dax-studio-export

Files (1)

• SKILL.md

SKILL.md (first lines)

# DAX export

1. The query starts with `EVALUATE`; wrap it in `TOPN(500, …)` while exploring. Pass it with `--query "<DAX>"`; write it to `.agent/dax/<KEY>-<purpose>.dax` and pass `--file` only when it is long or kept for a regression (in a fleet, a file edit needs the `write` grant the tile offers).
2. **Deployed model (the service):** `ad-pbi dax --workspace "<pbi_workspace>" --model "<pbi_model>" --query "<DAX>" --out .agent/out/<KEY>-<purpose>.tsv` (or `--file <path>.dax`)
   It renders TOON and keeps the file. Sign-in is the command's: Tabular Editor gets an az access token per launch, and a signed-out Azure CLI is signed in by the command (`az login --allow-no-subscriptions`, a browser window — say so in one line). Never open DAX Studio or Tabular Editor by hand to seed a sign-in. `not_signed_in` after that → `ad-pbi auth --probe` once, then `friction-log` type `tool-error`.
3. **Running Desktop (`localhost:<port>` from `ad-pbip desktop`, no sign-in):** `ad-pbip dax --server localhost:<port> --query "<DAX>" --out .agent/out/<KEY>-<purpose>.tsv` (or `--file`). It runs dscmd for you and renders TOON; never run dscmd or `python` yourself for a query. A `.csv` someone else exported: `ad-view <file>.csv`.
4. Regression: export the same query before and after a change (`--out …-before.tsv` / `…-after.tsv`) → `ad-diff before.tsv after.tsv --key <row key>`.
5. Model metrics — pwsh: `& "<dscmd_exe>" vpax ".agent/out/<KEY>.vpax" -s "<server>" -d "<pbi_model>"`
   bash: `"<dscmd_exe>" vpax ".agent/out/<KEY>.vpax" -s "<server>" -d "<pbi_model>"`. Report file size only; do not open. Against the service, `vpax` uses DAX Studio's own sign-in (open DAX Studio once and connect to the workspace, then re-run); against Desktop `localhost:<port>` it needs none.
6. `state-update` with paths. Return to caller.

From agentchieflou/this-next-please, MIT.

More from agentchieflou/this-next-please