feat: summarize changes with AI by AI (WIP)

[schmidtnz]

Motivation

We have large QGIS project files (XML files) that we want to version control.
When saving a project in QGIS, it does not serialize in a repeatable way. Some keys get ordered randomly.
This leads to large diffs with a lot of changes that do not influence the presentation of the file.
Also, some changes are insignificant, such as a changed viewport or different user saving the file.

This is an attempt to extract meaningful change from the XML and interpret it in human language using AI.

qgs files can be large and exceed the context of common models.

Modifications

• Python script to remove identifiable "churn" from qgs files and then diff these versions.
• Workflow to pass the reduced diff to AI for interpretation and posting to PR.

Verification

manual testing

[github-actions]

🗺️ QGIS project change summary

master-project/topo-editor-tiff50.qgs

• Project version updated: 4.0.0-Norrköping → 4.0.3-Norrköping.
• Layer display name changed: layer with source ../../topographic-data/topographic-data.gpkg|layername=geographic_name (providerKey="ogr", id="geographic_name_809a82b7_b97d_4932_829c_45e5bad435c7") had name "geographic_name" → "changed-name".

_{Generated by gpt-5-mini via GitHub Models from a churn-free diff - verify against the file diff.}