Zero-retention · One parse, four documents · AI-powered

One report,
four AI-written docs
in seconds.

Drop a .pbix or .pbip and PBICompass parses the semantic model once, then writes a Technical Documentation, an Audit & Health Report, an Executive Summary, and a Business User Guide — without your row-level data ever leaving the sandbox.

No data retained
4 documents, 1 parse
Claude Opus 4.8
.pbix & .pbip parsing DAX translation Data model narrative RLS & security Markdown · HTML · DOCX · PDF · JSON Audit & Health scoring Executive Summary Business User Guide .pbix & .pbip parsing DAX translation Data model narrative RLS & security Markdown · HTML · DOCX · PDF · JSON Audit & Health scoring Executive Summary Business User Guide
0
Document types
0
Export formats
0
Row data retained
0
Parse per report
Capabilities

Everything a BI team needs,
documented automatically

From one parsed semantic model, PBICompass writes four audience-specific documents — developer reference, governance audit, executive brief, and business user guide — so your reports never go undocumented again.

Zero-retention by design

Every upload is processed inside a per-job sandbox that is shredded the moment rendering finishes. Only metadata is read — never the VertiPaq row data — and nothing is logged or persisted.

finally → shred()

One parse, four documents

From the same parsed model, PBICompass writes a Technical Doc for developers, an Audit & Health Report for governance, an Executive Summary for leadership, and a Business User Guide for everyone else — pick one or generate all four.

1 parse → 4 docs

Five export formats

One run, every format you need: styled HTML, hand-written DOCX, clean Markdown, structured JSON, and PDF — no Word, no LaTeX, no external tooling required.

md · html · docx · pdf · json

Plain-English DAX

Multi-line measures are preserved and translated, with format strings, home tables and display folders intact — so anyone can read what a calculation actually does.

DIVIDE · CALCULATE · …

Full model coverage

Tables, columns, relationships, RLS roles, M / Power Query, data sources and report layout — the parser maps the common ~95% of real exports and never aborts on the odd construct.

TMDL · TMSL · PBIR

Audit & Health Report

A dedicated document with a 0–100 health score, best-practice checks, performance-risk heuristics, and prioritized recommendations — fully deterministic, so it works offline with no API key and never guesses.

health score · best practices · risks
Pipeline

From upload to handover
in four steps

parse → agents → render → download. The whole journey runs in an ephemeral sandbox.

1

Upload

Drop a .pbix or a zipped .pbip project. A zip-slip guard isolates every file inside its own per-job directory.

2

Parse

The stdlib-only parser extracts the semantic model and report layout into a canonical model.json — metadata only.

3

Generate

AI agents (or the offline engine) assemble whichever document types you chose — Technical, Audit, Executive, User Guide — from that one parsed model.

4

Download

Grab your docs in any format. The sandbox is shredded; only the rendered files survive, for a short TTL.

Live generator

Document your report now

Upload a file, choose which document(s) to generate, add optional governance metadata, and download enterprise-grade docs in seconds.

Secure sandbox · shredded after render

Power BI Knowledge Platform

Metadata-only · choose one document or generate all four

Click to upload or drag & drop

Accepts .pbix, or a .zip of a .pbip project folder

Files are processed in a secure sandbox. Uploaded metadata is shredded immediately after rendering.