Skip to content
GMST --h --m --s FR / EN
Logo of the Distributed WBPP
Beta Preprocessing

Distributed WBPP

Run PixInsight's WBPP across several PCs on your local network, transparently.

Download ↗ Source code (GitHub) ↗ Documentation ↗

Overview

Distributed WBPP runs PixInsight's WBPP (Weighted Batch Preprocessing) across several PCs on a local network, transparently.

Everything is configured in the native WBPP UI on the main machine; on the others, you simply start the script in client mode. The cluster auto-discovers available machines and moves frames over the network, with no shared folder required.

The heavy steps are spread across the machines: light calibration, measurements, registration, local normalization and its references, and every integration (calibration masters, final per-filter light and drizzle integrations).

A strict version handshake, identical PixInsight and WBPP on every node, and an automatic local fallback guarantee a result identical to a local run, validated bit for bit.

Features

Native WBPP UI, unchanged

Configure your sessions as usual in the native WBPP dialog: distribution grafts onto it, nothing changes in your habits. The only addition is starting the same script in client mode on the other PCs.

LAN auto-discovery

A lightweight companion (single executable, no runtime to install) discovers machines on the network and transfers frames with SHA-256 verification. No shared folder required.

Adaptive per-frame split

Light calibration, measurements, registration and local normalization are split according to each machine's measured cost, transfer included.

Parallel integrations

Bias, dark and flat master integrations, final per-filter light and drizzle integrations, and LN reference generation are each leased to a free machine while the server processes its own share locally.

Safe by design

Strict version handshake between nodes and automatic local fallback on any hiccup: never a silently wrong result. Validated bit-identically against a fully local run.

Live monitoring

A dashboard lists clients and their live activity, and WBPP's Execution Monitor tags every distributed step (frames processed locally / on the cluster).

Cross-platform, mixed clusters

Validated end-to-end on Windows, macOS and Linux, including OS-mixed clusters. Results are bit-identical across same-OS nodes and numerically equivalent in mixed runs (measured mean diff < 1e-8).

Screenshots

Distributed WBPP launcher: pick this PC's role, Server (drives WBPP) or Client (helps the server)
WBPP Execution Monitor during a distributed run: each heavy step is tagged [cluster: N local + M client]
Distributed WBPP Client window: waiting for work from the server, event log

Install

  1. 01

    Add the CaeloWorks repository

    In PixInsight, open Resources → Updates → Manage Repositories and add the CaeloWorks repository URL:

  2. 02

    Check for updates

    Run Resources → Updates → Check for Updates and accept: PixInsight downloads and installs the script.

  3. 03

    Restart PixInsight

    Relaunch the application: the script shows up in the Script menu. Future updates then arrive automatically.

Once installed, the script shows up under Script → CaeloWorks → Distributed WBPP: Server mode on the main machine (the native WBPP dialog opens), Client mode on the helper PCs. Every node must run the same PixInsight (1.9.4+) and WBPP versions. The repository still serves v1.0.0 to PixInsight 1.9.0 through 1.9.3.

Changelog

  • 1.8.0 2026-07-11

    The script now has its icon in PixInsight and a CaeloWorks header on all three windows (role picker, server dashboard, client log). It leaves Batch Processing for its own category: Script → CaeloWorks. Fixed: the menu icon never showed, because the package installed it in a folder reserved for the scripts bundled with PixInsight, and the feature identifier carried no name.

  • 1.7.0 2026-07-04

    macOS validated end-to-end (Sonoma, Intel): all three platforms are now validated, mixed-OS clusters included.

  • 1.6.0 2026-07-04

    Optimisations: drizzle jobs upload about 50% less data (registered frames are no longer shipped), and worker-produced drizzle data is path-fixed at collection so cached native re-runs stay correct. Revalidated end-to-end, 21 of 21 masters pixel-identical.

  • 1.5.0 2026-07-04

    Linux validated end-to-end (Ubuntu 24.04, PixInsight 1.9.4 / WBPP 3.0.1), including a mixed Windows + Linux cluster: every distributed operation collected from the Linux node, clean logs on both sides.

  • 1.4.0 2026-07-04

    Per-filter drizzle integrations are now distributed as whole jobs too. Masters validated pixel-identical against a fully local baseline on a real 2-node cluster.

  • 1.3.0 2026-07-04

    Local normalization (LN) reference generation is now distributed as well (one whole job per filter).

  • 1.2.0 2026-07-04

    Final per-filter light integrations are now distributed as well.

  • 1.1.0 2026-07-04

    PixInsight 1.9.4 / WBPP 3.0.1 compatibility (new WBPP 3.0 layout). Now requires PixInsight 1.9.4+.

  • 1.0.0 2026-07-02

    First public release (beta, GPL-3.0), distributed through the CaeloWorks update repository. Validated end-to-end on PixInsight 1.9.3 / WBPP 2.9.1. Still served by the repository to PixInsight 1.9.0 through 1.9.3.