SSI Scoreboard

Live stage-by-stage comparison tool for IPSC competitions on shootnscoreit.com.

The official site only lets you view one competitor at a time. This app lets you select any group of competitors and compare their results across every stage side-by-side -- including during an active match before official results are published.

Live instance: scoreboard.urdr.dev -- not affiliated with or endorsed by ShootNScoreIt.

Prerequisites

• Node.js 20+
• pnpm 10+ (corepack enable && corepack prepare pnpm@10.30.3 --activate)
• A ShootNScoreIt API key (account settings on shootnscoreit.com)

Local Setup

pnpm install
cp .env.local.example .env.local   # then fill in SSI_API_KEY
pnpm dev                            # http://localhost:3000

Deployment

The app supports two production targets, selected at build time via DEPLOY_TARGET. Both require a Redis instance for hot caching, and a relational store (SQLite or D1) for persistent shooter / match metadata that survives Redis flushes.

Docker / Docker Compose

cp .env.local.example .env.local   # fill in SSI_API_KEY, CACHE_PURGE_SECRET
pnpm docker:build                   # builds image
pnpm docker:up                      # starts redis + app on port 3000

REDIS_URL is set automatically via the compose service name. Two named volumes persist state across container restarts:

• redis_data -- Redis hot cache
• shooter_data -> /app/data -- SQLite store (shooter profiles, match indices, achievements, historical match cache)

Without Docker Compose (bare server, Kubernetes, Fly.io): run a Redis instance and set REDIS_URL to its connection string (rediss:// for TLS). Mount a writable volume at ./data/ (or set SHOOTER_DB_PATH) so the SQLite database persists. The app uses lazyConnect: true, so a missing Redis at startup is non-fatal -- requests fall back to direct GraphQL fetches.

Cloudflare Pages

The Cloudflare build swaps in HTTP-based adapters at build time:

• @upstash/redis instead of ioredis (Workers cannot open TCP)
• Cloudflare D1 instead of better-sqlite3 (Workers have no persistent disk)

You will need an Upstash Redis database and a D1 database.

pnpm cf:build    # DEPLOY_TARGET=cloudflare + @opennextjs/cloudflare build
pnpm cf:deploy   # build + wrangler deploy

Full step-by-step walkthrough -- wrangler login, secrets, custom subdomain, verification, troubleshooting -- in docs/deploy-cloudflare.md.

Note on serverless platforms without persistent disk (Vercel, Netlify default): the app's SQLite AppDatabase needs writable disk for shooter/match metadata. These platforms only work if you switch the persistence layer to D1 or a managed SQL store -- they are not currently supported targets.

Environment Variables

The most common ones an operator needs:

Full reference -- including TTL tunables, telemetry sampling, probe kill switches, and all internal flags -- in docs/env-vars.md.

Usage

1. Browse competitions month by month using the month navigator on the landing page, or paste a match URL (https://shootnscoreit.com/event/22/26547/) to jump straight to a match. The collapsible Filters panel narrows results by discipline, country, and level.
2. Select competitors individually by name, number, or club -- or tap Add squad to load an entire IPSC squad at once (up to 12 competitors total).
3. Multiple analysis views update immediately (stage table, hit-factor and HF% charts, speed/accuracy scatter, balance radar, coaching panels, style fingerprint, AI tip).
4. Use the Share button to copy the link with selections encoded -- recipients open the same match with the same competitors pre-selected.

Features

Comparison & rankings

• Side-by-side stage table with hit factor, time, hit zones, rank, percentile, shooting order, stage difficulty, run classification
• Group / Division / Overall ranking contexts
• Delta heatmap view (each cell relative to the group leader)
• Incomplete scorecard flag (IPSC rule 9.7.6.2)
• DQ banner and clean-match indicator

Charts

• Hit-factor bar chart with optional benchmark overlay
• HF% vs stage-winner line chart
• Speed-vs-accuracy scatter with iso-HF reference lines
• Stage-balance polar radar
• Style fingerprint (alpha ratio vs PPS) with archetype labels
• Style radar (composure, consistency, multi-axis profile)
• Info popovers on every chart explaining axes and reading tips

Coaching analysis

• Consistency score (CV of HF% across fired stages), efficiency (points per shot)
• Penalty rate, match-percentage impact, points left on the table
• "One stage away" what-if analysis
• AI coaching tips for completed matches (sparkle icon; requires AI provider configuration)

Cross-competition (shooter dashboard)

