infrared

infrared.city  ·  Simulations  ·  Knowledge base  ·  Install the skill  ·  Cookbook

---

What this is. Two ways to use the Infrared SDK without leaving the tools you already have:

1. An agent skill that teaches Claude Code, Cursor, Codex CLI, GitHub Copilot, and Windsurf how to drive the SDK and read its results.
2. A Jupyter cookbook that walks through every analysis end-to-end with embedded outputs.

Pick one. They share the same SDK, the same auth, the same conventions.

What you can simulate

The Infrared SDK runs eight microclimate analyses on any polygon you give it:

Analysis	Output
Wind speed	m/s grid
Pedestrian wind comfort (Lawson LDDC)	class 0–4 grid
Solar radiation	kWh/m² grid
Daylight availability	hours grid
Direct sun hours	hours grid
Sky view factor	0–100 fraction grid
Thermal comfort (UTCI)	°C grid
Thermal comfort statistics	% of time per UTCI band

Polygons can span one tile or many — the SDK fans out, runs jobs concurrently, and merges results.

What you can fetch

Run an analysis with what's already on the ground, or use the same layers in your own pipeline:

Layer	Format
Buildings	DotBim mesh + per-building heights
Trees / vegetation	Gridded canopy heights
Ground materials	Gridded surface-class IDs
Weather	Filtered EPW-style hourly fields

Layers are populated worldwide from OpenStreetMap-class sources; coverage and freshness vary by region.

Install the skill

The agent skill lives at plugins/infrared/skills/use-infrared/. It's a Claude Code plugin and a Cursor 2.5+ plugin — same content, twin manifests.

Claude Code

/plugin marketplace add Infrared-city/infrared-skills
/plugin install infrared@infrared-skills

Cursor

/plugin marketplace add Infrared-city/infrared-skills
/plugin install infrared@infrared-skills

Codex CLI / GitHub Copilot / Windsurf

These read AGENTS.md from the project root. Either clone this repo into your workspace, or copy plugins/infrared/skills/use-infrared/ into your project's .agents/skills/ directory.

The skill loads progressively — SKILL.md is an 80-line router that the agent reads first, then it pulls per-topic references from references/ only when needed (analysis specs, async/webhook patterns, error handling, geometry, weather, etc.).

Run the cookbook

Two flavours under cookbook/, both auto-synced from the SDK:

• cookbook/notebooks/ — eight pre-executed Jupyter notebooks against five preset cities (Munich, New York, São Paulo, Tokyo, Sydney). Standalone, runnable in any order. Cells ship with embedded outputs from a real run — flip through without executing to see exactly what the SDK produces.
• cookbook/scripts/ — runnable .py examples covering wind, UTCI, multi-analysis, vegetation/ground, tiling, fetch-layers, and the async + webhook lifecycle.

git clone [email protected]:Infrared-city/infrared-skills.git
cd infrared-skills/cookbook/notebooks   # or cookbook/scripts
python -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt
cp .env.example .env   # paste your INFRARED_API_KEY
jupyter lab            # for the notebooks

SDK access

Request an API key at https://infrared.city, then install the SDK from PyPI:

pip install infrared-sdk
export INFRARED_API_KEY=...

Full SDK reference: https://infrared.city/docs/sdk.

Learn more

• About the platform — https://infrared.city/
• What the simulations do — https://infrared.city/simulations/
• Validation, methods, model docs — https://infrared.city/knowledge-base/
• SDK docs — https://infrared.city/docs/sdk

Layout

infrared-skills/
├── plugins/infrared/skills/use-infrared/   # the agent skill
│   ├── SKILL.md                            # 80-line router
│   └── references/                         # per-topic deep-dives
├── cookbook/                               # 8 Jupyter notebooks (auto-mirrored from SDK)
├── AGENTS.md                               # for Codex / Copilot / Windsurf
└── docs/assets/                            # README artwork

License

Apache-2.0. See LICENSE.