dashboard-context

name	dashboard-context
description	Provides architectural context for the CI Dashboard Tracker project. Covers the Flask app structure, pluggable collector system, SQLite storage, AI analysis pipeline, Jira integration, and configuration model.

CI Dashboard Tracker Architecture

Overview

The CI Dashboard Tracker is a Python/Flask web application that monitors, analyzes, and tracks CI test failures in OpenShift environments. It provides real-time dashboards showing test failure metrics, historical trends, and platform-specific statistics.

Repository Layout

ci-failure-tracker/
  ci_failure_tracker.py          # Jira Bridge CLI (separate tool)
  teams/                         # Team-specific Jira Bridge configs
  examples/                      # Example configs for adopting teams
  dashboard/                     # Main dashboard application
    dashboard.py                 # CLI entry point (Click)
    wsgi.py                      # WSGI entry for Gunicorn
    config.yaml                  # Team-customizable configuration
    requirements.txt             # Python dependencies
    Dockerfile                   # Container image
    src/
      collectors/                # Pluggable data collectors
        base.py                  # BaseCollector ABC + data classes
        reportportal.py          # ReportPortal API collector
        gcsweb.py                # QE-Private-Deck GCS collector
        prow_gcs.py              # Direct GCS bucket collector
        prow_mcp.py              # MCP server collector
      storage/
        database.py              # SQLite schema, CRUD, migrations
      web/
        server.py                # Flask app, routes, background tasks
        templates/
          dashboard.html         # Single-page app (vanilla JS)
      ai/
        analyzer.py              # Vertex AI failure analysis (Claude)
      integrations/
        jira_integration.py      # Jira issue creation/search
      metrics/
        calculator.py            # Pass rate and trend calculations
      reports/
        weekly_report.py         # Weekly platform breakdown reports
    openshift/                   # OpenShift deployment manifests

Tech Stack

• Web Framework: Flask 3.1.3 + Gunicorn (single worker)
• Database: SQLite with WAL mode for concurrency
• Frontend: Vanilla JavaScript in a single HTML template
• AI: Claude Sonnet 4 via Google Vertex AI (anthropic[vertex])
• HTTP Client: requests 2.32.4
• Config: PyYAML 6.0
• Export: OpenPyXL 3.1.0 (XLSX), CSV, Markdown
• CLI: Click 8.1.0 + Rich 13.0.0

Data Flow

1. User clicks "Refresh Data" or cron triggers collection
2. Configured collector fetches job runs and test results from CI
3. Results normalized to TestResult/JobRun dataclasses
4. Stored in SQLite (indexed for fast queries)
5. Flask serves dashboard with real-time metrics
6. User can trigger AI analysis per test failure
7. User can create Jira tickets from analyzed failures

Database Schema

Five tables:

• job_runs -- job name, build ID, status, version, platform, pass rate
• test_results -- test name, status, error message, version, platform
• daily_metrics -- pre-aggregated daily pass rates
• test_metrics -- per-test aggregated statistics
• ai_analyses -- cached AI analysis results

Configuration Model

config.yaml sections:

• collector.type -- which collector to use (reportportal|gcsweb|prow_gcs|prow_mcp)
• tracking.versions -- OpenShift versions to track
• tracking.platforms -- cloud platforms (aws, azure, gcp, vsphere, etc.)
• tracking.test_suite_filter -- filter tests by suite name
• tracking.blocklist -- test IDs to exclude
• database.path -- SQLite file location
• web.host/port -- server binding

Key Design Patterns

• Pluggable collectors: All implement BaseCollector ABC with normalized data classes
• Manual collection: On-demand via UI button (no automatic scheduling by default)
• Lazy AI analysis: Users click to analyze (not automatic) to control costs
• Shared analysis: AI results are platform-agnostic, cached in DB
• Config-driven: Teams customize via config.yaml, no code changes needed