Documentation menu
API reference
Radiation
https://api.cogitan.ai/v1/fluxus; a local engine serves the same route under /api. GET routes are free on the hosted API; every POST consumes one call of quota./v1/fluxus/rad/assess
engine path /api/rad/assess · metered
Correlated-error footprint, burst rate and logical floor.
What it does, in full
Correlated-error footprint, burst rate and logical floor.
Composes the calibrated RAD chain: cached Geant4 deposition spectra -> reduced-order phonon transport (graded to 10.2% against G4CMP at a 6.64% noise floor) x the fitted eta trap correction (+/-9.2% conformal) -> per-qubit absorbed energy on the real pitch -> correlated floor and crossover distance.
Every section either carries a value with its ``kind`` or a ``refused`` block, and every refusal is repeated in the top-level ``refusals`` list so a caller reading one section cannot miss one. calibration is per section, and there is no scalar ``calibrated`` -- the burst rate was graded on two devices and the footprint on one, so a single flag would have to lie about one of them. CALIBRATED means the inputs sit AT a graded point; anything else reads BORROWED and names the axis it departs on and the device whose normalisation is reused. calibration_ref and known_misses are always served: the response never looks more certain than the chain is.
Calibrated against Google's gap-engineered Willow (arXiv:2506.18228) and Li et al.'s 63q sapphire device (arXiv:2402.04245). Note the gamma channel is the weakest observable: the incident flux is inferred from device 2 and the cross-check runs 0.29-0.70x of measured.
Request
Body: RADRequest
Correlated-error assessment: die, layout, mitigation, environment, code.
detection_threshold_keV is deliberately OPTIONAL WITHOUT A DEFAULT. It is a device property, not a design output (quoted by Google for Willow at 180 keV, derived from a two-observable ratio for Li et al. at 55 keV), so omitting it yields a refusal on the burst rate rather than an invented one.
Example
curl -s "$FLUXUS_API/rad/assess" -H "Authorization: Bearer $COGITAN_API_KEY" \
-H 'content-type: application/json' \
-d '{"detection_threshold_keV": 180, "thickness_um": 300, "source": "muon"}'No CLI command answers this route.
Example body: Willow's quoted detection threshold (180 keV, from the request model's docstring)
What it refuses
- detection_threshold_keV omitted -> burst_rate refused; it is a DEVICE property and the chain cannot predict it
- trap geometry outside the eta box -> trap_correction refused rather than extrapolated
- thickness outside 0.05-2.0 mm -> burst_rate and footprint refused; inside it, interpolated between Geant4 runs and the response names the two runs
- unknown source -> HTTP 422 from the request model
- qubits off the die, duplicate positions, or n_qubits disagreeing with qubit_xy_mm -> footprint_qubits refused, with the die size a generated grid would need; never clamped. Any layout other than Willow's 60 measured sites reads BORROWED
- inputs away from a graded point -> that section reads calibration.status = BORROWED, naming the axis it departs on and the device whose normalisation is reused. NOT a refusal: the number is still returned.
Responses: 200 Successful Response; 422 Validation Error.
Validated for
fidelity reduced_order_chain_graded_against_G4CMP
TWO PUBLISHED DEVICES, which is the unusual part: Google's gap-engineered Willow (arXiv:2506.18228) and Li et al.'s 63q sapphire device (arXiv:2402.04245). The phonon transport kernel is separately graded against G4CMP (Geant4 11.4.1) at 10.2% of distribution mass misplaced, against a 6.64% statistical noise floor.
The full rad_assess card/v1/fluxus/rad/assess/batch
engine path /api/rad/assess/batch · metered
/api/rad/assess over several designs. Same path, in a loop.
What it does, in full
/api/rad/assess over several designs. Same path, in a loop.
Equality with the single call is enforced by tests/test_rad_service.py::test_batch_is_the_single_path_in_a_loop.
Request
Body: RADBatchRequest
Several assessments in one call. The batch path IS the single path.
Rejected above a measured work budget: see BUDGET_S. The limit is on TOTAL work, not on item count, because 25 cheap designs and 25 expensive ones differ by 20x and only one of them is a problem.
Object: RADRequest
Correlated-error assessment: die, layout, mitigation, environment, code.
detection_threshold_keV is deliberately OPTIONAL WITHOUT A DEFAULT. It is a device property, not a design output (quoted by Google for Willow at 180 keV, derived from a two-observable ratio for Li et al. at 55 keV), so omitting it yields a refusal on the burst rate rather than an invented one.
Example
curl -s "$FLUXUS_API/rad/assess/batch" -H "Authorization: Bearer $COGITAN_API_KEY" \
-H 'content-type: application/json' \
-d '{"items": [{"detection_threshold_keV": 180}, {"detection_threshold_keV": 180, "backside_cu_coverage": 0.5}]}'No CLI command answers this route.
Example body: two mitigation levels side by side
What it refuses
No refusal list is served for this route. Invalid input is a 400 or 422 with the error envelope; any refusal it makes is described above.
Responses: 200 Successful Response; 422 Validation Error.
Validated for
This route has no card in the capability manifest; read its description above for what it rests on.
Generated from Fluxus 0.3.0 at commit b9aa57d on 2026-10-20 by python -m fluxus.docs_bundle.