mskit-simu 0.7.0

Mini Simulation Kit — terrain + traffic simulations powered by Simu AI, OpenCTV, JAXA AW3D30

MSKit-Simu — Mini Simulation Kit

A lightweight Python library for terrain-based simulations powered by real-world elevation data, live traffic feeds, open camera streams, Simu (an embedded AI assistant you talk to in plain English), and a 3-D vehicle visualiser with a live spectator mode.

---

✨ What's Inside

Module	Description
DEMLoader	Streams JAXA AW3D30 30 m elevation tiles lazily from HuggingFace — no manual download needed
TrafficRouter	Auto-selects OpenTraffic → UTD19 → synthetic fallback, globally
OpenCTVLayer	Aggregates free public traffic camera feeds (Singapore, London, NYC)
Simu	Embedded AI assistant — understands plain English, picks and runs the right simulation
VehicleScene	3-D vehicle models (car=rectangle, bike=cube, bus=hexagon) + map legend
SpectateSession	Live animated top-down traffic view — vehicles drive around in real time
CLI (mskit)	One command — boots an interactive hub that loads every feature at once

---

🚀 Installation

# Core only — terrain, traffic, cameras, rule-based Simu
pip install mskit-simu

# + AI brain (SmolLM2-360M, ~700 MB download once)
pip install mskit-simu[llm]

# + 3-D visualisation + spectate + GIF export
pip install mskit-simu[viz]

# + GGUF / quantised model support
pip install mskit-simu[gguf]

# Everything
pip install mskit-simu[all]

---

🖥️ One Command — Everything Loaded

mskit

That's it. One command boots the interactive hub and pre-warms every module:

  Loading modules…
  ✅ Loading DEMLoader (JAXA AW3D30)… done
  ✅ Loading TrafficRouter… done
  ✅ Loading OpenCTVLayer… done
  ✅ All systems ready — DEM · Traffic · Cameras · Vehicles · Simu

  mskit [Tokyo] >

Then drive everything from one prompt:

Command	What it does
traffic london	Live speed, flow, congestion + nearby cameras
cameras singapore	OpenCTV live camera feed near Singapore
vehicles	3-D vehicle model scene + map legend
spec session	Live animated top-down traffic view
spec session save out.gif	Export spectate animation as GIF
elevation fuji	JAXA elevation lookup at Mount Fuji
goto paris	Switch active location to Paris
random walk in Tokyo 500 steps	→ Simu AI runs it
water flow near Zurich	→ Simu AI runs it
run everything in London	→ Simu runs all 6 sims
demo	Full feature walkthrough
info	Version + data source summary
help	Full command list
exit	Quit

Start at a specific location

mskit --city tokyo
mskit --city singapore
mskit --lat 51.5074 --lon -0.1278      # London

---

🎥 Spectate Session

Watch vehicles drive around a live animated arena — cars, bikes, and buses all rendered as their correct 3-D shapes in a top-down view.

# In the hub:
mskit [Tokyo] > spec session                     # live window (close to exit)
mskit [Tokyo] > spec session save traffic.gif    # export as animated GIF

from mskit.viz import SpectateSession

# Live window
SpectateSession().run()

# More vehicles, longer
SpectateSession(n_vehicles=20, fps=15).run()

# Export animated GIF (requires Pillow)
SpectateSession(duration_s=5).run(save_gif="spectate.gif")

Vehicle shapes:

Shape	Type	Colour	Size
Flat rectangle	🚗 Car	Blue	4.5 × 2.0 m
Tall thin cube	🏍️ Two-wheeler	Amber	2.0 × 0.6 m
Hexagon prism	🚌 Bus	Green	r = 1.5 m

---

🚗 3-D Vehicle Scene + Map Legend

mskit [Tokyo] > vehicles                         # interactive 3-D window
mskit [Tokyo] > vehicles save my_scene.png       # save to PNG

from mskit.viz import VehicleScene, Vehicle, VehicleType, draw_vehicle_scene

# Quick demo — 2 of each type
draw_vehicle_scene()
draw_vehicle_scene(save="scene.png", show=False)

# Build your own scene
scene = VehicleScene(title="Tokyo Intersection")
scene.add(Vehicle(VehicleType.CAR,         x=0,   y=0,   heading=0,   label="Taxi"))
scene.add(Vehicle(VehicleType.TWO_WHEELER, x=6,   y=2,   heading=45,  label="Bike"))
scene.add(Vehicle(VehicleType.BUS,         x=-8,  y=5,   heading=90,  label="Bus 12"))
scene.show()
scene.save("intersection.png")

---

🖥️ Legacy CLI Subcommands

The following subcommands still work alongside the hub:

mskit simu                                   # full interactive Simu session
mskit run "water flow simulation in Zurich"  # one-shot from terminal
mskit cameras --city singapore               # cameras near Singapore
mskit cameras --city london --radius 3
mskit traffic --city london                  # traffic report
mskit traffic --lat 35.6762 --lon 139.6503   # by coordinates
mskit demo                                   # full feature walkthrough
mskit info                                   # version + data sources

---

🐍 Python API

Talk to Simu (plain English)

from mskit import Simu

simu = Simu(auto_select="untrained", auto_simmode="custom")
result = simu.chat("random walk in Tokyo for 300 steps")
print(result["output"])     # sim results dict
print(result["intent"])     # what Simu understood

