Eddy3D Component list

Opens the Eddy3D setup window: what each simulation capability needs, what is missing on this machine, and how to install it. Replaces the Install Engines components.

Open the native Eddy3D assistant in Codex, Claude, or Antigravity. Esinti discusses simulation goals, explains workflow tradeoffs, and can apply confirmed component, wiring, geometry, or template changes to this Grasshopper canvas.

Load example Grasshopper definitions for common workflows. Templates include microclimate simulations, outdoor comfort studies, and CFD analysis setups. Version: 1.12.0.827

Download an EPW weather file from a direct URL, or search climate.onebuilding.org by station name, WMO ID, or dataset year.

Engine settings for Morph Weather: climate products, spatial interpolation, uncertainty case and solar methods. Every value left empty stays on the Future Weather Generator's own default.

Morph a present-day EPW into future-climate EPWs with the Future Weather Generator (future-weather-generator.adai.pt), then feed the result to any Eddy3D workflow. Needs Java 17+ and the generator's .jar, which Eddy3D does not ship: download the distribution you need (CMIP6 Global, CORDEX-CMIP5 Europe, …) into ~/Eddy3D/FWG. The tool is licensed CC BY-NC-SA 4.0 — noncommercial use, attribution required.

Deconstruct a Weather object into hourly time series values. OutdoorPlus

Visualize monthly dry-bulb temperature and relative humidity from Eddy3D Weather as two aligned min/mean/max charts.

Convert specific humidity (w) and temperature (T) to relative humidity (%). OutdoorPlus

Read an EPW file and create a Weather object for the simulation. OutdoorPlus

Visualize a wind direction on a compass circle. Direction is meteorological degrees (0=N, 90=E, 180=S, 270=W); outputs the flow vector and the 16-point cardinal name.

Cluster annual wind conditions into a budget of representative directions using k-means over hourly wind vectors (speed x direction): frequent, strong conditions attract the budget, and each cluster reports an observed direction/speed pair plus its frequency. Without wired speeds, clusters directions alone (unit vectors).

Define an analysis period (from/to day of year, start/end hour of day) and output the hour-of-year indices it covers, for filtering annual results.

Interactively select a month, day, and hour for annual analysis. Click the controls on the component to change the date and time.

Convert a start date/time and optional end date/time into hour-of-year values (1–8760) for indexing annual hourly data.

Translate a Ladybug analysis period to hours of the year.

Export geometry to STL format for OpenFOAM or other CFD tools. Supports meshes and Breps (auto-meshed); binary or ASCII, single or multiple files.

A boolean toggle that is always FALSE when a file is opened. Useful for preventing automatic execution of heavy work. When Run is connected to several component inputs, they run one Grasshopper solution at a time in canvas order. Double-click to toggle.

Hold the last output while the left mouse button is pressed, then publish the newest value when the button is released. Useful between an MD Slider and an expensive component.

Define atmospheric boundary layer inflow conditions for Eddy3D.

Define inflow boundary conditions from a manually entered vertical profile (z/zR, U/UR, k/UR^2) instead of the parametric ABL log-law. Writes fixedProfile inlet conditions for U, k and epsilon. epsilon is derived from the profile as epsilon(z) = Cmu^0.5 * k(z) * d(U)/dz.

Create a uniform (constant velocity) inflow boundary condition for Eddy3D.

Compute the snappyHexMesh refinement level needed to reach a target cell size (each level halves the cell size).

Define a cylindrical simulation domain for Eddy3D. One cylindrical mesh serves all wind directions; the cylinder side faces switch between inlet and outlet per direction. The auto radius targets the 3% frontal-blockage limit of ASCE/SEI CWE Prestandard AC 6-8b, which the case component verifies. Model surrounding buildings within ~240 m of the study area (ASCE 49 proximity guidance) before trusting results near the context edge.

Configure mesh refinement, layers, and grading for Eddy3D.

