PHAROS DATA ARK · 2026-10-29-01024f5b59

Developing and maintaining the Ark on the Ark

Inventory checked October 23, 2026. Source is /opt/dataark; the additional remote copy is DavJenkins/pharos. Use the admin Linux account for development, not root. This is the live working checkout; make Git checkpoints and review changes before deploying.

New terminal sessions provide a brief orientation. Run ark-help anywhere, or choose ark-help status, library, develop, recovery or basics. See shell welcome and administration guide.

Available tools

VS Code 1.143.1 ARM64: desktop application menu or code /opt/dataark.
IntelliJ IDEA 2026.2.3 ARM64: desktop application menu or idea. Installed under /opt/jetbrains with /opt/idea as the stable link. First launch may require license/setup choices; these are left to the user. Free core features and optional paid features have different licensing.
GHex 46.2: graphical binary/hex file editor, application menu or ghex. Work on a copy before editing an archive or system binary.
OpenJDK 21: java, javac, and standard JDK diagnostics. IntelliJ also has a separate bundled runtime.
Python 3.13, venv/pip, the existing /opt/dataark/venv environment and pinned requirements.txt; GCC, make and development headers.
Git and local history; SSH/SFTP, rsync, curl, archive utilities, Docker, systemd and Apache administration tools.
Chromium for manual local browser verification. No Node/npm, browser automation framework, or JavaScript lint tool was found in this audit; the current Ark serves ordinary JavaScript without a Node build step.
Codex CLI: official ARM64 standalone release, launched with codex from the source directory. The user must sign in separately. Do not copy another machine's authentication files or put credentials in the repository. Normal hosted-model use requires internet and an eligible account; installation does not provide an offline AI model.

Local workflow

Run the Python tests from a terminal on the Ark:


cd /opt/dataark
venv/bin/python -m unittest discover -s tests

They use Python unittest rather than pytest. Use the same virtual environment for application dependencies. Docs are rendered by tools/publish_docs.py without a separate Markdown toolchain. The existing deploy/deploy-web.sh validates/renders the dashboard and publishes the static site and documentation. Deployment and system service changes need the existing administrative privileges; editing code alone does not restart the API.

Use Chromium to check the actual landing page, library manager, maps, pop-ups and documentation. A passing unit suite is not proof that every reader, network route or recovery scenario works. Consult the acceptance tests and the source/handover guide before changing running services.

Memory and recovery limits

This Ark is a Raspberry Pi 4 Model B with 2 GB soldered RAM and about 2 GB swap. Its RAM is not an ordinary replaceable module. IntelliJ is below the vendor's published 8 GB total RAM requirement; installation is not a promise of usable performance. Keep heavy IDEs closed when not needed and avoid running both together. IDE graphical first-run and project indexing have not been acceptance-tested. No new resource quotas have been introduced.

Vendor archives, checksums and provenance for standalone tools are retained in /var/cache/dataark/ide. APT-installed packages come from the configured signed Debian/Raspberry Pi repositories. This cache is not a complete offline package mirror. Existing Python packages and Docker images support the current installation, but all wheels, OS packages, plugin dependencies and build artifacts needed for a fresh offline rebuild have not been collected and tested. Java/Maven/Gradle projects may need further dependencies.

The current installation can edit, test, document and deploy its source locally. A verified one-script clean-image bootstrap and a completely offline rebuild remain unfinished goals. See bootstrap plan.