Operation: Find [Name] — Live Investigation Board

A tongue-in-cheek out-of-office experience. Colleagues submit "witness statements" about sightings; Claude extracts entities and plots them on a live knowledge graph and map.

Built by: Cameron Stiller · Nuix
Graph library: Ogma by Linkurious

---

What it does

• Colleagues visit the live site and see a real-time investigation board tracking the missing person's movements
• Staff can submit witness statements (e.g. "I saw Cameron at the airport with a suspicious amount of sunscreen")
• Claude API moderates each statement and extracts entities (locations, activities, associates) and relationships
• A background GitHub poller imports pre-planned sightings from a public JSON feed at a scheduled time — so the board updates itself while you're actually on holiday
• Everything lands on an interactive Ogma knowledge graph and Leaflet map that updates in real time via WebSockets

---

Stack

Layer	Tech
Frontend	React 18 + TypeScript + Vite
Graph viz	Ogma (Linkurious) — commercial licence required
Map	Leaflet + OpenStreetMap
Backend	Node.js + Express + Socket.io
AI	None — staff form submits structured fields directly
Geocoding	Nominatim (OpenStreetMap) — rate-limited at 1 req/s
Database	PostgreSQL
Deployment	Docker Compose (local) · Helm + k3s (production)
Real-time	Socket.io (WebSocket broadcast on new sightings)

---

Quick start (Docker)

cp .env.example .env
# Fill in: TARGET_NAME, LAST_SEEN_DATE, POLL_URL
# Optionally set: DB_USER, DB_PASSWORD, DB_NAME (defaults are fine for local dev)

docker-compose up -d
# Frontend → http://localhost:3000
# Backend  → http://localhost:3001

Ogma licence

Ogma is a commercial graph library. You need a licence from Linkurious.
Once you have the .zip, extract the JS files into frontend/public/ogma/ (the path vite.config.ts expects).

---

Environment variables

Copy .env.example and fill in:

Variable	Description
TARGET_NAME	Name of the "missing" person (e.g. Cameron Stiller)
LAST_SEEN_DATE	Display date in the header (e.g. 2026-06-26)
POLL_URL	URL to a JSON array of pre-planned sightings (see below)
POLL_INTERVAL_MS	How often to check POLL_URL (default 300000 = 5 min)
DB_USER / DB_PASSWORD / DB_NAME	PostgreSQL credentials (defaults fine for local)

---

Pre-planned sightings feed

The poller fetches a JSON file from POLL_URL every POLL_INTERVAL_MS. Format:

[
  {
    "time": "2026-06-27T09:20:00",
    "location": "Sydney Airport Terminal 1, Mascot NSW 2020",
    "activity": "Checking in at the departure gate",
    "associate": "Optional — person or thing they were with"
  }
]

• time is treated as AEST (UTC+10) and is also the dedup key — duplicate times are skipped
• Future-dated entries are ignored until their time arrives, so you can commit the whole trip upfront
• Geocoding is done via Nominatim; no coordinates needed in the feed

A good workflow: keep sightings in a separate public GitHub raw URL so you can edit them from your phone while you're away.

---

Kubernetes / Helm deployment

The helm/operation-find-cameron/ chart deploys frontend, backend, PostgreSQL (Bitnami subchart), and Traefik ingress with a /demo subpath.

helm upgrade --install operation-find-cameron ./helm/operation-find-cameron \
  -n your-namespace \
  --set backend.env.TARGET_NAME="Your Name" \
  --set backend.env.POLL_URL="https://raw.githubusercontent.com/..." \
  --set postgresql.auth.password=changeme

Images are expected to be pre-built and imported on the node (imagePullPolicy: Never) — see DEPLOYMENT.md for the full build-and-import workflow.

---

Project structure

├── backend/
│   ├── src/
│   │   ├── index.ts              Express + Socket.io server
│   │   ├── routes/witness.ts     API endpoints
│   │   └── services/
│   │       ├── database.ts       PostgreSQL operations
│   │       ├── geo.ts            Nominatim geocoding + reverse geocoding
│   │       └── poller.ts         GitHub sightings feed poller
│   └── Dockerfile
├── frontend/
│   ├── src/
│   │   ├── App.tsx               Shell, timeline playback, header stats
│   │   └── components/
│   │       ├── GraphVisualization.tsx   Ogma knowledge graph
│   │       ├── MapView.tsx              Leaflet map with suburb clustering
│   │       ├── WitnessForm.tsx          Statement submission (staff only)
│   │       └── LocationAutocomplete.tsx Nominatim address search
│   └── Dockerfile
├── helm/operation-find-cameron/  Helm chart for Kubernetes
├── docker-compose.yml            Local dev / demo
└── .env.example                  Environment variable template

---

Easter egg — submitting sightings

The main site is read-only for most visitors. But if you know the URL, you can submit sightings yourself:

https://your-domain/demo/staff

The /staff route unlocks a witness statement form. Type what you saw and Claude will extract the entities and add them to the graph in real time. Keep it work-appropriate — the API moderates submissions.

This path is intentionally undocumented in the UI. Share it only with people you want contributing to the investigation.

---

Adapting for your own OOO

1. Fork this repo
2. Set TARGET_NAME to your name
3. Create a public GitHub repo with your sightings JSON (see format above)
4. Set POLL_URL to the raw URL of that file
5. Deploy with Docker Compose or Helm
6. Put the URL in your out-of-office email

---

Licence

Source code: MIT
Ogma graph library: commercial licence required from Linkurious — not included in this repo.