Define a custom snappyHexMesh refinement region (a box, solid or surface) — refines the cells inside/near the geometry to the chosen level. Wire Extras into a case component so the region is written every time the case is written; or wire a written Case in and press Apply to edit the dictionaries in place (which a later re-write undoes).

Generate centered surface samples on the actual faces of Brep, surface, or mesh geometry.

Assign a multi-face ground plate to the wind tunnel: each face gets its own aerodynamic roughness length z0 and becomes its own ground patch (nutkAtmRoughWallFunction). Feed into the wind case component's Ground Roughness input.

Fetch land-cover polygons around a location from OpenStreetMap (open data, Overpass API) and classify each into an aerodynamic roughness length via the Davenport-Wieringa terrain classification — plus the terrain elevation around the site (AWS Terrain Tiles, open data). Outputs ready-made ground roughness zones and a terrain mesh for the wind case.

Define a pollutant emission source for the wind study: a closed volume (stack tip, traffic corridor box, exhaust vent) releasing a named species at a mass rate. Wire into the Eddy3D Case component's Sources input; the species is transported as a passive scalar with turbulent diffusivity (Sct) on every direction case, and the concentration field (kg/m3) is read back by probing the species name.

Represents a tree as a porous zone for wind blocking (Darcy-Forchheimer). Feed into the wind case component.

Combine a multi-part building mesh into a single watertight, CFD-ready solid via the bundled Python mesh service (trimesh/manifold3d/pymeshfix). The server auto-starts locally on the first run (uv-managed Python environment; first start installs it, 1-2 minutes) and is reused afterwards.

Define a custom OpenFOAM function object the solver runs at runtime — fieldAverage, yPlus, wallShearStress, forces, surfaceFieldValue, a coded FO, etc. Wire Extras into a case component so it is written every time the case is written; or wire a written Case in and press Apply to edit controlDict in place (which a later re-write undoes).

Create, write, and manage an Eddy3D outdoor wind simulation case.

Configure solver run controls for Eddy3D.

Mesh and run an OpenFOAM case on the selected engine (wind / indoor / UMF).

Writes meshing and simulation scripts (.bat / .sh) into a Scripts/ folder under the wind study, so the workflow can be launched manually outside Grasshopper. The scripts match what the Run component executes. Write the study to disk first (Wind Case 'Write').

Create an air region for the UMF case. OutdoorPlus

Define simulation domain extents and refinement padding. OutdoorPlus

Select a building material from the list and override its properties.

Configure mesh refinement for building regions.

Build a solid building region for the UMF case: from the façade surface meshes, two material wall layers (outer + inner) are extruded inward to model heat and moisture transport through the building envelope.

Creates a STL mesh from geometry using the gmsh application. Useful to create healthy mesh topologies for building elements.

Generate a multi-resolution terrain mesh from input geometry with a solid base. OutdoorPlus

Create a soil-backed grass surface material for the urbanMicroclimateFoam terrain heat, radiation, and evapotranspiration model. OutdoorPlus

Configure mesh settings for terrain and underground regions. OutdoorPlus

Create a terrain region with materials and depth settings. OutdoorPlus

Select a terrain surface material from the list and override its properties. OutdoorPlus

Coupled evaporating water surface for urbanMicroclimateFoam. The water geometry becomes a named air patch that exchanges sensible heat and moisture with the air; water motion and a moving free surface are not solved. Requires an UMF build that includes the simpleWater model.

