Documentation
Algorithm evidence, runtime verification, and confidentiality boundaries in one auditable surface.
This page follows an evidence-forward documentation pattern used by top product organizations: public implementation proofs, reproducible verification steps, and explicit separation between what is disclosed and what remains proprietary.
SmartSwap+ Portfolio Optimizer
Optimal Search Overlay
StableX Depeg Optimization
Cryptographic Source Attestation
Algorithm Implementation Registry
Each row maps an algorithm ID to its deployed entrypoint, source file attestation hash, and verification endpoint for runtime confirmation.
| Algorithm ID | Component | Entrypoint | Source File | SHA-256 | Verification Endpoint | Verification Field |
|---|---|---|---|---|---|---|
| Loading algorithm evidence... | ||||||
Loading evidence policy...
Runtime Verification Protocol
How to validate claims without source disclosure.
Verification is performed against live API output fields and source-file hashes exposed by the documentation proof endpoint. This proves deployed algorithm lineage while preserving proprietary internals.
curl -sS https://amarq.app/api/app/documentation/proof
curl -sS https://amarq.app/api/smartswap/portfolio-optimize \\
-H "Content-Type: application/json" \\
-d '{"mu":[0.05,0.04],"sigma":[[0.1,0.02],[0.02,0.08]]}'Confidentiality Boundary
What is disclosed and what remains proprietary.
amaRQ publishes algorithm identity, entrypoint provenance, source attestation hashes, and runtime verification paths. Proprietary model-tuning logic, search heuristics, and protected implementation details remain confidential by design.
Documentation Style
Built for executive review and engineering audit.
This section is intentionally structured like top product docs centers: concise claims, verifiable artifacts, and direct links between architecture statements and executable runtime checks.
Pre-print Alignment (Public)
This section provides a whitepaper-style bridge between the public pre-print claims and deployed runtime behavior, while keeping protected implementation and key material out of public documentation.
Optimization Objective
Constrained portfolio utility under risk control.
The deployed optimizer aligns with the published mean-variance utility framing on the portfolio simplex, where allocation vectors satisfy non-negativity and unit-sum constraints.
maximize L(w) = mu^T w - (lambda/2) w^T Sigma w
subject to sum_i w_i = 1, w_i >= 0Boltzmann Sampling Layer
Temperature-scaled concentration around high-utility states.
Scenario sampling follows an inverse-temperature weighting scheme where larger beta concentrates probability mass near global utility maximizers, supporting robust search behavior across candidate allocations.
P_beta(w) ∝ exp(beta · L(w))
beta -> infinity => mass concentrates on argmax L(w)StableX Labeling Logic
Deviation and duration are jointly enforced.
The primary operational label follows sustained deviation logic on 5-minute candles, with published severity tiers maintained for audit transparency.
primary: |VWAP_close - 1.00| > 0.005 for >= 3 consecutive 5m candles
moderate: |price - 1.00| > 0.01 for >= 15 minutes
severe: |price - 1.00| > 0.05 (any duration)Disclosure policy: no private keys, secrets, provider credentials, internal billing identifiers, or protected model internals are published on this page. Public evidence is limited to attestation hashes, endpoint-level runtime checks, and high-level methodology summaries.