2.1 KiB
2.1 KiB
ArbSphere MVP Contracts: Toy DSL Seeds and Canonical IR
This document provides small, self-contained seeds for ArbSphere primitives that can be used by adapters to bootstrap interoperability in a two-venue MVP. The seeds follow the CatOpt-inspired canonical IR mapping used by EnergiBridge-like bridges.
Primitives (DSL seeds)
-
LocalArbProblemSeed
- id: string
- venue: string
- asset_pair: string (e.g. "AAPL/MSFT")
- target_misprice: float
- max_exposure: float
- latency_budget_ms: int
-
SharedSignalsSeed
- version: int
- price_delta: float
- cross_venue_corr: float
- liquidity: float
-
PlanDeltaSeed
- actions: list of dicts with keys {venue_from, venue_to, instrument, size, time}
- timestamp: ISO8601 string
- delta_id: string
- signer: string (optional)
- signature: string (optional)
-
DualVariablesSeed
- multipliers: dict
- version: int
-
PrivacyBudgetSeed
- limit: float
- remaining: float
- expiration: ISO8601 string
-
AuditLogSeed
- entries: list of AuditLogEntrySeed
-
AuditLogEntrySeed
- ts: ISO8601 string or numeric
- event: string
- details: dict
- signature: string
Example: two-venue MVP seed composition
LocalArbProblemSeed: {
id: "seed-venue-A-1",
venue: "venue-A",
asset_pair: "AAPL/MSFT",
target_misprice: 0.5,
max_exposure: 10000.0,
latency_budget_ms: 50
}
SharedSignalsSeed: {
version: 1,
price_delta: 0.12,
cross_venue_corr: 0.75,
liquidity: 20000.0
}
PlanDeltaSeed: {
actions: [ {venue_from: "venue-A", venue_to: "CROSS-VENUE", instrument: "AAPL/MSFT", size: 100, time: "2026-01-01T00:00:00Z"} ],
timestamp: "2026-01-01T00:00:00Z",
delta_id: "delta-001",
signer: "arb-issuer",
signature: "sig"
}
Usage guidance
- Adapters should map their vendor-specific data to these seeds before translation into the canonical IR.
- The seeds are intended to be minimal, deterministic, and extensible as more primitives are added.
Sanity checks
- Validate that LocalArbProblemSeed.asset_pair is two-tuple string when exported to IR.
- Ensure that PlanDeltaSeed.actions are deduplicated in downstream merges to support deterministic replay.