Configure mesh refinement for vegetation regions. OutdoorPlus

Define vegetation property coefficients for canopy modeling. Shows the recommended coefficients (Leaf Length, rsMin, kc) by default; right-click to show all coefficients (Cd, C, nEvapSides). OutdoorPlus

Create a vegetation region from tree crown solids. Pick a Tree Type preset for a typical leaf area density and foliage drag coefficient, or choose Custom and wire your own LAD. The output drives BOTH the Outdoor+ (OpenFOAM/UMF) case and the LBM Run component's Vegetation input, with consistent canopy drag. Library sources — LAD (crown-average, m²/m³): Sjöman et al. (2021) Arboricult. Urban For. 47(6), plant area index of 64 urban species; Klingberg et al. (2017) Urban For. Urban Green. 26, Gothenburg leaf area mapping; ENVI-met Albero plant database conventions; Zhang et al. (2018) Atmosphere 9(5):198 and Beijing For. Univ. J. (2017) ENVI-met validations for subtropical evergreens. Cd (per frontal leaf area, 0.1–0.3): Katul et al. (2004) Boundary-Layer Meteorol. 113; Mayhead (1973) Agric. Meteorol. 12, conifer wind-tunnel drag; Gillies et al. (2002) J. Geophys. Res. 107(D24), drag vs. wind speed and crown streamlining. Note: NIST TN 2039 (2019) reports Cd ≈ 2.8 against total projected area per volume — a different normalization; do not mix it with the Cd·LAD convention used here. OutdoorPlus

Prepare and run a UMF case. OutdoorPlus

Create, read, and manage an Outdoor+ (UMF microclimate) case. OutdoorPlus

Configure snappyHexMesh settings for the simulation. OutdoorPlus

Configure simulation control settings for UMF. OutdoorPlus

Define simulation timing and optional weather-driven time series. OutdoorPlus

Configure the view-factor discretization for radiation modeling. OutdoorPlus

Tests for buildings and trees for Eddy3D-OutdoorPlus simulation.

Deconstruct a MetaFOAM Region instance.

External wind pressure on a facade per EN 1991-1-4, and the opening flow it drives.

Build an isothermal indoor ventilation case (room + inlets + outlets + sinks) for OpenFOAM 12.

CO2 generation rate of one occupant by age, activity and sex (Persily & de Jonge 2017).

Air exchange through an open window (Maas 1995) and the steady-state CO2 it supports.

Ventilation inlet — defines where air enters the room (diffuser, window, door). Direction is computed perpendicular to the surface, pointing into the room.

Ventilation outlet — defines where air exhausts from the room (return grille, open window).

Set an indoor wall temperature (°C). Wire a room surface to give just that surface its own temperature patch; leave it empty for a single case-wide wall temperature.

A breathing occupant (LOD-0 body with a separate mouth patch) for the Indoor Species Case.

A CO2 passive-scalar source box for an indoor ventilation case.

A volumetric heat source box for an indoor ventilation case (transported temperature scalar).

A Darcy-Forchheimer momentum sink (filter/screen) box for an indoor ventilation case.

A fan/jet momentum source (mean velocity) box for an indoor ventilation case.

An airborne-pathogen passive-scalar source box for an indoor ventilation case.

Build a CO2 species case (OpenFOAM 12 multicomponentFluid) with a breathing manikin.

Grade indoor CO2 (ppm) against EN 16798-1 or another CO2-based IAQ standard.

Sleep-adapted Gagge two-node model (Yan et al. 2022) for bedrooms.

Gagge two-node thermal comfort: SET, ET, PMV, TSENS and DISC.

Vegetation that attenuates the sun instead of blocking it, with an optional leaf-on/leaf-off season. Feed the Canopy input of Sun Hours or Solar Irradiation.

Analysis points with outward normals and per-cell areas over surfaces, at a spacing suited to a building rather than a district.

Daily sun arcs, analemmas and sun positions as curves and points — north-aware, drawn from the same solar geometry the sun studies use.

Sun positions for a period at a chosen time step (6-minute and finer), from site and clock rather than from an EPW's hourly rows.

Per-point sun exposure (and, for an explicit HOY selection, the per-instant trees) from a Shadow Result — without putting a whole year of lit flags on the canvas.

Lit or shaded at each analysis point for one or more sun instants — the shadow's position, not accumulated sun hours.

