MetricsReporter

Three data sources. One dashboard. Measurable improvement.

---

MetricsReporter is a .NET 8 CLI tool that aggregates code coverage, complexity, coupling, and analyzer violations from three independent sources into one interactive dashboard — then lets you (or your AI agent) fix everything via a structured refactoring loop.

  coverage.xml  +  metrics.xml  +  violations.sarif   →   one interactive HTML
  (OpenCover)      (Roslyn)        (Analyzers)              + unified JSON

Production-Ready Example

MetricsReporter is now also demonstrated on baidakovil/eShop, a production-ready reference application forked from dotnet/eShop. That repository shows the full end-to-end setup on a non-trivial external solution: Roslyn metrics, SARIF diagnostics, OpenCover coverage, ReportGenerator output, AI-assisted refactoring, and a published HTML dashboard. Explore the example repository or open the interactive demo for eShop.

Dashboard Output

_{Production-ready example on Ubuntu in GitHub Codespaces, viewed from macOS}
▶ Open Production-Ready eShop Demo

_{MetricsReporter is OS-independent: Dashboard Output on Windows ARM64}
▶ Open Live Interactive Demo

The Problem

Your C# project has growing tech debt, but:

• Coverage, metrics, and violations live in three separate files — OpenCover XML, Roslyn XML, and SARIF JSON
• No single view shows coupling, complexity, coverage, and analyzer violations together
• You can't measure whether a refactoring actually helped
• AI agents don't know which method to fix first or whether the fix worked

Quick Start

# Install
dotnet tool install --global MetricsReporter.Tool

# Generate dashboard from your three data sources
metricsreporter generate --opencover coverage.xml --roslyn metrics.xml --sarif analyzers.sarif --output-html report.html

# Query violations from CLI — returns JSON
metricsreporter read --namespace MyApp.Services --metric Coupling
# → [{"kind":"Type","fullyQualifiedName":"MyApp.Services.OrderService","metrics":{"RoslynClassCoupling":{"value":14,"status":"Warning"}}}]

# Verify a fix passes thresholds
metricsreporter test --symbol MyApp.Services.OrderService.Process --metric Complexity
# → {"isOk":true}

Open report.html — you'll see the dashboard from the screenshot above.

Next step → Full tutorial: produce your first dashboard · CLI reference · Configuration reference

Key Features

Interactive HTML Dashboard — Drill down from Solution → Assembly → Namespace → Type → Method. Filter instantly, toggle warning/error awareness, hover for metric details. Self-contained HTML, no server, handles 50k+ symbols. → details

AI-Driven Refactoring — Give any AI agent (Copilot, Cursor, Claude) a namespace and a metric. It reads the violation, edits the code, rebuilds, and verifies — autonomously. Ready-to-use prompt files included. → workflow

Threshold Gates for CI — Define warning/error thresholds per metric per symbol level. CLI exits with 0 (pass) or non-zero (fail) — plug it straight into your pipeline. → details

Baseline & Delta Tracking — Every run saves a baseline. Next run computes deltas automatically, per method — complexity, coverage, violations. → details

Suppression System — Mark intentional exceptions with [SuppressMessage]. They show up in the dashboard with justifications, not as false alarms. → details

ReportGenerator Integration — Link to line-by-line HTML coverage maps. Script hooks trigger full rebuild + coverage recollection as part of the AI verify step. → details

_{Rich interactive UI — adjustable thresholds, delta tracking, suppressions, tooltips, ReportGenerator deep links, and more}

AI Agent Workflow

MetricsReporter ships with ready-to-use prompt files for Copilot, Cursor, or any AI agent:

# 1. Agent asks: "what's broken?"
metricsreporter read --namespace MyApp.Services --metric Coupling

# 2. Agent reads the source, edits files, runs `dotnet build`

# 3. Agent verifies the fix:
metricsreporter test --symbol MyApp.Services.OrderProcessor --metric Coupling
# → { "isOk": true }  ✅

# 4. Repeat until clean

Prompt file	What the agent does
refactor-complexity.md	Reduce cyclomatic complexity below thresholds
refactor-coupling.md	Reduce class coupling with DI, interfaces, DTOs
refactor-coverage.md	Write tests until branch coverage passes
refactor-sarif.md	Fix CA/IDE analyzer violations

Technical Highlights

• Cross-format symbol resolution — cross-links symbols across three unrelated XML/JSON formats by fully-qualified name, handling compiler-generated state machines and namespace mismatches (details)
• Clean architecture — command handlers, dependency injection, interface segregation, NUnit + NSubstitute test suite
• Zero-dependency HTML — renders 50k+ symbols with vanilla JS, no frameworks, no CDN — one self-contained file
• Three-layer configuration — CLI flags → env vars → JSON config → defaults, with JSON schema validation (reference)
• Structured CLI — four commands (generate, read, readsarif, test) with JSON output to stdout — designed for both human and machine consumption

Architecture

flowchart LR
    cov["OpenCover XML"]
    ros["Roslyn XML"]
    sar["SARIF files"]

    subgraph gen["metricsreporter generate"]
        par["Parsers"] --> agg["Aggregator"] --> ren["Renderer"]
    end

    cov --> par
    ros --> par
    sar --> par

    ren --> json["report.json"]
    ren --> html["report.html"]

    json --> read["read / test / readsarif"]
    read --> ai["AI agent loop"]

Loading

Layer	Key classes	Responsibility
Parsers	OpenCoverMetricsParser, RoslynMetricsParser, SarifMetricsParser	Parse each format into a unified symbol model
Aggregator	MetricsAggregationService	Merge into Solution→Assembly→Namespace→Type→Member tree, apply thresholds, compute deltas
Renderer	NodeHierarchyRenderer, MetricValueRenderer	Serialize queryable JSON + self-contained HTML
CLI	CommandHandler<TOptions>	Four commands backed by DI-composed pipeline

→ Architecture deep dive

Who Is This For

• Solo developers tracking tech debt across coverage, complexity, and violations in one place
• Teams enforcing quality gates in CI with threshold-based exit codes
• AI agent operators running autonomous refactoring loops — no API keys, no servers, just CLI

Documentation

Tutorials	Get your first dashboard running
How-To Guides	Config files, scripts, shipping updates
Reference	CLI commands, config schema, report format, suppressions
Explanation	Architecture, coverage pipeline, namespace inference
Feature Deep Dive	Dashboard, AI integration, suppressions, reconciliation engine

Contributing

git clone https://github.com/baidakovil/metricsreporter.git
cd metricsreporter
dotnet restore && dotnet build && dotnet test

The project follows SOLID principles, uses DI throughout, and maintains a comprehensive NUnit test suite. See CONTRIBUTING.md for code style, branch workflow, and PR guidelines.

License

MIT