Malten

Spatial AI for the real world. Foursquare if built in the AI era.

Contextually aware of what's around you. Fully agentic - agents continuously index and maintain the world view for each area.

What It Does

Open the app → instantly see:

• 📍 Where you are (street, postcode)
• ⛅ Weather + rain forecast
• 🕌 Current prayer time, next prayer
• 🚏 Live bus/train arrivals with countdown
• ☕ Nearby cafes, restaurants, pharmacies, shops

Move → it updates automatically. No searching, no typing. Just awareness.

Ask anything → AI with spatial context answers.

The Model

See ARCHITECTURE.md and claude.md for the full spacetime model.

events.jsonl     = cosmic ledger (facts about the world, append-only)
spatial.json     = materialized quadtree (current state, rebuildable)
stream/websocket = real-time event propagation
localStorage     = your private timeline (your worldline)

Primitives: Streams, Agents, Commands, Database (quadtree), Events

Privacy: Your conversations stay in localStorage. Only facts about the world go to the server.

How It Works

Instant data: When you arrive somewhere, Malten fetches what you need immediately - weather, transport, places. No waiting.

Smart caching: Data is cached spatially in a quadtree. Move 500m? Get fresh local data. Stay put? Use cache.

Background agents: Agents per area continuously index - more places, more detail. You never wait for them.

Adaptive updates: Moving fast (driving)? Updates every 5s. Walking? 10s. Stationary? 30s.

Try It

go build -o malten .
./malten

Open localhost:9090, enable location.

AI Integration

# Fanar (production)
FANAR_API_KEY=xxx FANAR_API_URL=https://api.fanar.qa/v1 ./malten

# OpenAI (fallback)
OPENAI_API_KEY=xxx ./malten

Data Sources

• Location: OpenStreetMap Nominatim
• Weather: Open-Meteo
• Prayer times: Aladhan
• Transport: TfL (London buses, tubes, trains)
• Places: OpenStreetMap + Foursquare
• Traffic: TfL disruptions

Architecture

User at location
  → Query quadtree (instant)
  → Cache miss? Fetch, store as event, return
  → Agent enriches area in background
  → events.jsonl = source of truth
  → spatial.json = rebuildable from events

No waiting. Cache-first. Fetch on demand. Enrich in background.

Important Files

• ARCHITECTURE.md - The spacetime model (read first)
• claude.md - Full development context for AI assistants
• events.jsonl - Append-only event log (never delete)
• spatial.json - Quadtree state (rebuildable)

License

AGPL-3.0