Direct sun hours and shadow by ray casting — an interactive preview, not a Radiance simulation.

Daily Light Integral per point in mol/m²/day — the photosynthetically active photons (400-700 nm) landing on each point per day, averaged over the analysed period. Shares Solar Irradiation's geometry: beam plus isotropic sky plus a ground term, no interreflection. The ground albedo here is a PAR albedo and is much lower than the broadband one. Vegetation is deliberately not accepted — see the Canopy note in the docs.

Per-point daily light integral — period mean and worst month — and, for an explicit Month selection, the per-month tree, from a Daily Light Integral Result.

Cumulative incident solar energy per point in kWh/m² — beam plus isotropic sky plus ground reflection. An interactive preview, not a Radiance simulation.

Climate-based annual daylight metrics from Radiance annual illuminance matrices: sDA spatial daylight autonomy — fraction of the grid daylit for at least half the occupied hours (LM-83 default sDA300/50%) ASE annual sunlight exposure — fraction of the grid seeing direct sun for more than 250 occupied hours (a glare proxy; LOWER is better) UDI useful daylight illuminance — fraction of occupied hours inside a useful band ASE needs the DIRECT matrix. Without it, ASE is reported as zero and the Report says so — the total matrix would count bright overcast sky as glare.

Daylight factor: interior illuminance under the CIE standard overcast sky as a percentage of the simultaneous unobstructed exterior horizontal illuminance. Feed it the illuminance at each sensor and the exterior illuminance the sky was built for. The ratio is invariant to sky brightness, so the absolute value only has to MATCH the one used for the render — it does not have to be any particular number.

Vertical Sky Component per point, in percent: the diffuse illuminance a vertical plane receives directly from a CIE standard overcast sky, over the illuminance an unobstructed horizontal plane would receive under the same sky. Geometry only — no weather, no orientation, no time of year. An unobstructed vertical plane reads 39.6%; BRE BR 209 reads 27% as the guideline for conventional window design.

Fraction of a target's surface each observer point can see — which seats see the park, which units see the water. The target is sampled by area; rays are blocked by the context and by the target's own body.

Visual openness by ray casting: the unobstructed fraction of a view cone around each point's normal. A design-orientation tool for seats, windows and routes — how open does it FEEL here — not a daylight metric.

Area-weighted min/mean/median/max over a sun result, plus the area and fraction reaching a threshold.

EN 17037 sunlight exposure and BRE amenity overshadowing / APSH against a sun-hours result. Design aid — thresholds are editable and not certified.

Configuration for the MRT + UTCI analysis.

Predefined multi-layer construction (assembly) for the SurfaceTemp admittance solve.

Thermal + optical material properties for a building/ground MRT surface.

Canopy material properties for an MRT tree surface.

Leaf/canopy material properties for an MRT vegetation surface.

Point-specific statistics (and optionally the raw hours) from a SurfaceTemp Result, without putting the full 8760-hour year on the canvas.

Mean radiant temperature at each sensor, hour by hour. MRT = shortwave + longwave. SHORTWAVE — what the sensor absorbs from sun and sky. Either a direct raycast (default, pure C#) or the Radiance DDS chain (MRT Settings -> Radiance Reflections), which adds diffuse sky and interreflection off the surroundings. Radiance returns annual total and direct illuminance per sensor, which are mapped onto the probes' shortwave series. LONGWAVE — what the sensor exchanges with everything around it, weighted by view factors traced from each sensor against the scene and the sky dome. SURFACE TEMPERATURES come from ENERGYPLUS, not Radiance — Radiance is a light transport engine and computes no temperatures at all. With MRT Settings -> EnergyPlus Surfaces on, Eddy3D builds an epJSON from the polygons that actually matter to the sensors (those inside the cumulative view-factor percentile and above the small-face cutoff; everything else is demoted to a shading surface), runs EnergyPlus against the EPW, and maps the surface-specific temperatures out of the ESO back onto the geometry. With it off, every surrounding surface is simply assumed to sit at air temperature. SKY TEMPERATURE is always Clark-Allen from dew point, dry bulb, opaque cloud cover and relative humidity — it needs no engine.

