Build Process Watcher dashboard preview
Action snippets updated for v0.7.0

Build Process Watcher

Capture Gradle, Kotlin, and JVM build-process metrics in Local Mode, watch a Remote Mode run live, then replay or compare the exported data when you need to explain what changed.

Open Demo Dashboard View Repository
What is new

More context at the end of every build.

The current release makes the monitoring choices explicit while keeping the JSON and dashboard workflows compatible with older runs.

Richer run dashboards

Remote runs show memory, heap, GC, JIT, class-loading, process metadata, and VM flags when the JVM exposes them. Hosted reports are retained for 24 hours.

See the mode options →

Replay and Compare

Local JSON artifacts can be opened offline in Replay or lined up against another build in Compare. Remote Mode is not required for either workflow.

Explore the analysis tools →

Experimental prediction

Opt-in predictive reliability checkpoints are available as advisory signals for triage. They are disabled by default and may change while experimental.

Read the safeguards →
Collection modes

Choose where a run should live.

Every snippet uses the same action. Local Mode keeps outputs in GitHub Actions artifacts; Remote Mode adds the live hosted dashboard. Every snippet is pinned to v0.7.0.

Local Mode · artifacts only

Generates logs, JSON, CSV, SVG charts, and the Actions summary in the workflow. Download the JSON later for Replay or Compare; nothing is sent to the hosted dashboard.

- name: Build Process Watcher
  uses: cdsap/build-process-watcher@v0.7.0

- name: Build project
  run: ./gradlew clean build

Remote Mode · live dashboard

Streams samples to the hosted dashboard, prints a dashboard URL, and retains the hosted report for 24 hours. The action can still produce local artifacts when available.

- name: Build Process Watcher
  uses: cdsap/build-process-watcher@v0.7.0
  with:
    remote_monitoring: 'true'

- name: Build project
  run: ./gradlew clean build

BigQuery export · optional

Not another monitoring mode. It is an independent export choice that requires Remote Mode and asks the backend to store finished-run metrics in your configured dataset.

- name: Build Process Watcher
  uses: cdsap/build-process-watcher@v0.7.0
  with:
    remote_monitoring: 'true'
    export_to_bigquery: 'true'

- name: Build project
  run: ./gradlew clean build
Analysis tools

The action inputs work together in predictable ways.

Replay

Input: one exported build_process_watcher.json. Needs: no Remote Mode. Scrub metrics offline in the browser.

Open Replay →

Compare

Input: two exported JSON files. Needs: no Remote Mode. Compare processes and timelines side by side.

Open Compare →

Prediction Experimental

Input: predictive_reliability: 'true'. Opt in to advisory predictive reliability checkpoints; this implies Remote Mode and requests BigQuery export. It is disabled by default.

Read the public contract →
Experimental prediction: prediction is an opt-in advisory signal for triage, not a guaranteed OOM, timeout, or build-failure outcome. Availability and behavior may change while experimental. Do not use it as a production decision or reliability guarantee, and do not infer private model details from the public dashboard.
Workflow

Built for comparing real CI behavior.

The UI now centers on the three analysis loops the action supports: watch a run, replay an exported run, and compare two builds side by side.

01

Run dashboard

Track peak RSS, heap, GC time, process metadata, VM flags, and live refresh status.

02

Replay JSON

Upload an artifact JSON file and scrub through memory and GC timelines offline.

03

Compare runs

Overlay two exported runs with a shared elapsed timeline and per-process colors.

04

Optional BigQuery export

Collect finished-run metrics in BigQuery when `export_to_bigquery` is enabled.

Screens

The data stays inspectable after the job finishes.

Replay JSON run demo

Replay exported runs

Use workflow artifacts without enabling remote mode.

Compare JSON runs demo

Compare two builds

Spot changes in memory pressure, GC, and process configuration.

Compare JSON Runs Replay JSON Run
Inputs

Configuration that matters most.

These inputs are independent unless noted. See the README input contract for backend deployment details.

Input Description Default Required
remote_monitoring Enable cloud dashboard ingestion. false No
export_to_bigquery Optional finished-run metrics export; requires remote_monitoring: 'true'. false No
predictive_reliability Experimental, opt-in advisory signals; implies Remote Mode and requests BigQuery export. false No
interval Polling interval in seconds. 5 No
disable_summary_output Suppress the GitHub Actions job summary. false No

Open a real dashboard.

The demo uses recorded build data so you can inspect the run UI without waiting for a workflow.

View Example