Git is excellent for text and selected exported artifacts, but many PLC and HMI projects also contain proprietary binaries, generated files and online state. A safe workflow defines what can be diffed, what must be packaged, and which engineering evidence Git alone cannot provide.
The short answer
Start with vendor support and a tested export/import path. Store deterministic text sources, configuration, scripts and documentation in Git; exclude caches, credentials and machine-specific files. For proprietary binary projects, retain controlled native packages and checksums beside an export that reviewers can inspect. Use short-lived branches, meaningful commits, peer review, release tags and a reproducible manifest. Prove restoration in an isolated engineering environment before relying on the repository for recovery.
- Version only artifacts whose source, ownership and restoration route are understood.
- A commit hash is not a substitute for a deployable native backup.
- Reviewable exports and native packages serve different purposes.
- Every release baseline is tied to software, hardware, tools and test evidence.
Decision guide
| Artifact | Best control | Reason |
|---|---|---|
| Structured text/config | Git repository | Meaningful diff, review and merge |
| Vendor binary project | Immutable package plus checksum | Often cannot merge or inspect reliably |
| Generated/cache files | Regenerate or exclude | Noise and machine dependence |
| Release baseline | Tag plus manifest and native package | Reproducible handover and recovery |
A practical workflow
- 01
Inventory authoring tools, file types, export/import support and secrets.
- 02
Define repository layout, ignore rules, naming, access and review policy.
- 03
Create a clean baseline with native package, inspectable exports and manifest.
- 04
Use branches, peer review and deterministic checks for each change.
- 05
Tag approved releases and run restore plus acceptance tests independently.
Evidence and acceptance
Keep repository URL and access owner, tool and version, target hardware and firmware, source and export paths, ignore policy, commit and review record, native project checksum, dependency manifest, build/export instructions, release tag, signed approval, restore environment, restoration result, test outcome, exceptions and rollback package.
Common mistakes
- Putting passwords, certificates or private endpoints in the repository.
- Assuming proprietary binary projects can be merged safely.
- Committing generated files that change on every save.
- Tagging a release without its native package and toolchain.
- Calling a successful clone a tested recovery.
Responsibility boundary
Primary sources
Technical claims and product boundaries were checked against these primary sources. Fact-checked
Frequently asked questions
Can Git store a PLC project file?
Yes as a binary, but useful diff and merge may be unavailable; keep an inspectable export and tested native package too.
Should online uploads be committed?
Only through an approved, safe and documented acquisition process with provenance and no exposed secrets.
Does a Git tag prove what ran in the controller?
No. Deployment and as-found evidence are separate records that must reference the intended release.
Does AutoNexit replace a Git server?
No. It connects repository and package identities to project decisions, releases and evidence.