Create comfort sensor probes from meshes (one probe per face center, facing the face normal) and/or points (facing corresponding Normals), mixed freely on one input.

Tags Breps or Meshes as a radiation surface for an MRT analysis. Breps are meshed at Patch Size; Meshes are used face-for-face as given.

Computes the Sky View Factor (SVF) for each input point using the Tregenza 145-patch sky subdivision. Casts 145 rays toward the upper hemisphere and returns the fraction of unobstructed sky directions (0 = fully obstructed, 1 = fully open sky).

Solves outdoor surface temperature per analysis point via the frequency-domain admittance method (no thermal mesh, no warm-up). Feeds a future MRT component alongside Sky Exposure. Method: Beckett, O., Owens, S. and Acred, A. (2026). Applying Frequency Domain Methods for Calculating Outdoor Surface Temperatures. Proceedings of the 12th National Conference of IBPSA-USA, Minneapolis, MN. https://publications.ibpsa.org/conference/paper/?id=simbuild2026_1312

Probe-specific statistics (and optionally the raw hours) from an MRT Result, without putting the full 8760-hour year on the canvas.

Solves MRT on a prepared VF Model: shortwave (direct raycast, or Radiance DDS when MRT Settings enables reflections) + view-factor longwave. Wire the VF Model straight from MRT View Factors for ambient/FFT surface temperatures, or through SurfaceTemp (EnergyPlus) for E+ temperatures. Result feeds Deconstruct MRT and UTCI.

Assembles tagged surfaces + sensors into a radiation model, builds the sky dome, and solves probe-to-polygon view factors. Feed the VF Model to SurfaceTemp (EnergyPlus) and/or MRT Solve. The sweep is the expensive part of an MRT run — solving it once here lets the downstream stages re-run without repeating it.

Surface temperatures via EnergyPlus, mapped onto a solved VF Model. The counterpart of the FFT SurfaceTemp component for the staged MRT pipeline: it consumes MRT View Factors' output (the E+ surface selection depends on the view factors) and its output feeds MRT Solve. Skipping this stage leaves surfaces at ambient temperature unless they carry FFT temperatures from MRT Surface.

Colors the model's surfaces by view factor: the mean each face receives from all sensors, or one chosen sensor's view factors to every face. Values output is per face, in mesh face order, for custom gradients.

Equal-angle fisheye of the hemisphere above one sensor, as a colored mesh — a flat disk or a 3D dome (Display dropdown): sky, building, ground and vegetation per direction, plus the sensor's cosine-weighted sky view fraction.

Inspect any Eddy3D case: Outdoor wind study, Indoor case, or OutdoorPlus (UMF) case.

Reference an existing wind case folder (mesh/ + case_NNN) for post-processing.

Sample fields at points on a solved case, post-hoc. With Run it writes a probes function and runs postProcess on the requested Time (latest by default), then reads the results; without Run it reads existing results. Works on a wind case (one sub-result per direction) or a loaded case.

Extract solver-side streamlines from a solved case with OpenFOAM's streamlines function object (particles tracked through the actual mesh — accurate in refined near-building cells, unlike tracing a probed field). With Run it writes the function-object dict and runs postProcess on the requested Time, then reads the tracks; without Run it reads existing tracks.

Export probed facade pressure coefficients into the EnergyPlus AirflowNetwork as an .idf snippet: WindPressureCoefficientArray (the simulated directions), per-node WindPressureCoefficientValues and ExternalNode objects, ready to paste/merge into a Ladybug Tools (or hand-built) AirflowNetwork model. Enable Pressure Coefficient in Run Settings, probe the Cp field at facade points, and wire the probe tree here.

