A satirical incident reporting system for condemned digital artifacts. Upload an image. Hotfix processes it. Output: molten slag.
The system analyzes uploaded images using Gemini Vision and files a formal postmortem — classification, severity, failure origin, disposition, archive note — before thermally decommissioning the artifact via dragon-based remediation.
- Gemini Vision analysis — 16-field structured incident schema delivered via Gemini's constrained JSON mode
- Hotfix animation — PixiJS dragon idle, fly-in, and smelt sequence with audio
- Incident postmortem — full structured report overlay with social share (X, Bluesky, Reddit, LinkedIn) plus copy-link
- Global incident manifest — real-time Firestore feed of all thermally decommissioned artifacts
- Decommission index — live cumulative pixel count across all incidents
- Camera support — deploy field scanner via device camera or file upload
| Layer | Technology |
|---|---|
| Framework | React 19 + TypeScript + Vite |
| Animation | PixiJS 8 |
| AI | Gemini (gemini-3.1-flash-lite-preview) via @google/genai |
| Database | Firebase Firestore |
| Audio | Howler.js |
| Styling | Tailwind CSS v4 |
Prerequisites: Node.js
- Install dependencies:
npm install - Create
.envfrom.env.exampleand set your server-side Gemini API key:cp .env.example .env GEMINI_API_KEY=your_key_here - Start the API server (terminal 1):
make server - Start the Vite dev server (terminal 2):
npm run dev
The app runs at http://localhost:3000.
src/
├── App.tsx Main smelter page
├── main.tsx Root + hash-based page routing
├── types.ts SmeltLog + GlobalStats types
├── firebase.ts Firestore client
├── components/
│ ├── SmelterCanvas.tsx PixiJS dragon animation
│ ├── IncidentReportOverlay.tsx Postmortem modal + share
│ ├── IncidentManifest.tsx Global incident manifest page
│ ├── ManifestIncidentCard.tsx Manifest-specific incident card
│ ├── IncidentLogCard.tsx Shared incident log card
│ ├── DataHealthIndicator.tsx Data integrity indicator
│ ├── DecommissionIndex.tsx Cumulative pixel decommission counter
│ ├── HeaderPill.tsx Header pill display component
│ ├── StatItem.tsx Shared stat display component
│ ├── SeverityBadge.tsx Severity tier badge
│ ├── SanctionBadge.tsx Sanction indicator badge
│ ├── P0Badge.tsx P0 priority badge
│ └── SiteFooter.tsx Site footer
├── hooks/
│ ├── useEscalation.ts Escalation vote state
│ ├── useGlobalStats.ts Global stats subscription
│ ├── useLiveIncidentCounts.ts Live incident count subscription
│ ├── useManifestLogs.ts Manifest feed pagination + sorting
│ ├── useRecentIncidentLogs.ts Top-3 incident subscription for P0 badge
│ └── useModalDialog.ts Native <dialog> modal control
├── services/
│ ├── geminiService.ts Gemini prompt, schema, analysis
│ ├── escalationService.ts Escalation vote writes
│ └── breachService.ts Breach tracking writes
└── lib/
├── utils.ts Pixel formatting + color utilities
├── firestoreErrors.ts Firestore error handling
├── smeltLogSchema.ts Strict SmeltLog schema + parser
├── animationWindow.ts Animation timing utilities
├── storageJson.ts LocalStorage helpers
├── typeGuards.ts Runtime type guards
├── postmortemAutoOpen.ts Auto-open postmortem overlay logic
├── impactGlow.ts Impact-score-based glow tiers
└── incidentReportModel.ts Incident report view-model builder
shared/
├── impactScore.js Impact score formula (IMPACT_WEIGHTS, computeImpactScore)
├── colors.js Fallback color palette
└── admin-init.js Firebase Admin SDK initialization
scripts/
├── trigger-sanction.ts Manually trigger a sanction batch
├── reset-sanctions.ts Reset sanction state on all docs
└── firestore.rules.integration.test.ts Firestore rules integration tests
docs/
