pbi-extractlisted
Install: claude install-skill starmynd-org/powerbi-to-ai-visuals
# Get a Power BI file open
A `.pbix` is a zip. The part that matters is `DataModel`: an XPress9-compressed
Analysis Services backup holding the VertiPaq column store. A `.pbip` (File >
Save as > Power BI project) is the same model written as text - TMDL for the
semantic model, PBIR JSON for the report - and it usually keeps the identical
binary at `*.SemanticModel/.pbi/cache.abf`.
So there are two things to pull, and they come from different places:
| what | best source | fallback |
|---|---|---|
| row data | `DataModel` in the .pbix, or `cache.abf` in the .pbip | none - without one of these there is no data |
| tables, measures, hierarchies, sort-by columns | TMDL in the .pbip | metadata inside the binary |
| pages, visuals, filters, theme | PBIR in the .pbip, or `Report/definition` or `Report/Layout` in the .pbix | none |
## Do this
```bash
python3 tools/pbi_extract.py "<file.pbix | file.pbip | PBIP folder>" -o build/model
python3 tools/pbir_parse.py "<same source>" -o build/model/report-spec.json
```
Then confirm what you got before building anything on it:
```bash
python3 tools/pbi_chat.py schema | head -60
python3 tools/pbi_chat.py measures
```
## What to check, and why
**Where the data actually lives.** The Power Query M in a partition usually
points at a path that does not exist on your machine - here, an Excel file on
another person's OneDrive. Do not try to refresh it. The rows you want are the
ones already imported into the model file.
**Auto date tables.