Probe-specific statistics, per-hour values, a colored probe mesh and an inline legend from an annual wind field or Annual VAF object, without putting the full 8760-hour year on the canvas.

Compute volumetric flow rates (m³/s) across a mesh, treating its vertices as velocity probes. Per face: average vertex velocities × face area × cos(angle to face normal).

Compute Velocity Amplification Factors (VAF) and annual wind speed at probes from CFD or ML wind-prediction results and EPW weather data. VAF (the term used in the wind-engineering literature for what Eddy3D historically called "wind factors") is the local wind speed normalized by the reference speed.

Create a visualization mesh from polyMesh point/face data. OutdoorPlus

Remove ground mesh faces that intersect buildings, creating an analysis ground mesh with building footprints cut out.

Create a metric-aware color legend and an optional colored mesh from point/value samples. Supports wind, solar, sun-hours, temperature and other environmental data.

Read cell connectivity and cell zones for a region. OutdoorPlus

Visualize a probed scalar field — CO2, temperature, age of air, Cp, pressure — as a colored point cloud, a heatmap mesh, or a translucent volumetric cloud. Feed the Probe component's points and one value per point. For velocity, use the Vector Field Viewer.

Visualize a probed vector field: colored velocity arrows, a point cloud, a heatmap mesh, streamlines, or volumetric smoke (pick via Display Mode). Feed the Probe component's points + velocity vectors (Field = U), or any points + vectors. For a field without direction — CO2, temperature, age of air, Cp — use the Scalar Field Viewer.

Export an OBJ mesh from a polyMesh description. OutdoorPlus

Write probed wind results as a CSV for the Eddy3D Visualizer (https://viz.eddy3d.com/): columns X, Y, Z_relative, U_at_z, mag_U, U_x, U_y, U_z — one row per probe point. Upload the file at https://viz.eddy3d.com/ to view the 3D field, coloured by velocity magnitude, with the vector components powering the viewer's particle-flow overlay.

Open a wind case's direction cases in ParaView. All directions are added to the ParaView pipeline browser; click Apply on the ones you want to load (nothing is loaded automatically).

Run the OpenFOAM checkMesh command for a case region. OutdoorPlus

Visualize faces that fail tet decomposition during topoSet. OutdoorPlus

Draws a wind case's residual convergence directly on the Grasshopper canvas, with lightweight timed updates. Wire the case and toggle 'Live' to monitor a running simulation without an external plotter window. When a warm-up ramp is enabled the solver restarts mid-run and writes a separate residual file per phase; all phases are stitched into one continuous curve so you see the full history (warm-up + main), not just the latest phase.

Monitor blockMesh, surfaceFeatures, and snappyHexMesh progress from the mesh case logs.

Parses log files in a case folder and reports any FOAM errors. OutdoorPlus

Open the web-based residual plotter for a wind case's convergence history (one trace per direction).

Read and visualize sets produced by checkMesh. OutdoorPlus

Generate a Markdown report documenting the wind case to the QA discipline of the ASCE/SEI CWE Prestandard (within its steady-RANS pedestrian-comfort allowance): solver and OpenFOAM version, domain and blockage, boundary conditions, numerics and the 2nd-order verdict, mesh quality, convergence, y+, Reynolds, an optional comfort section, limitations, and a compliance summary table.

Compute annual probe-specific PET (Höppe) from simulation outputs: MRT and wind-speed data trees, plus air temperature, relative humidity and the person. Solves in the background; cancel from the right-click menu. For a single point, use "Thermal Comfort".

Compute a thermal comfort metric at a point: UTCI (Ta, RH, wind, MRT), PET (adds the personal inputs), or NOAA Heat Index (Ta, RH only). Pick the metric from the dropdown — the inputs adapt. Wire hourly lists (e.g. EPW series) to compute annual values.

Compute annual probe-specific UTCI from simulation outputs: MRT and wind-speed data trees, plus air temperature and relative humidity. For a weather-only calculator, use "UTCI (Weather)".