• Per-shooter career view with match history, aggregate stats
• Tiered achievements that survive the 200-match prune window
• Manual match URL submission, dashboard backfill from cached data

Search & navigation

• Month-by-month browser; typing a query switches to full-history search mode
• Filters: firearms, country (ISO 3166-1 alpha-3, defaults to SWE), level (L2+, L3+, L4+, all)
• Recent matches list (localStorage)
• Shareable URLs encode the full competitor selection

Platform

• Server-side cache with smart TTL tiers (future / pre-match / active / complete)
• Persistent SQLite/D1 store for shooter profiles, match indices, achievements, historical match cache
• Dynamic Open Graph images per match (Satori / next/og)
• MCP server (HTTP + stdio transports) -- 7 tools for AI assistant integration
• New-version banner; PWA installable; mobile-first (390px); WCAG 2.1 AA; dark mode

Development Commands

pnpm dev          # start dev server on port 3000
pnpm build        # production build
pnpm lint         # ESLint -- zero warnings required
pnpm typecheck    # TypeScript type check -- zero errors required
pnpm test         # Vitest unit + component tests
pnpm test:watch   # Vitest in watch mode
pnpm test:e2e     # Playwright E2E tests (mocked API, no live key needed)
pnpm test:e2e:ui  # Playwright with interactive UI

Quality bar: pnpm lint, pnpm typecheck, and pnpm test must all pass with zero errors and zero warnings before merging. CI enforces this.

Intentionally pinned major versions

Held back from latest -- don't upgrade without a migration plan:

MCP Server

SSI Scoreboard exposes an MCP server so Claude and other AI assistants can query competition data directly. Seven tools: search_events, get_match, compare_competitors, get_stage_times, get_popular_matches, get_shooter_dashboard, find_shooter.

HTTP transport (public, no local setup needed):

POST https://scoreboard.urdr.dev/api/mcp

stdio transport (local subprocess) -- .mcp.json at repo root pre-registers two stdio servers:

• ssi-scoreboard -- calls the live production instance (no local server needed)
• ssi-scoreboard-local -- calls localhost:3000 (requires pnpm dev running)

Claude Code picks up .mcp.json automatically when you open the repo.

For client-specific setup (Claude Desktop, generic HTTP clients), example prompts, and troubleshooting, see docs/mcp.md.

Architecture

Browser -> Next.js Route Handlers -> shootnscoreit.com/graphql/
         ^
    POST /api/mcp (MCP clients / Claude)

The SSI_API_KEY lives server-side only and is never sent to the browser. Route Handlers are the only place that touches the upstream GraphQL API.

Two persistence layers:

• Redis (ioredis on Docker, @upstash/redis on Cloudflare) -- hot cache for active matches and per-request data.
• AppDatabase (SQLite on Docker, D1 on Cloudflare) -- durable store for shooter profiles, match indices, achievements, and historical match data offloaded from Redis.

Key entry-points:

• app/api/match/[ct]/[id]/ -- match metadata (stages, competitors, scoring progress)
• app/api/compare/ -- fans out scorecard queries, merges ranking data
• app/api/events/ -- event search with date-range, firearms, country, level filters
• app/api/shooter/[shooterId]/ -- cross-competition shooter dashboard
• app/api/og/match/[ct]/[id]/ -- dynamic Open Graph images
• app/api/mcp/ -- MCP HTTP endpoint (JSON-RPC, single-shot transport)
• app/api/coaching/ -- AI coaching tips
• app/api/admin/cache/ -- authenticated cache purge / health / force-refresh

For the full module layout and contributor conventions, see CLAUDE.md.

Design System

This project uses Tailwind v4 with CSS custom property design tokens (OKLCH color space) defined in app/globals.css. Always use semantic token classes (bg-background, text-muted-foreground, etc.) rather than raw palette classes (bg-gray-100). Dark mode is supported via next-themes and the .dark class -- all tokens have light and dark values. A theme toggle in the toolbar lets users switch at runtime.

shadcn/ui is the primary component library. Never modify files in components/ui/ directly; use pnpm dlx shadcn@latest add to update them.

Accessibility

This project targets WCAG 2.1 AA compliance:

• All interactive elements are keyboard-navigable with a visible focus ring
• Minimum touch target size 44x44px (enforced globally in globals.css)
• Error states use role="alert" for screen reader announcements
• Color is never the only means of conveying information -- pairs with shape, icon, or pattern
• Semantic HTML throughout (<button>, <table>, <th scope>, etc.)

Support

If this tool saves you time at the range, consider buying me a coffee: