ViSAGE

An interactive 3D visualization package for SAGE26 semi-analytic galaxy formation outputs.

Renders dark matter haloes and SAGE galaxies together in a browser-based interactive viewer powered by PyVista and Trame.

Features

Rendering

• World-space gaussian splat rendering of haloes and galaxies — splats scale with camera distance and stay physically meaningful at any zoom
• Structure render mode: each galaxy drawn as a layered composition — cold-gas envelope (blue, sized by ColdGas) + outer envelope (green for CGM-regime sized by CGMgas, red for Hot-regime sized by HotGas)
• 27 selectable matplotlib colormaps, identical lists for halo and galaxy layers
• Live colormap, colour-by mode, opacity and visibility controls per layer
• Colour-by dropdowns are model-aware — only modes whose underlying field is present in the loaded model appear in the list; they update automatically on model switch
• Full still-quality rendering at all times — no resolution drop during camera drag or playback

Playback & camera

• Play / Pause / Stop / Reverse / Repeat transport at 0.1× – 5× speeds
• Continuous camera rotation (CW / CCW at 15° / 30° / 60° per second)
• Reset / Centre / Focus buttons
• Fly to halo, galaxy, coordinates, or sub-box (with focus mode that masks everything outside)
• Draw Sphere (Coords tab): place a live two-handle sphere in the viewport — drag the centre ball to translate, drag the edge ball to resize; Lock Sphere commits it as the active focus region
• Draw Box (Box tab): place a live resizable box widget — drag any face or corner handle; Lock Box commits it; Clear on both tabs cancels the widget without navigating
• Switching models always lands at z=0 of the new model; slider and snap chip update immediately
• Camera bookmarks (save, restore, delete)

Selection & inspection

• Galaxy Info panel (Target tab) — GalaxyID, type, halo Mvir, stellar mass, sSFR, cold gas, B/T, BH mass, H2 mass, gas regime, FFB regime, environment classification, mass-weighted stellar age
• Group Info panel (Environment tab) — FOF-aggregate stats: classification, member breakdown (centrals vs satellites), host Mvir, total stellar / cold gas / SFR, mean B/T, spatial extent, target role, BCG stellar mass
• Highlight Galaxy / Highlight Members buttons add regime-coloured splat overlays — CGM-regime members in dodgerblue, Hot-regime in tomato; the selected galaxy is marked with a white border ring
• Galaxy Info, Group Info, and highlight splats all appear in screenshots and recordings — info panels are composited as overlays; highlight actors are baked into the pre-rendered playback frames
• Double-click any point in the viewport (any tab) to populate the Target tab's halo + galaxy IDs and draw a red marker on the selection. Only currently visible galaxies (passing all filters and focus) are selectable. If Focus is active, the camera carries to the new selection at the last-used radius.
• Enter to run in every input field — Halo idx, Galaxy idx, Coords X/Y/Z, Box bounds, Console command, script path, screenshot/movie label all submit on Enter, equivalent to clicking the paired Go / Zoom / Run / Take Screenshot button

Filtering

• Halo filters: Mvir (log10), Rvir (Mpc/h), Vvir (km/s)
• Galaxy filters: stellar mass, sSFR, B/T, age, BH mass, ICS mass, type (centrals / satellites), FFB regime, CGM / Hot regime, environment class (Field / Isolated / Group / Cluster, via checkboxes in the Environment tab)
• Filters are active-only — a slider sitting at its full-range endpoints has no effect; move it inward to filter. Every galaxy with detectable mass is visible at startup.
• Filters auto-disable when the loaded model doesn't contain the underlying field
• Reset Filters button restores defaults
• Playback respects all scene state — the pre-render frame cache is keyed on filter values, focus region, layer visibility/opacity/color-mode, and highlight indicator state; changing any of these and pressing Play again always produces fresh frames

Side-by-side multi-box comparison

• Load two or more SAGE models side-by-side in a single viewport with +SBS in the Models section of the hamburger menu
• Each box is fully independent: its own snapshot, filters, colormaps, opacity, and visibility settings
• A box strip at the bottom of the viewport shows all loaded boxes; click any box label to make it active — the entire right panel (Structure, Filters, Target, Console, …) then controls that box
• Active box label is green; idle boxes are white
• Play, step, and the snapshot slider advance only the active box's snapshot
• Rotation is disabled in multi-box mode (all boxes share one camera; independent rotation is not supported)
• Halo Mvir colour mode is always locked to Viridis; the colormap selector is greyed out when Mvir is selected
• CLR button in the box strip resets that box to its defaults without affecting others

Multi-model (overlays)

• Auto-scans <sage_root>/output/ for SAGE model subfolders
• Switch the primary model from the hamburger menu (any box size)
• Overlay a second compatible model on top (same box size + snap count)
• Loading spinner during model swaps; warning snackbar for incompatible overlays

Output

• Screenshots in PNG / JPG / TIFF
• Movie recording in GIF / MOV (H.264, via ffmpeg) / PNG sequence
• Configurable FPS (1 – 60) and resolution (Native / 2× / 4× supersampled)
• Optional user-typed label per capture; everything saves into sage_outputs/session_<timestamp>/ in your current working directory
• Overlay compositing — Galaxy Info, Group Info, console pop-out, and open Library cards are all composited into screenshots and recordings exactly as they appear on screen
• Catalogue export (CSV, HDF5, FITS, TXT) for the current filter selection, target, or box region — or, in Lightcone Mode, the whole cone (including any synthetic-photometry columns) — saves to sage_outputs/catalogues/

Story Mode

• Play JSON-defined stories — ordered scenes of captured viewer state (snapshot, camera, layers, filters, focus) with text/media overlays — as full presentations over the live 3D view
• Scenes carry titles, headings, body text, citations, LaTeX equations (vendored KaTeX, offline), images, videos, audio, and a clickable scene-menu grid
• Camera motions per scene: orbit, snapshot sweeps through cosmic time (with pre-rendering for instant playback), and fly-throughs with optional snapshot rewind to a target redshift
• Portable stories: symbolic snapshot references ("first", "last", "40%", "z=1.5") resolve against whatever model is loaded
• Stories load from sage_stories/ in your launch directory; a bundled Presentation Template provides a full talk skeleton to copy and fill in — see the Story Mode guide

Launch Mode wizard

• Guided setup flow, accessible standalone (visage with no --par/--lightcone) or from the Launch-Mode dropdown (SAGE-logo button, top-left of the toolbar)
• Three guided flows share the same wizard shell, switchable at any time ("Back" / dedicated toolbar buttons):
  • SAGE26 setup — clone, compile, configure, and run SAGE26
  • SAGEswarm — clone SAGEswarm, install its Python requirements, configure run_pso.sh, and run the PSO calibration with a live plot gallery
  • LightSAGE — clone LightSAGE (upstream repo sage-home/sage-lightcone), build only the sage2kdtree / cli_lightcone tools (SAGE itself is never rebuilt — ViSAGE feeds it your existing SAGE26 output), configure and run the two-stage pipeline, then jump straight into Lightcone Mode on the result — with an optional third stage that synthesizes broadband photometry (SED) via FSPS (pip install "sage-viewer[sed]")
• Step chips in the header track progress per flow (cyan = current step, green = done, white = pending)
• Rescan button re-runs the environment scan from scratch at any point
• Clone SAGE26 option clones the SAGE26 repository from GitHub — prompts for the parent directory (defaults to home folder) before cloning
• Create config file option generates a new .par from a template pre-filled with paths for your SAGE26 directory; choose a custom filename before writing
• Every editable config (.par, run_pso.sh, run_lightcone.sh) is shown as a parameter form — one labelled box per option, pre-filled with its default — instead of raw text; edits fold back into the file on Save & Run, preserving comments and layout
• The LightSAGE build auto-detects a macOS Apple-clang/SDK mismatch and falls back to a compatible SDK; the generated build/run scripts live in ~/.visage/, never inside the third-party checkout
• Screenshots, recordings, catalogue exports, and LightSAGE lightcone output all save to sage_outputs/ in the directory you launched from
• Wizard always resets cleanly when reopened

Lightcone Mode

• visage --lightcone FILE opens a cli_lightcone HDF5 output in the exact same Explore UI as a SAGE box — same toolbar, navigation panel (every colour-by mode), info panel, and gaussian-splat rendering
• Reads every SAGE field carried in the flat lightcone file into a full galaxy snapshot, plus host haloes built from the Type == 0 centrals
• The snapshot slider becomes a redshift/time cut: it only spans the snapshots actually present in the cone, and moving it removes the near (lower-redshift) side of the cone, keeping the far side — the full cone shows at the slider's maximum
• Camera — reset frames the cone zoomed-in, horizontal and centred, end to end; the go-to-centre button stands you at the observer (coordinate origin) looking outward along the cone
• Photometry tab (replaces the Box tab) — if the cone carries synthetic photometry, a separate, independent splat layer that builds a false-colour image from a stack of the ticked filters (each tinted its representative colour; mass-to-light stackable too). Own Visible/Opacity; show it with the galaxies on, off, or on its own
• Reach it from the Launch-Mode wizard's "Visualize lightcone" step after a run, the wizard's Load Existing Lightcone button, the Session Models list (see below), or directly via --lightcone

Session models

• The Launch-Mode dropdown lists every box and lightcone opened so far this session under Session Models, with a box or telescope icon per kind and the active one marked
• Click any entry to jump straight back to it (a quick relaunch on the same port) — so loading a box after a lightcone (or vice versa) never loses track of what you had open
• Persisted in ~/.visage/session_models.json across relaunches

