helixordevelopers

Try it now

Pick a decision pack, send it a payload, and read the decision the runtime returns: the action, the rules that fired, the remedy, a receipt hash and the measured latency. Nothing to install and no license needed.

No installNo licenseRuntime version: loading

Use synthetic data only

Do not paste real personal, customer or financial data. The playground stores nothing and does not log payloads, but it is a shared public endpoint. The examples below are made up.

Playground#

Loading the packs from the hosted runtime…

Beyond rules#

The packs above decide over the facts you send. These examples show an engine finding an answer that no fixed rule contains: searching ahead and proving a move, switching to an exact solve when the position allows it, and picking a strategy per situation while learning from outcomes. Every call is shown in an API inspector. Preview

Search with proofs

Four in a row

Play the engine. Each move shows the columns it searched and whether it has a proof of a forced win.

 Choosing the technique

Reversi

An estimate in the midgame, an exact solve with the final disc margin near the end, and the move says which.

 Strategy and learning

Arena duel

Watch the engine rank seven strategies every tick and carry what it learned from each episode into the next.

What runs behind the playground#

The playground is the same Decision Runtime you install, running on a Helixor server. The version it reports is the runtime version serving your request. Nothing is simulated in the page: every result shown above is the runtime's answer, and when the service cannot be reached the page says so instead of showing results.

  • Several decision types, one runtime. The data protection guard is the built-in example pack. The lending packs are compiled from playbooks and sealed with a server-side license, then evaluated over typed fields: one decides eligibility and searches your candidate changes for the lowest-cost recourse, the other verifies algebraic constraints with the interval solver.
  • Every result is one action from the pack's declared set, with every rule that fired, a remedy where the pack defines one, and a receipt hash over the pack, the decision and a digest of your input. See Receipts.
  • Latency. For the built-in pack the runtime reports it; for the licensed packs it is measured around the runtime call on the server. Network time to the server is not included.
  • Limits. Bodies up to 8 KB, a per-address rate limit, and a short timeout per evaluation. Nothing is stored.

In your own process the call does not leave your application: see the Quickstart to install the runtime and Write your own decision pack to author one.