This laboratory is intentionally split into three epistemic layers:
The real-time API target is https://ialirt.imap-mission.com/space-weather. The parser is defensive: it preserves raw records, discovers likely time/numeric fields, and only maps quantities when their key names match known physical aliases. If CORS, network, or schema conditions prevent browser access, the lab enters clearly labeled DEMO mode rather than failing silently.
Propagation is a ballistic Parker approximation with layered termination-shock, heliosheath, heliopause and upstream-flow geometry; it is not an MHD simulation. Candidate event labels are heuristic classifiers—not cataloged event identifications. The heliopause surface is schematic and adjustable. ENA inverse reconstruction uses regularized angular smoothing/backprojection for exploration; imported ENA data preserve their provenance and energy bins when present. Calibrated science products should be brought in through public CDAWeb/mission exports rather than treated as I‑ALiRT equivalents.
v8 causal graph / ancestry: the persistent ledger is a model-receipt layer built from candidate events, not a curated event catalog. Global graph time is logarithmically compressed for display while milestone timestamps remain physical. A spacecraft-worldline proximity score combines radial-front and angular Parker proximity only where loaded ephemeris overlaps the modeled outbound interval. Reverse ancestry is a compatibility ranking, not causal identification; temporal echo agreement dominates, while event strength and ENA directional intensity act only as secondary weights. Synthetic/model ENA maps are marked as such and cannot upgrade a hypothesis to measured support.
v7 causal trace: selectable event traces are causal bookkeeping models, not assertions that a particular plasma parcel survives unchanged to the heliopause. Earth arrival uses ballistic L1 propagation. Outward travel uses the selected solar-wind speed plus Parker-spiral geometry. ENA echo windows add modeled outward plasma travel to energy-dependent neutral-hydrogen return time. Cross-mission associations are displayed only when independent mission data exist and remain probabilistic constraints.
v6 Earth response / tomography: the dayside magnetopause uses the Shue et al. empirical functional form driven by upstream dynamic pressure and IMF Bz when available. The bow-shock standoff is a Farris–Russell-style Mach-dependent approximation using the available Alfvén Mach number as a proxy when magnetosonic Mach is unavailable; it is labeled accordingly. Unless a coordinate-qualified GSM science product is loaded, mapped Bz is treated as an upstream directional proxy rather than asserted GSM Bz. The Newell coupling function is shown as a relative solar-wind/magnetosphere driver and is not converted into an asserted geomagnetic-index forecast. Epoch tomography is an exploratory multi-energy radial inversion: each ENA energy/radius pair receives its own ballistic source lookback, an iterative positive reconstruction is regularized radially, and residuals are reported. It is not a full MHD/kinetic 4D inversion.
v8.6 fidelity pass: the native AU-scale renderer adds depth-sorted boundary facets, stronger perspective, shaded bodies, motion-trail plasma parcels, and a higher-detail spacecraft glyph. Cesium prefers satellite imagery with the scientific grid retained underneath, adds lightweight outline-only Earth-limb and range geometry, and keeps the v8.5 manual render-loop / invalid-polyline stability fixes intact.
v8.5 Cesium stability audit: the Earth→L1 corridor no longer contains a zero-length Cartesian endpoint; all arbitrary space-vector polylines use ArcType.NONE rather than surface-geodesic interpolation. Cesium is driven by a Helio-owned manual render loop so a single render exception cannot permanently disable the viewer. On a render error, optional dynamic/model layers are quarantined while the core globe/grid/L1 corridor remains active. External imagery retains the token-free grid underneath and listens for asynchronous provider errors. The bridge is pinned to CesiumJS 1.143 for this stabilization pass. Original v5 bridge: CesiumJS is used for Earth-centered spatial context only: Earth orientation, Earth–Sun direction, nominal L1 geometry, modeled magnetosphere shells, and ballistic L1→Earth parcel positions. The AU-scale heliosphere remains in the local renderer because a geospatial globe is not an appropriate numerical frame for the compressed outer-heliosphere reconstruction.
v4 reconstruction layer: the built-in HAPI adapter can inspect public CDAWeb dataset metadata, request a bounded subset of likely physical variables, expand common scalar/vector/sky-map structures, and retain external missions as separate constraints. ENA time-of-flight uses nonrelativistic neutral hydrogen kinematics and therefore represents source-age geometry, not a causal back-propagation of the plasma state. The 3D boundary probability cloud is an exploratory sensitivity envelope, not a Bayesian posterior.