Embedded console (Console tab)

• Terminal mode — a live xterm.js terminal backed by a real PTY ($SHELL -l); full ANSI colour, cursor control, and interactive programs (vim, top, htop, less) all work
• SAGE command mode — natural-language SAGE commands (show only clusters, go to halo 42, snap 30, screenshot, …); switch via the SAGE Cmds button, terminal returns to the shell
• Multiple sessions with a + button — each console has its own PTY process and command history
• Pop-out floats a movable / resizable console card over the viewport so you can keep typing while watching the render

Self-contained metadata

• Cosmology (h, Ω_m, Ω_Λ), box size, and snapshot redshifts are read directly from model_0.hdf5's Header/Simulation
• The .par file is now only needed for tree-file paths

Supported simulations

Both supported automatically — point at the .par file and ViSAGE figures out the rest from the HDF5.

Quick start

pip install sage-viewer

Explore Mode — view existing SAGE26 results by pointing at a .par file:

visage --par /path/to/SAGE26/input/millennium.par

Launch Mode — configure and run SAGE26 via the guided wizard. Run from your SAGE26 root so the wizard can find your .par files and executable automatically:

cd /path/to/SAGE26
visage

Open the printed URL in any browser. To launch on a remote cluster and view locally, use SSH port-forwarding:

# On the cluster
visage --par millennium.par --port 8080

# In a local terminal
ssh -L 8080:localhost:8080 user@cluster
# Then open http://localhost:8080 in your browser

Command-line options

--par FILE              Path to a SAGE .par file — omit (with --lightcone) to launch in Launch Mode (wizard)
--lightcone FILE        Path to a cli_lightcone HDF5 output file — opens the
                        full Explore UI on the lightcone instead of a SAGE box
--par-dir DIR           Directory to scan for additional .par files
                        (defaults to the parent of --par; used for the
                        multi-model dropdown)
--snap N                Initial snapshot number (default: last = z=0)
--port N                Trame server port (default: 8080)
--n-jobs N              Worker threads for parallel halo file reads
--max-halos N           Downsample ceiling per snapshot
--min-halo-mass MSUN    Minimum halo mass to load
--min-stellar-mass MSUN Minimum stellar mass to load

Galaxies have no display cap — every galaxy above the mass floor loads, since all snapshots are preloaded up front.

Multi-model workflow

If your SAGE root looks like:

SAGE26/
├── input/
│   ├── millennium.par
│   ├── millennium_vanilla.par
│   └── microuchuu.par
└── output/
    ├── millennium/model_0.hdf5
    ├── millennium_vanilla/model_0.hdf5
    └── microuchuu/model_0.hdf5

then visage --par input/millennium.par discovers all three models automatically. Click the hamburger icon (top-left) → Models section to switch, or click "+ overlay" next to a compatible model to render both at once.

Installation

PyPI (recommended)

pip install sage-viewer

Requires Python ≥ 3.10. After install, visage is available as a command. If your shell can't find it, add the user bin directory to your PATH:

# macOS (Python 3.12 user install)
export PATH="$HOME/Library/Python/3.12/bin:$PATH"

Movie recording in MOV format requires ffmpeg in your PATH.

Synthetic photometry (SED synthesis for lightcones) is an optional extra — it pulls in FSPS + astropy, which the base install omits:

pip install "sage-viewer[sed]"

FSPS also needs its stellar-population data on disk, pointed at by the SPS_HOME environment variable (see the python-fsps docs). Nothing else in ViSAGE depends on it.

From source

git clone https://github.com/MBradley1985/ViSAGE
cd ViSAGE
pip install .

HPC / supercomputer

A helper script is included for module-system clusters (Slurm, PBS, etc.):

# Load a Python module first (name varies by cluster)
module load python/3.12.0

# Create a venv and install ViSAGE in one step
./install_hpc.sh

# Optional: place the venv on scratch for faster I/O
./install_hpc.sh /scratch/$USER/visage-env

The install is editable (pip install -e .) so a git pull updates the code immediately with no reinstall. Load ffmpeg via your module system if you need MOV recording.

In every session:

source .venv/bin/activate
visage --par /path/to/millennium.par --port 8080
# SSH-tunnel the port to your local browser

Documentation

Full documentation at mbradley1985.github.io/ViSAGE.

Tabs at a glance

When multiple boxes are loaded a box strip appears at the bottom of the viewport. Click any box to make it active (green label). All tab controls then target that box.

The Focus button (top of the right panel) is tab-aware: it focuses on whatever's active in the current tab (target galaxy, environment halo, coords point, or box region).

License

MIT — see LICENSE.