powerbi-documentation

Files (3)

• LICENSE
• SKILL.md
• local.patch

SKILL.md (first lines)

# Power BI Documentation Skill

Generate comprehensive, standardized documentation for Power BI semantic models using an interactive workflow.

## Prerequisites

Before starting documentation, ensure:
1. The powerbi-modeling-mcp server is connected and available
2. Connection to the target semantic model is established (Power BI Desktop, Fabric workspace, or PBIP folder)

## Interactive Documentation Workflow

**CRITICAL:** This skill uses a conversational approach. The assistant MUST ask questions one at a time and wait for user responses before proceeding. Do NOT generate the document until all questions are answered. All questions use numbered options (1/2/3/...) for consistency — never use Yes/No style answers.

### Phase 1: Connect and Gather Metadata

Before asking any questions, the assistant should:

1. **Connect to the semantic model** using `connection_operations`
2. **Gather basic model information** using `model_operations`:
   - Model name
   - Table count (and table names for classification)
   - Measure count (including display folders)
   - Relationship count
   - File name (for document title)
   - Last modified date
3. **Check for calculation groups** using `calculation_group_operations` (ListGroups)
4. **Classify tables** by analyzing naming conventions and structure:
   - Fact tables (e.g., `fact_*`, `fct_*`, or tables with numeric measures and foreign keys)
   - Dimension tables (e.g., `dim_*`, `d_*`, or lookup/reference tables)
   - Parameter tables (e.g., `para_*`, `param_*`, or tables with What-If parameters)
   - Bridge/mapping tables
   - Date tables
   - Calculated tables
   - Other/Utility tables

This information is needed to ask informed questions and provide sensible defaults.

### Phase 2: Interactive Configuration

Start the interactive configuration with welcoming the user to the 'The Data Forest's interactive Power BI documentation generator'.
Then ask these questions **one at a time**, waiting for user response before proceeding to the next.

#### Pre-Question: Additional Business Context
```
Before we start configuring the documentation, would you like to upload any additional business context?

This can be any kind of documentation — business requirements, data dictionaries,
stakeholder briefs, existing wiki pages, design documents, or anything else that
describes the purpose and context of this model.

Providing this context helps me generate richer, more accurate descriptions,
executive summaries, and recommendations tailored to your actual business needs.

1. Yes, I'd like to upload some documents (I'll attach them now)
2. No, let's proceed without additional context

[1/2]:
```

**If the user selects 1:** Wait for the user to upload their file(s). Once received, read and analyze the uploaded documents from `/mnt/user-data/uploads/`. Extract key information such as:
- Business domain and purpose of the model
- Key stakeholders and their reporting needs
- Important business processes and KPIs
- Data source descriptions and ownership
- Any naming conventions or terminology definitions
- Known limitations or scope boundaries

Store this extracted context mentally and use it throughout the documentation generation to:
- **Executive Summary:** Incorporate actual business purpose, stakeholders, and scope rather than inferring from model structure alone
- **Table & Measure Descriptions:** Use domain-specific language and business terminology from the uploaded context
- **Recommendations:** Tailor suggestions to the actual business requirements and stakeholder needs
- **Business Glossary:** Pre-populate with terms and definitions found in the uploaded documents
- **Data Sources:** Enrich with any data lineage or ownership information from the context

After reading the documents, briefly confirm what context was extracted: "Thanks! I've reviewed your documents and found context about [brief summary]. I'll use this to enrich the documentation. Let's continue with the configuration."

**If the user selects 2:** Proceed normally. The documentation will be generated based solely on the semantic model metadata.

#### Question 0: Documentation Template
```
How would you like to set up the documentation?

I can offer you a preset template or let you configure each section manually:

1. Internal documentation preset (comprehensive, includes all technical details, recommendations, and DAX code)
2. Client documentation preset (polished, business-focused with executive summary, glossary, and clean formatting — less technical detail)
3. Manual configuration (I'll walk you through each section one by one)

[1/2/3]:
```

**If the user selects 1 (Internal) or 2 (Client):** Show the preset configuration summary (see Preset Templates section below) and proceed to Phase 3 after confirming. Still ask the Title question (Question 2) and Language question (Question 1) before showing the preset summary. If the user wants to customize any part of the preset, allow them to override individual settings.

**If the user selects 3 (Manual):** Continue with all questions below.

#### Question 1: Documentation Language
```
What language should the documentation be written in?

1. English
2. Dutch (Nederlands)
3. French (Français)
4. German (Deutsch)
5. Other (please specify)

[1/2/3/4/5]:
```

The selected language applies to all generated prose, section headers, descriptions, and analysis. Table/column/measure names remain as defined in the model. DAX code is never translated.

#### Question 2: Document Title
```
What should be the title of the documentation?

I detected the model name is "[Model Name]" from [File Name].
Default: "Power BI Semantic Model Documentation: [Model Name]"

1. Use default title

From Kilo-Org/kilo-marketplace, Apache-2.0.

More from Kilo-Org/kilo-marketplace