v0.7.0: The Monday Brew Update

Grab a fresh cup of coffee, because v0.7.0 introduces one of our biggest workflow upgrades yet!

With the new AI Prompt-Pack Generator, you now have the option to bridge your semantic models to external LLMs (like ChatGPT or Claude) to auto-generate Data Dictionary definitions and Business Logic.

🔒 Zero-Network Guarantee: This application remains strictly 100% OFFLINE. The app itself makes zero API calls, uses zero network connections, and sends zero data. The new AI features rely entirely on a manual "Bring Your Own LLM" copy-paste workflow, ensuring you maintain absolute control over your enterprise data.

Here is everything packed into this release:

✨ What's New

• AI Prompt-Pack Export & Import (Preview): Full disclosure: This feature is currently in its early stages. We've introduced this optional AI export/import loop to let you test the workflow and see if the look and feel pleases you. It's limited right now, but full functionality and deeper model integration will be rolling out in upcoming v0.7.x minor updates!
  • Export: Select "Both" or "AI Prompt-Pack (.md) only" to generate a prompt_pack.md file alongside your HTML. It contains a zero-shot LLM prompt with your model's exact TMDL payload.
  • Auto-Inject AI Definitions: Copy the JSON array generated by your AI directly into the new HTML feature! The HTML viewer now includes a sidebar button to instantly inject the AI-generated descriptions directly into the documentation UI, syncing tables, columns, and measures entirely locally in your browser memory.

🛠️ Bug Fixes & Reliability

• Universal Excel Parity: Fixed a KeyError: 'TableName' crash that prevented Excel (.xlsx) datasets from generating AI Prompt Packs. Excel is now 100% supported in the AI workflow.
• Fixed an issue where the PyInstaller executable would fail to bundle correctly.
• Resolved Pandas dependency issues that caused NameError crashes in certain environments.
• AI Prompt instructions have been heavily refined to prevent the LLM from hallucinating conversational text instead of raw JSON.

📥 Installation

No installation required! Just download PBI_Doc_Generator_v0.7.exe from the assets below and drop your files into it.

v0.6.1 (Hotfix & Excel DAX Expansion)

Release Notes: v0.6.1 (Hotfix & Excel DAX Expansion)

This update expands our analytical capabilities by unlocking full DAX measure extraction for Excel Power Pivot models, while also delivering critical UI/UX and performance optimizations to make the v0.6 release perfectly stable.

🌟 What's New

• Excel Power Pivot DAX Detection: The generator now seamlessly detects all Data Analysis Expressions (DAX) measures natively embedded within your Excel .xlsx Power Pivot Semantic Models. Note: While all measures and calculated columns are successfully inventoried, the extraction of the raw DAX expression code from the binary stream is currently a work in progress and will appear empty in the documentation boxes.
• Drastically Reduced Application Footprint: Through deep dependency auditing and migration to natively compiled xpress C-extensions, the total application executable size has been shrunk by ~40% (from 61 MB down to 36 MB) resulting in faster startup times and lower memory usage!

🛠️ Fixes & Improvements

• Foreground Matrix Splash Screen: Fixed an issue where the Matrix startup screen would occasionally spawn behind other active application windows; the loading visual is now strictly enforced to the topmost layer.
• Progress Bar Smoothing: Eliminated the jarring "0% to 100% jump" during extraction. The progress bar now employs realistic time-weighted stages (prioritizing the heaviest workloads like DAX Measure and Calculated Column extraction) combined with a sleek animation easing function for a buttery smooth visual readout.
• Subprocess Window Suppression: Resolved a visual defect where a black Command Prompt terminal window would abruptly flicker on-screen during the Node.js layout verification phase.
• PyInstaller Bundling Stability: Fixed a local PyInstaller cache exclusion rule that previously caused module failures on initial .exe execution.

v0.6.0

Release Notes: v0.6.0 (Major Update)

This update brings massive visual upgrades to the generated documentation, a brand new-style startup experience, and full offline crash safety.

🌟 What's New

• Enhanced Report Pages (SVG Wireframes): The generator now draws full, to-scale SVG wireframe layouts for every single page in your report. You can visually inspect overlapping elements, pop-up panels, and visual placements without ever opening Power BI!
  

• Data Fields Detection: The visual inventory tables now extract the exact data fields assigned to each visual (e.g. knowing exactly which measure is on the X-axis, Y-axis, or Legend).

• Matrix Startup Splash & Real-Time Progress: The generator now boots up with a Matrix terminal that displays the real-time progress of your extraction.

• Excel Power Query Extraction: The tool now fully supports dropping .xlsx files! It will cleanly extract and format all Power Query (M-code) scripts running behind your Excel sheets. (Note: DAX extraction for Excel models is currently a work in progress).

• Bulletproof Crash Safety: The extraction pipeline is now fully resilient. If an unexpected error occurs during parsing, the app will not crash but instead generate a safe fallback HTML document containing the error logs, so you can easily report the issue.