Load elevation data

from mskit import DEMLoader

dem = DEMLoader()
elev = dem.elevation_at(35.6762, 139.6503)           # Tokyo → metres
patch = dem.patch(lat=51.5074, lon=-0.1278, radius_km=5)  # numpy array

Run simulations directly

from mskit import DEMLoader, RandomWalk, Projectile, WaterFlow, TerrainAgent

dem = DEMLoader()

rw   = RandomWalk(dem, lat=35.6762, lon=139.6503, slope_bias=0.6)
path = rw.run(steps=500)
print(f"Distance: {path['total_distance_km']:.2f} km")

proj = Projectile(dem, lat=35.3606, lon=138.7274,
                  elevation_deg=45, azimuth_deg=90, speed_ms=80)
traj = proj.run()
print(f"Range: {traj['range_km']:.2f} km")

wf   = WaterFlow(dem, patch_km=10)
flow = wf.run(lat=47.3769, lon=8.5417)          # Zurich

agent   = TerrainAgent(dem, 35.6762, 139.6503, 35.73, 139.74)
episode = agent.generate_episode(max_steps=300)

Query live traffic

from mskit import TrafficRouter, DEMLoader

router = TrafficRouter(DEMLoader())
info   = router.traffic_at(51.5074, -0.1278)   # London
print(f"{info.speed_kmh:.1f} km/h — {info.congestion_level} ({info.source})")

OpenCTV traffic cameras

from mskit import OpenCTVLayer

ctv    = OpenCTVLayer()
report = ctv.traffic_report(lat=1.3521, lon=103.8198, radius_km=3)
print(f"{report.camera_count} cameras near Singapore")
if report.nearest:
    print(report.nearest.image_url)

---

📷 OpenCTV Camera Sources

Source	Region	Cameras	Refresh	Key needed
Singapore LTA	Singapore	87	20 s	❌
TfL JamCam	London	900+	~30 s	❌ (optional)
NYC DOT	New York City	~900	~60 s	❌
Synthetic	Everywhere else	∞	—	❌

---

🚦 Traffic Source Priority

Priority	Source	Coverage
1	OpenTraffic / OSRM	Global road network
2	UTD19 (ETH Zurich)	40 cities, 23,541 loop detectors
3	Synthetic	Everywhere — slope + time-of-day

---

🤖 Simu Brain Options

Option	Activate	Download
untrained	auto_select="untrained"	None — instant
huggingface	auto_select="huggingface"	~700 MB once
custom (HF repo)	auto_select="custom", custom_model="org/repo"	varies
custom (local)	auto_select="custom", custom_model="/path"	None
custom (GGUF)	auto_select="custom", custom_model="file.gguf"	None

---

🗂 Project Structure

mskit-simu/
├── mskit/
│   ├── __init__.py          ← top-level exports (all public API)
│   ├── cli.py               ← unified hub CLI  (mskit command)
│   ├── dem.py               ← DEMTile + DEMLoader (JAXA AW3D30)
│   ├── sims/
│   │   ├── random_walk.py   ← slope-biased terrain walk
│   │   ├── projectile.py    ← ballistic trajectory
│   │   ├── flow.py          ← D8 water runoff routing
│   │   └── agent.py         ← RL terrain navigator
│   ├── traffic/
│   │   ├── opentraffic.py   ← OpenTraffic / OSRM layer
│   │   ├── utd19.py         ← UTD19 loop detector layer
│   │   ├── openctv.py       ← OpenCTV camera aggregator
│   │   └── router.py        ← TrafficRouter (auto source selection)
│   ├── simu/
│   │   ├── simu.py          ← Simu AI assistant
│   │   └── intent.py        ← NLU intent parser
│   └── viz/
│       └── vehicles.py      ← VehicleScene + SpectateSession
└── pyproject.toml

---

📋 Changelog

v0.7.0

• Single mskit command boots the all-in-one interactive hub
• spec session — live animated top-down spectator view with direction arrows, live stats, and map legend
• GIF export — spec session save out.gif (requires Pillow)
• SpectateSession class added to mskit.viz
• Hub pre-warms all modules (DEM, Traffic, Cameras, Simu, Viz) on startup
• Hub commands: traffic, cameras, vehicles, spec session, elevation, goto, info, demo, help
• Pillow added to [viz] extra dependency
• Status bumped to Beta

v0.6.x

• 3-D vehicle models + map legend (VehicleScene, draw_vehicle_scene)
• Car → rectangle · Two-wheeler → cube · Bus → hexagon prism
• mskit.viz module introduced

v0.5.x – v0.6.x

• OpenCTV layer (Singapore LTA, TfL, NYC DOT)
• TrafficRouter multi-source auto-selection
• Unified CLI with simu, run, cameras, traffic, demo, info subcommands
• Simu two-tier startup (brain selection + sim mode selection)

v0.4.x and earlier

• Core DEM streaming, RandomWalk, Projectile, WaterFlow, TerrainAgent
• UTD19 integration
• OpenTraffic / OSRM integration

---

📄 License

MIT — see LICENSE.

Resource	Link
PyPI	https://pypi.org/project/mskit-simu/
GitHub	https://github.com/MegaBites-AI/MSKit
Dataset	https://huggingface.co/datasets/MegaBites-AI/AW3D30-DEM-Tiles
Simu model	https://huggingface.co/HuggingFaceTB/SmolLM2-360M-Instruct