Why We Chose Rust for Our Core Engine
NodeFox uses a Rust and WebAssembly execution core to run workflows in the browser. The engine coordinates a graph of model calls, code, data, integrations and human decisions. Its job is to determine which work can execute, carry outputs between steps and expose the state of the run.
Why the runtime matters
A workflow can wait for an API response, combine results from multiple branches, pause for a person and then write to another system. Those dependencies need an explicit execution model. In NodeFox, data connections and activation conditions define that behavior in the graph itself.
This gives builders a concrete place to inspect the process: the steps, their inputs and outputs, and the conditions that allow an action.
Rust and WebAssembly
Rust supports a compiled execution core without a garbage collector. Its ownership and type systems help prevent classes of memory-safety and concurrency errors in safe Rust. These properties support demanding runtime work; they do not eliminate every possible application bug or guarantee a particular latency.
WebAssembly makes that core usable in the browser. NodeFox can author and execute workflows in a browser session, with worker-based operations. The browser must remain open for that execution mode. For background operation, NodeFox also provides a headless runner on supported machines; see the deployment guide for its requirements and integration differences.
Predictable orchestration, variable inputs
Deterministic orchestration means explicit dependencies, synchronization and conditions govern execution. It does not mean that a model will always return the same answer, that an external API will return the same data, or that every run takes the same time.
Use code for exact checks, validate model outputs, and wire human decisions before consequential actions. Execution records help teams examine what happened and test a changed workflow against representative inputs.
What to evaluate
Measure the complete workflow with your own data and providers. Model response time, API limits, document size, branching and review delays all affect completion time. Evaluate normal cases, missing inputs, conflicting records and service failures, alongside the runtime's behavior.
See the architecture reference, recorded product walkthrough, or discuss a workflow with us.