Hefner  and I built this site for me, and you! The goal of this site was to provide an easy way for me to check the stats on my npm packages, both for prioritizing issues and updates, and to give me a little kick in the pants to keep up on stuff.

As I was building it, I realized that I was actually using the tool to build the tool, and figured I might as well put this out there and hopefully others will find it to be a fast and useful way to search and browse npm packages as I have.

If you’re interested in other things I’m working on, follow me on Twitter or check out the open source projects I’ve been publishing on GitHub.

I am also working on a Twitter bot for this site to tweet the most popular, newest, random packages from npm. Please follow that account now and it will start sending out packages soon–ish.

Open Software & Tools

This site wouldn’t be possible without the immense generosity and tireless efforts from the people who make contributions to the world and share their work via open source initiatives. Thank you 🙏

© 2026 – Pkg Stats / Ryan Hefner

dag-ml-wasm

v0.3.44

Published

WASM bindings for DAG-ML JSON contracts.

Readme

dag-ml WASM bindings

Browser-friendly bindings for DAG-ML JSON contracts.

The WASM package exposes validation, DSL compilation, execution-plan construction and synchronous host-controller execution over UTF-8 JSON strings. Artifacts and data-buffer ownership remain outside the binding.

For parallel browser HPO, call host_hpo_search_parallel_json with a dispatcher that returns a Promise for each serialized candidate task. Each Web Worker loads its own WASM instance and calls host_hpo_evaluate_worker_task_json with candidate-local controller state. The core dispatches a whole trial window before awaiting results, validates scores and checkpoints, then tells the optimizer in trial order. host_hpo_search_json remains the synchronous single-worker route. For progressive pruning, tagged fold tasks call host_hpo_evaluate_worker_fold_json and return to the coordinator after each fold; pruned candidates never execute later folds. Complete candidates run one full native FIT_CV pass for the exact global OOF result.

contract_manifest_json() returns a stable JSON manifest with the package version, supported contract ids, exported Python/WASM function names and shared fixture digests. Browser integrations should check it before accepting cached pipelines or persisted nirs4all-core workspaces.

Build

cargo test -p dag-ml-wasm
node_out_dir="$PWD/target/wasm/dag-ml-wasm"
wasm-pack build crates/dag-ml-wasm --target nodejs --out-dir "$node_out_dir" --release
node scripts/smoke_wasm_bindings.cjs "$node_out_dir"
web_out_dir="$PWD/target/wasm-web/dag-ml-wasm"
wasm-pack build crates/dag-ml-wasm --target web --out-dir "$web_out_dir" --release
node scripts/smoke_wasm_web_bindings.mjs "$web_out_dir"

JavaScript Surface

import init, {
  LocalImplementationRegistry,
  contract_manifest_json,
  loss_execution_attestation_json,
  validate_pipeline_dsl_json,
  compile_pipeline_dsl_artifact_json,
} from "./pkg/dag_ml_wasm.js";

await init();
const manifest = JSON.parse(contract_manifest_json());
validate_pipeline_dsl_json(JSON.stringify(dsl));
const artifact = JSON.parse(compile_pipeline_dsl_artifact_json(JSON.stringify(dsl)));

Local JavaScript losses and metrics are retained by a WASM registry and never serialized into DAG-ML contracts:

const implementations = new LocalImplementationRegistry();
implementations.register_loss(JSON.stringify(lossReference), weightedLoss);

const loss = implementations.resolve_training_loss(
  JSON.stringify(trainingLossRole),
  "FIT_CV",
);
const value = loss(target, prediction);
const attestation = JSON.parse(
  loss_execution_attestation_json(JSON.stringify(trainingLossRole), "FIT_CV"),
);

Controllers should bind from the native NodeTask rather than resolving a role and constructing lineage independently:

const binding = implementations.bind_training_loss(
  JSON.stringify(nodeTask),
  0,
);
const loss = binding.invoke;
const value = loss(target, prediction);
const requiredAttestation = JSON.parse(binding.required_attestation_json);
binding.free();

The role index is zero-based. Binding validates that the task's ordered loss requirements exactly match its active roles. A controller may copy the returned attestation into NodeResult.lineage.loss_attestations only after loss(...) returns successfully.

For scheduler execution, lower the roles into the native plan and execute that plan rather than rebuilding a loss-free campaign plan:

const planJson = build_execution_plan_with_training_losses_json(
  planId,
  JSON.stringify(graph),
  JSON.stringify(campaign),
  JSON.stringify(controllerManifests),
  JSON.stringify(trainingLossRoles),
);

const resultsJson = execute_execution_plan_phase_json(
  planJson,
  JSON.stringify(controllerManifests),
  runId,
  rootSeed,
  "FIT_CV",
  (controllerId, taskJson, exactSeed) =>
    controller.invoke(controllerId, taskJson, exactSeed),
);

The native lowerer replaces all plan loss roles, groups them by node and sorts them canonically before validating controller capabilities. At execution, every manifest embedded in the plan must exactly match the independently supplied trusted controller registry before any callback runs. The scheduler then checks that each callback result contains exactly the task's required loss attestations in the same order.

The callback's exactSeed argument is a decimal string (or null), avoiding precision loss for native u64 seeds beyond JavaScript's safe-integer range. The callback may set NodeResult.lineage.seed to null; the WASM bridge then injects the authoritative native seed before scheduler validation. The returned resultsJson still contains native numeric u64 values; use a lossless JSON integer parser, or preserve the raw JSON, when inspecting lineage seeds exactly.

For host hyperparameter search, host_hpo_search_json(planJson, manifestsJson, envelopeJson, requestJson, checkpointJson, controller, optimizer) runs the same sequential native search as Rust and PyO3. The controller has the signature shown above. The synchronous optimizer(operation, payloadJson) must return a JSON string: ask returns {"params": {...}} or {"params": null}; report_intermediate returns {"prune": true|false}; tell, pruned, fail, and prepare_terminal return {"ok": true}; checkpoint returns {"continue": true|false}. prepare_terminal receives a sealed prospective checkpoint before an optimizer transition, so a browser host should persist it durably before replying. To resume, pass the last published native checkpoint as checkpointJson and reconcile any prepared optimizer transition first with recover_host_hpo_checkpoint_json(checkpointJson, preparedJson, interruptedJson); use "null" and "[]" when there is no prepared terminal or interrupted trial. This export runs one candidate at a time; browser worker parallelism remains an adapter gap.

JavaScript-local descriptors use binding:javascript and a host_local or portable_registered lifecycle. A Web Worker must populate its own registry; functions are not cloned, posted, or embedded in replay artifacts. Resolution is rejected when the exact descriptor or phase does not match.

Rust-side validation failures are returned as JSON strings with the ADR-11 descriptor fields category, code, severity, message, remediation_hint and context.