XrmToolSuite.SolutionComplexityScore 1.2026.7.12

📊 Solution Complexity Score

An XrmToolBox plugin that inventories a single solution's components and computes a weighted 0–100 complexity score — with a maintainability score, upgrade/migration/testing effort in person-days, a rough annual support-cost estimate, and a sibling build-quality score — from the component tallies via a UI-free, SDK-free scoring engine. Read-only (counts components; never modifies the solution).

Features

Exports

• PDF (native, MigraDoc/PdfSharp-GDI)
• HTML dashboard
• Excel workbook
• JSON
• Markdown
• An executive summary — offline-templated by default, with an opt-in AI path

Help & Support

A right-aligned Help button on the toolbar opens a Help & Support dialog (Documentation, Report an issue, and a support link, each opened in the browser). The tool implements IHelpPlugin and IGitHubPlugin pointing at repository kkora/XrmToolSuite.

Build & install

This tool is not single-DLL — it ships the Excel/PDF export dependency chain (ClosedXML + PdfSharp/MigraDoc-GDI) into the Plugins root next to the tool DLL. The one-step build copies the whole chain automatically:

dotnet build src\Tools\XrmToolSuite.SolutionComplexityScore\XrmToolSuite.SolutionComplexityScore.csproj -c Release -p:DeployToXTB=true

Restart XrmToolBox and open Solution Complexity Score. For a manual copy to another machine, copy every DLL from the tool's bin\Release\net48\ folder — flat in the Plugins root, never a subfolder, or XrmToolBox silently drops the tool. See ./DEPLOYMENT.md and the suite guide Deployment_Guide_XrmToolBox.md.

Usage

1. Connect to your Dataverse environment.
2. Open Solution Complexity Score and pick a solution (visible, non-system solutions load off-thread).
3. Run the inventory — collection runs on a background worker with progress.
4. Review the dashboard: complexity + maintainability + build-quality scores, effort/cost estimates, the metric strip and the hotspot grid.
5. Export to PDF / HTML / Excel / JSON / Markdown.

Notes & limitations

• Read-only — no destructive operations, so no confirmation dialog is required.
• The complexity/effort model, the ComponentCounts POCO and the report projection are UI-free and SDK-free; the collector degrades query failures to zero counts rather than throwing.
• The optional AI executive summary is opt-in behind a session-only API key (env var or the AI-settings dialog) that is never persisted, and a payload-preview consent dialog shows the anonymized JSON (no record data, credentials or environment names) before anything is sent. The chosen provider/model-id persist; the key does not.
• Local models (Ollama): run the summary against a local model — no API key, nothing leaves the machine.
  1. Install: winget install Ollama.Ollama (or ollama.com/download). Ollama serves http://localhost:11434 automatically; start it manually with ollama serve if needed.
  2. Pull a model: ollama pull qwen2.5:7b (also gemma3:4b, llama3.2:3b, qwen2.5-coder:7b). ollama list shows installed models, ollama ps shows loaded ones, ollama run <model> "hi" warms one up.
  3. AI settings ▸ Provider = Ollama (local), Model = qwen2.5:7b, leave API key blank. The first request loads the model (waits up to 5 min); if Ollama isn't running it falls back to the offline summary. nomic-embed-text is an embedding model — use a chat model (qwen2.5/gemma3/llama3.2).
• Deferred quality signals (naming-prefix consistency, description coverage, managed/unmanaged layering) need a collector change and are a phase-2 item.