Probe-specific statistics and comfort hours/% from a UTCI or PET (Simulation) Result, without putting the full 8760-hour year on the canvas. Data only — wire the Result (or these outputs) into the Thermal Comfort Legend to draw a map.

Classifies pedestrian wind comfort per point from an annual hourly wind-speed series (the Wind Speed output of the Velocity Amplification Factors (VAF) component) against a comfort criterion (Lawson, Davenport, NEN8100). Returns the comfort category, class letter, and activity description for each point.

Bin an hourly point-specific series (e.g. UTCI) into a comfort range or the UTCI thermal-stress categories, per analysis period, and report hours/percent in each band. Feed it a point-specific DataTree (e.g. the UTCI component's output) and, optionally, one Analysis Period per branch (see the Analysis Period / Analysis Period To Hours components); an unwired period covers the whole series as one implicit Annual period.

Color UTCI, PET or NOAA Heat Index temperatures by their thermal-stress categories, or a comfort share by its own bands, and create a matching legend.

Solver controls for the FluidX3D GPU engine (memory, simulated time, export interval, and an interactive real-time window).

Read the time-averaged pedestrian wind field from an LBM case directory. Outputs world-frame points and velocity vectors — plug both into the Vector Field Viewer.

Prepare and launch a container-based OpenLB wind simulation (Smagorinsky LES, time-averaged pedestrian wind field). Uses the same ABL inflow object as the OpenFOAM and FluidX3D engines. Needs Docker Desktop or podman; the solver image is pulled on first run.

Solver controls for the container-based LBM wind engine (grid spacing, warmup and averaging windows, probe layer, GPU, container runtime).

Watch a FluidX3D wind solve live in the viewport: colors an analysis mesh with the velocity magnitude of the newest exported frame while the GPU solver runs, updating as each frame lands. Also shows the final field of a completed run. Wire either the Run component's Case or Folder output into Case, and supply the mesh to read the wind on (e.g. a pedestrian-level plane).

Download an ONNX wind-prediction model from HuggingFace for the Wind Predictor component. Yel 2.0 is public; Esen 1.0 and Poyraz 1.0 need a HuggingFace token. All are 8-channel Wind Predictor models. (Yel 1.0 is a different architecture — the GAN image model the hosted Wind Predictor (Cloud) component runs via its API — and cannot be loaded here.) Models cache in ~/Eddy3D/Models/ and are reused on subsequent runs (a model already downloaded to the old ~/SUS_LAB/ folder is moved over, not re-fetched).

Run ONNX wind-field prediction end-to-end. Computes SDF, building height, Zrelative, U/Uref, direction features from geometry, assembles the 8-channel input tensor, runs ONNX inference, and outputs predicted wind speeds. Supports legacy 1ch (U), 2ch (U + k) and new 4ch (U + k + Uroof + kroof) models.

Predict a pedestrian wind-speed field from buildings without running CFD, using the hosted Eddy3D model (Yel 1.0, a 512x512 image GAN). Rasterizes the buildings and the analysis plane into the model's input image, sends it to the API, and returns the predicted wind speeds plus a colored result mesh. Runs on Eddy3D's server: needs internet, no GPU and no model download, and the free server may need a minute to wake up. For a local GPU run over arbitrary points and multiple wind directions, use Wind Predictor with a model from ML Model instead.

Calculate Pedestrian Wind Comfort using predicted wind fields from the ONNX model.

Compute and export wind dataset features from analysis points and building geometry.

Read processed CSV datasets back into Grasshopper. Supports mag_U and all spatial features.

Resample direction-specific wind-magnitude fields onto a new point grid (nearest-neighbour average, with direction-specific rotation). Prepares grids for GAN applications.

Export the solved MRT field as a machine-learning dataset: one row per sensor per hour with spatial features, hourly climate drivers and the MRT/UTCI targets.