• Zero-Network Strict Mode: The application is fully portable and makes absolutely zero network connections. Safe for strict enterprise environments.

🛠️ Fixes & Improvements

• Fixed an issue with the layout parser not accounting for certain nested visual geometries.
• Updated UI to center the processing screen over the main application window instead of the primary monitor.

v0.5

Release v0.5 🚀 - Native DAX Translation & Visual Lineage

Thank you to everyone in the community for the incredible feedback! Version 0.5 introduces the two most highly requested features to make documenting your Power BI models easier and more transparent than ever before.

✨ What's New in v0.5

1. Native DAX-to-English Translation

Say goodbye to deciphering complex, nested DAX code. The tool now features a custom-built, shape-aware DAX parser!

• It automatically reads your VAR, RETURN, CALCULATE, and iterator functions (SUMX, FILTER, etc.).
• It translates the logic into plain, readable English bullet points.
• Extremely complex measures are broken down step-by-step so anyone (even non-technical stakeholders) can understand what a measure is calculating.

2. Visual-Level Lineage

Ever wondered "Is it safe to delete this measure?"

• The HTML documentation now actively scans all your report pages.
• It maps exactly which measures and columns are used in which visuals across your entire .pbix file.
• Usage context is injected directly under each field in the documentation.

3. Quality of Life Improvements

• Splash Screen: Added a sleek, instant loading screen to the .exe so you immediately know the tool is starting up while it unpacks in the background.
• UI Polish: Updated window titles and version tracking to ensure a smoother experience.
• Robustness: Fixed an edge-case bug where the NOT operator without parentheses would crash the DAX parser. The parser is now bulletproof against complex models.

💾 How to use

1. Download the PBI_Doc_Generator_v0.5.exe file below.
2. Double-click to run (no installation or Python environment required).
3. Select your .pbix or .pbip file and let the tool do the magic!

🔮 What's next? (v0.6)

We are actively investigating bringing the exact same Power Query (M-code) auto-documentation capabilities to Excel (.xlsx) files with Power Pivot Data Models. Stay tuned!

PBI Doc Generator v0.4

🆕 Report Pages & Visuals

The big one: documentation now includes a full visual inventory per report page.

• Each page lists its visuals as a structured, easy-to-read bullet list
• Every visual includes semantic data roles: X-axis, Y-axis, Legend, Values, Tooltips, etc.
• Tables and matrices now show Columns (N) with full field listings
• Slicers are summarized as a compact filter chip line per page
• Technical/helper elements (shapes, buttons, !-prefix cards, ServerString, etc.) are excluded from counts and grouped in a footnote

---

🔍 Smarter Observations

• Unused measures → detects measures defined but not used in visuals
• Duplicate DAX → flags identical measure expressions
• Hardcoded connections & excessive bidirectional filters (existing from v0.3)

---

🛠️ Improvements

• Full DAX & M expressions
  No truncation — all logic is shown completely in <pre><code> blocks

• Binary payload detection
  Inline base64 blobs replaced with: [binary payload omitted]

• No horizontal scroll
  Code, tables, and diagrams wrap neatly at any screen size

• Mermaid diagrams (offline)
  Relationship and lineage diagrams now render fully offline (mermaid bundled in the exe)

• .Report folder matching fix
  Fixed issue where multiple .pbip files in the same directory could mix report pages

• GUI enhancements
  Clickable LinkedIn and GitHub links in the footer

---

📦 Download

PBIDocGen_v0.4.exe

• Portable — no installation required
• Runs fully offline

---

✅ Supported Formats

Format | Support -- | -- .pbix | ✅ .pbip + PBIR | ✅ .pbip + PBIR-legacy | ✅ .SemanticModel (TMDL) | ✅

`

v0.3-beta - Offline Mermaid & Improved Data Sources

• Bundled Mermaid.js (now works 100% offline without internet connection).
• Categorized Data Sources exactly like Power BI's "Get Data" dialog (File, Database, Azure, etc.).
• Pinned pandas version for stability.

Full Changelog: v0.2-beta...v0.3-beta

v0.2-beta

New

• English UI — interface is now fully English.
• .pbip / TMDL support — select a Power BI Project (.pbip) or its .SemanticModel folder, in addition to .pbix. Full TMDL parsing including tables, Power Query/M, measures, calculated columns, relationships, and Calculation Groups!
• Measure Lineage — new section visualizing measure->measure and measure->column dependencies as a Mermaid graph, plus a dependency table.

Fixed

• Robust .pbip resolution (.SemanticModel / .Dataset / model.bim).
• Invalid build-backend in pyproject.toml that blocked pip/pipx installs.

Install

Run from source: pip install pbixray pandas then python gui.py,
or pipx install git+https://github.com/djrien-ai/pbi-doc-generator.git

(Or download the attached standalone executable below)