├── ux-copy.md Voice, persona, and copy rules
├── design-decisions.md Deliberate spec deviations
├── sanction-system-design.md Sanction pipeline architecture
├── sanction-judging-patch.md Sanction judging rebuild brief
├── sanction-test-plan.md Sanction test scenarios
├── sanction-deploy-checklist.md Sanction deploy runbook
└── archive/
├── classification-prompt.md Original AI generation prompt (superseded)
├── judging-prompt.md Original sanction judging prompt (superseded)
├── sanction-rebuild-prompt.md Original sanction rebuild brief (superseded)
├── gemini-implementation-and-share-spec.md Original Gemini spec (superseded)
└── original-ai-studio-prompt.md Original AI Studio scaffold prompt
docs/ux-copy.md— Voice, persona rules, writing constraints for UI and AI copydocs/design-decisions.md— Deliberate deviations from the original spec and their rationaledocs/sanction-system-design.md— Sanction pipeline architecture and data modeldocs/sanction-judging-patch.md— Sanction judging rebuild briefdocs/sanction-test-plan.md— Sanction test scenarios and coverage plandocs/sanction-deploy-checklist.md— Sanction deploy runbook
# Cloud Run (Express server + client)
make deploy # uses .env, deploys to gcloud default project
./deploy.sh --env-file .env.staging # specific env file
# Firebase functions only
firebase deploy --only functions --project anchildress1-unstableThe onIncidentCreated Cloud Function fires on every incident_logs document create. It batches 5 unevaluated incidents, asks Gemini to pick one winner, and commits the result. Fewer than 5 unevaluated → no-op.
Manually trigger a sanction batch (without creating a new incident):
curl -s -X POST "https://onincidentcreated-u36ut3r63a-ue.a.run.app" \
-H "Authorization: Bearer $(gcloud auth print-identity-token --project=anchildress1-unstable)" \
-H "Content-Type: application/json" \
-H "ce-id: manual-$(date +%s)" \
-H "ce-source: //firestore.googleapis.com/projects/anchildress1-unstable/databases/legacy-smelter" \
-H "ce-type: google.cloud.firestore.document.v1.created" \
-H "ce-specversion: 1.0" \
-H "ce-datacontenttype: application/json" \
-H "ce-time: $(date -u +%Y-%m-%dT%H:%M:%SZ)" \
-H "ce-subject: documents/incident_logs/manual-trigger" \
-d '{"value":{"name":"projects/anchildress1-unstable/databases/legacy-smelter/documents/incident_logs/manual-trigger","fields":{"evaluated":{"booleanValue":false}}}}'Returns 204 on success. The function independently queries Firestore for unevaluated docs — the event payload is just a gate.
Check function logs:
gcloud functions logs read onIncidentCreated \
--gen2 --region=us-east1 --project=anchildress1-unstable --limit=15Key fields per incident doc:
| Field | Type | Written by | Purpose |
|---|---|---|---|
evaluated |
boolean | server.js (false) → sanction.js (true) | Claim flag — false = eligible for judging |
sanction_lease_at |
timestamp/null | sanction.js | Claim lease; swept after 5 min TTL |
sanctioned |
boolean | sanction.js | Winner flag |
sanction_rationale |
string/null | sanction.js | Gemini's one-sentence justification |
sanction_count |
number | sanction.js | 0 or 1, used in impact_score formula |
impact_score |
number | sanction.js | 5×sanction + 3×escalation + 2×breach |
Three terminals:
make functions # Firebase emulator: auth:9099, firestore:9180, functions:5001
make server # Express API on :8080
make dev # Vite on :3000 (proxies /api → :8080)Requires VITE_USE_FIREBASE_EMULATOR="true" and FIRESTORE_EMULATOR_HOST="127.0.0.1:9180" in .env. The server auto-pairs FIREBASE_AUTH_EMULATOR_HOST from the Firestore host.
Polyform Shield 1.0.0 — free to use, fork, and adapt; no monetization without permission.
