Open source walk-forward analysis and quant trading tools open source searches often land on repositories before products. The Kiploks engine is published under Apache 2.0 so you can audit math, reproduce metrics, and pin versions (Apache 2.0).

This page is a getting started orientation, not a substitute for the canonical README in the repo.

Why run locally

Verify definitions match what you see in hosted reports
Integrate analysis into your CI or research pipelines
Debug discrepancies between an export and a metric printout
Build internal tooling without depending on a UI session

Prerequisites

A supported Node.js version (see engine repository)
Deterministic inputs: same data revision, same bar rules, same package versions
A habit of pinning semver: engine packages move; silent upgrades break audits

Minimal path

Clone the engine repository (https://github.com/kiploks/engine).
Install dependencies per package instructions in the repo.
Run the package build steps if required by that revision.
Follow analyze examples in open engine docs.

Reproducibility checklist (short)

commit hash recorded
package-lock.json (or equivalent) stored
dataset file hash recorded
config JSON stored next to outputs

If you cannot reproduce a number a month later, you do not have an audit trail.

What local engine is not

Local engine access does not automatically grant hosted product features (accounts, storage, collaboration). Those are separate (Engine overview).

Common failure modes

mixing browser bundling with server-only imports
running analysis on non-deterministic data pulls without snapshotting
comparing outputs across two engine versions without noticing

Use a small golden fixture

Pick one tiny dataset and one known config as a regression fixture.

When engine upgrades happen, diff metrics on the fixture before you trust large batch reruns.

Docker can help, but volumes need discipline

Mounting a live-synced data folder can introduce silent file changes between runs.

Prefer immutable snapshot files for anything you cite as evidence.

Environment parity with CI

Run the same Node version locally and in CI. Version skew is a classic source of "works on my machine" metric drift.