Lineage · Audit · Wiki-ready Markdown

Open a PBIP project folder — get a searchable dashboard plus nine Markdown docs ready for ADO Wiki or GitHub. Runs entirely in your browser (nothing uploads) or as a local CLI.

• Interactive dashboard — 11 tabs: Sources, Tables, Columns, Relationships, Measures, Calc Groups, Functions, Pages (with layout wireframe), Unused, Lineage (search-driven), Documentation
• Wiki-ready Markdown — up to 9 generated docs, paste-ready for ADO Wiki or GitHub. Lite / Detailed toggle per doc — paste-into-wiki summary for stakeholders, full reference for engineers. Cross-doc links between them resolve in the dashboard, on GitHub, and on ADO Wiki
• Improvements audit — 16 checks across five severity tiers (high / medium / low / info / strengths) including broken-DAX-reference detection
• Source Map — flat PBI-column → physical-source lineage with CSV export
• Page layout wireframe — SVG thumbnail of each page showing visuals at true canvas positions
• Privacy by default — browser-mode files never leave your machine; CLI binds to loopback only

Measures — DAX dependencies, where-used per visual + page, direct/indirect/unused status, descriptions inline. CSV export honours filters.	Columns — types, usage counts, status, sortable + searchable. Same CSV export pattern.
Tables — grouped by role (Fact / Dimension / Bridge / Calc Group). Per-table stats for columns, measures, keys, FKs.	Tables drill-down — expand any table to see columns, measures, and per-relationship cardinality chips (*:1, ↔ for bidi).

Relationships — active + inactive, cardinality (one/many on each side), cross-filter direction (single → / both ↔). CSV export.	Sources — connection buckets, partition modes, physical-source index (what breaks if this source goes away?), hand-written SQL from Value.NativeQuery, M-step breakdown classifying every ETL step, raw M expressions on demand.
Lineage — search any measure or column from the tab itself, or click an entity from any other tab. Shows upstream DAX dependencies + source tables + functions, and downstream visuals + measures that feed into it.	Unused — orphan measures, dead-chain measures, indirect-use detection. Pairs with the Improvements audit's broken-reference detection.

Pages overview — every page in the report with visual count, slicers, measure / column bindings.	SVG wireframe — each page's visuals at true canvas positions. Charts, cards, slicers, tables, maps — shapes and buttons deliberately filtered.
Per-visual bindings — every measure + column each visual consumes, with type chips. Hover the wireframe for field-well tooltips.	UDF reference — parameters, body, and every measure that calls each function. Tabular 1702+.
Calc Groups — items, precedence, format-string expressions per item.

Nine Markdown documents — paste-ready for ADO Wiki or GitHub, anchor-stable, with cross-doc links that resolve everywhere.

Plus a Changelog tab exposing the project's release history.

→ Open the browser build — requires Chrome, Edge, or Opera (File System Access API).

1. Click Open folder — pick the PBIP project parent folder (the one that contains both <Name>.Report and <Name>.SemanticModel). Or click Try a sample for a one-click demo.
2. When multiple project pairs exist under one parent, a pair-picker lets you choose which Report + Semantic Model to load. Report-only and Model-only modes are also supported.
3. Dashboard renders in the same page.

Firefox / Safari users: run the CLI (see below) — the File System Access API is Chromium-only.

Requires Node.js 18+ and a .Report folder with its .SemanticModel sibling.

Double-click (Windows):

launch.bat

First run does npm install + npm run build, then starts the app and opens your browser.

From the terminal:

npm install
npm run build
node dist/app.js

The app listens on http://127.0.0.1:5679 (loopback only — nothing leaves your machine). Paste the .Report path or use the picker. Recent reports are remembered.

The same parser + data-builder that power the CLI run client-side via three small shims:

• src/browser/fs-shim.ts — pretends to be fs but reads from an in-memory Map
• src/browser/path-shim.ts — POSIX-style path replacement
• src/browser/fsa-walk.ts — walks a FileSystemDirectoryHandle into the Map

An import-map in docs/index.html redirects bare import … from "fs" / "path" to the shims. The parser never knows the difference.

npm run build:browser    # → ./docs/
npm run serve:browser    # → 127.0.0.1:5700

.github/workflows/pages.yml auto-publishes docs/ to GitHub Pages on every push to main.

src/
  pbir-reader.ts       Read-only access to PBIR report/page/visual JSON
  model-parser.ts      findSemanticModelPath + TMDL + BIM parsers
  report-scanner.ts    Walks visuals/filters/objects to extract field bindings + positions
  data-builder.ts      Cross-references model + report into FullData
  md-generator.ts      9 MD docs (Lite + Detailed modes), ADO Wiki-safe anchors, cross-doc xref links
  improvements.ts      16-check audit (incl. broken-ref detection) — severity-tiered
  html-generator.ts    Dashboard HTML template
  client/              Dashboard runtime (tabs, search, sort, lineage view, wireframe)
                         main.ts · render/escape.ts · render/md.ts · globals.d.ts
  render/              Shared server-side escape helpers
  app.ts               CLI HTTP server + landing page
  browser/             FSA walker + shims + entry shell + pair-picker logic
                         entry.ts · fsa-walk.ts · fs-shim.ts · path-shim.ts · pair-picker.ts

scripts/
  build-browser.mjs    Assembles docs/ from dist/ + browser TS output
  bake-sample.mjs      Bakes a sample PBIP into docs/sample-data.json
  serve-browser.mjs    Tiny static server for local testing

changelog/             Per-version release notes (one file per release)
tests/                 node:test suites — 311 tests, zero framework deps

Runtime deps: none. Only Node builtins (fs, path, http, crypto, child_process). Dev-deps are typescript + @types/node only — needed to build.

npm run typecheck    # tsc --noEmit
npm test             # compile + run Node's built-in test runner
npm run build        # compile CLI to dist/
npm run build:browser  # compile + assemble docs/ + bake sample

Tests use the stdlib node:test module. .github/workflows/ci.yml runs typecheck + tests + build on every push and PR across Node 18 / 20 / 22.

Each of the nine generated MDs starts with an HTML comment suggesting its wiki page name:

<!-- Suggested ADO Wiki page name: Health_and_Safety/Measures -->
# Measures Reference

To publish:

1. Open the dashboard for your report, switch to the Documentation tab
2. For each doc (Model / Data Dictionary / Sources / Measures / Functions / Calc Groups / Pages / Improvements / Index):
   • Click ⎘ Copy to copy the markdown
   • In ADO Wiki, create a new page with the name from the <!-- Suggested ADO Wiki page name: ... --> hint
   • Paste
3. If you want ADO's auto-TOC instead of the hand-rolled one, type [[_TOC_]] at the top of the page

Every anchor link is verified by tests/md-anchors.test.ts — drift fires CI.

changelog/ — one file per release. Full index in changelog/README.md. Future direction is open-ended; feature requests welcome via Issues.

Latest release: v0.11.7 — Default tab on first load is now Sources (was Measures), and Measures + Columns sort no longer flips to ascending on report reload. Small UX defaults that were quietly wrong.