Compares what Path of Building and a LocalExile checkout calculate for one saved build. It is a developer tool for finding stats the server gets wrong or does not implement, and it is not part of the portable application.
.\scripts\build-exporter.ps1
python scripts/parity.py --localexile C:\LocalExile --build C:\Builds\Example.xmlIt needs the .NET 10 SDK, Python 3.10+ and a LocalExile checkout with its game data extracted.
A run takes about 10 seconds, plus about 20 for the first build of the server probe. --no-build reuses the probe.
--show-skipped lists the modifier lines the exporter could not write, and --json <file> saves the result.
No client, Gateway or real save is involved, and it does not change your LocalExile checkout. It is safe while you play.
PoB runs headless through the bundled runtime. The build is exported to a temporary database, and
tools/LocalExile.Parity runs the checkout's own login path on it (PrepareCharacterForArea, then PoolsFor).
That project builds against your checkout under the assembly name the Gateway's InternalsVisibleTo already trusts,
so it can break when those internals change.
Both sides must describe one character, so PoB is first reduced to what the server received:
| Left out of PoB | Why |
|---|---|
| Passive jewels, ascendancy and cluster nodes | The exporter does not write them. PoB uses the saved passive list. |
| Tattoos the exporter skipped | The server never received them; the passive keeps its ordinary effect. Supported tattoos are written and stay on both sides. |
| Skills and auras | A fresh server character has none active. |
| Active flasks, bandit and Pantheon | A fresh server character has none of them. |
| The Kitava resistance penalties | They need quest flags the character does not have. |
| Modifier lines the exporter could not write | The server never received them. Lines match by base item and wording. |
| Items the server made inactive | They are listed as one finding, so the table shows formulas only. |
Reading a line in the table:
- Blank status: both sides agree.
- rounding: they differ by 1.
- mismatch: a real difference. Check it before calling it a bug: PoB is the project's formula reference, not proof of the official game.
- missing: one side did not report this stat.
A low attribute can make the server switch off an item, usually through a socketed gem whose requirement is not met. That happens more often here than in game, because the attributes from jewels and skipped modifiers are gone.
When armour, evasion or energy shield differ on one item, the report names that item. That usually means the exporter
chose the wrong modifier, so compare the item's text with what was saved before suspecting the server.
For anything else, list PoB's sources for that stat with modDB:Tabulate and compare them with the server's total.
A missing percentage often points straight at one passive or mastery.
The check covers the character sheet: attributes, life, mana, energy shield, ward, armour, evasion, accuracy, resistances, block and spell suppression. Skill damage, speed and critical strikes are not compared yet. Builds that rely on modifiers the exporter skips test less of the server, because those lines leave both sides.