Orbitra Prime
Trading intelligence
Mark prices, index prices and funding references in Orbitra Prime come from Prism, and confidence is available wherever a decision depends on it — from a liquidation-distance preview to a portfolio valuation.
PrismOrbitra L1
Oracle mesh for prices and external data
Every price arrives with its confidence.
Prism Oracle Mesh is the price and external-data pipeline of ORBITRA ONE™. It gathers observations from exchanges, dealers and benchmarks, normalizes and aggregates them with robust statistics, and publishes signed market state with an explicit confidence score. Risk, matching, funding and settlement then act on each price according to how far it can be trusted.
How it moves
Many observations, one signed value
- 01SourcesExchanges, dealers, benchmarks
- 02NormalizeTime, units, quality
- 03AggregateRobust median + weighting
- 04ConfidenceFreshness and dispersion
- 05PublishSigned market state
The illustration follows price observations through the five stages of the mesh. Sources — exchanges, dealers and benchmarks — enter on the left and are refracted into a single stream as they are normalized for time, units and quality, aggregated with a robust median and weighting, and scored for confidence from freshness and dispersion. On the right, one signed market-state record emerges, carrying value, confidence and provenance.
The problem
A market is only as sound as its reference prices.
Margin requirements, liquidations, funding payments and collateral valuations all depend on reference prices that originate outside the network. When a reference is wrong — stale, manipulated or measuring something slightly different — the damage lands on users who did nothing wrong.
Single-source feeds fail with their source. Simple averages let one bad print drag the result. And many feeds publish a number without saying how reliable it is, so downstream systems treat a thin, quiet quote exactly like a deep, consistent one.
Prism treats external data as evidence to be weighed, not fact to be accepted. Every published value carries its freshness, dispersion and confidence, and every consumer — Aegis, ApexMatch, funding, liquidation and asset valuation — has explicit rules for what to do when confidence falls.
Operating sequence
Five stages from observation to market state.
Each published value is the output of a fixed pipeline, and each stage leaves a record that can be inspected afterwards.
- 01
Sources
Independent observations arrive from exchanges, dealers and benchmark providers. Each source is registered with its coverage, method and track record, and no single source can set a price alone.
- 02
Normalize
Observations are aligned to a common clock, converted to common units and quote conventions, and tagged with quality signals such as age, depth and source status. Malformed inputs stop here.
- 03
Aggregate
A robust median anchors the estimate, so a minority of faulty or manipulated sources cannot pull it far. Weights reflect each source’s liquidity, reliability and recency.
- 04
Confidence
The freshness of the inputs and their dispersion produce a confidence score. Close agreement among fresh sources scores high; stale inputs or wide disagreement score low.
- 05
Publish
Value, confidence and provenance are signed and published as market state, read inside the same deterministic execution as every transaction that depends on them.
Architecture
Components of the mesh
Prism is built as a chain of narrow responsibilities, so each can be monitored, tested and challenged on its own.
- 01
Source registry
Records every source, the instruments it covers, its methodology and its observed reliability. Sources are admitted, weighted and retired under defined rules.
- 02
Normalization layer
Aligns timestamps, units, contract multipliers and quote conventions, so that a benchmark fixing and a dealer quote describe the same quantity before they are compared.
- 03
Robust aggregator
Computes a median-anchored estimate weighted for liquidity and reliability, bounding the influence any minority of sources can have on the result.
- 04
Confidence engine
Scores each value from the age of its inputs and their dispersion, and publishes that score beside the price rather than hiding it inside the number.
- 05
Signed publication
Publishes each value as signed market state with its provenance, so consumers and reviewers can trace any number back to the observations that produced it.
- 06
Consumer policies
Each consuming module declares how it responds to confidence: margins can widen, price bands can tighten, matching can pause and liquidation can switch to protective rules.
Security and failure control
Designed for the day a source goes wrong
Oracle failures are rarely dramatic. A feed goes quiet, a venue prints an off-market trade, a benchmark arrives late. Prism is engineered to notice these conditions and to make them visible to every system that relies on a price.
- Source diversityIndependent source types — exchanges, dealers and benchmarks — reduce the chance that one outage or one manipulated venue determines the result.
- Outlier resistanceThe robust median sets extreme observations aside by construction. Moving the published value requires corrupting a majority of the weighted sources, not one of them.
- Staleness detectionEvery input carries its age. Values built from stale inputs lose confidence automatically instead of silently persisting.
- Confidence circuit breakersWhen confidence falls below a market’s threshold, dependent actions change mode: price bands tighten, matching can pause and liquidation logic applies protective rules instead of acting on a single uncertain print.
- Surveillance feedbackDivergence between Prism values and trading on the network is monitored as a market-surveillance signal and flagged for review.
Across the three systems
One reference truth for three systems
Orbitra L1
Settlement and compute
Prism is a market-services module of Orbitra L1. Its values are committed as state, read deterministically on VectorLanes and finalized under QSE together with the transitions that use them. Assets from other chains enter on a separate path, through GateMesh.
Orbitra Realm
Applications and commerce
Applications in Orbitra Realm read the same signed market state through NexusSDK — a lending application checking collateral, a payment converting between currencies or a tokenized fund publishing its valuation.
Value
What confidence changes
- Traders and users
- Margin, funding and liquidation that respond to reliable prices rather than a single anomalous print. The confidence behind a mark is shown, not assumed.
- Institutions
- Traceable valuation. Each reference price links to its sources, method and confidence, supporting risk oversight, valuation processes and dispute resolution. Data use follows the market-data terms.
- Developers
- One interface for prices with provenance and confidence built in. Contracts on NexusWASM can require a minimum confidence before acting instead of building their own oracle defenses.
Specifications
Specifications
- Pipeline
- Sources → Normalize → Aggregate → Confidence → Publish
- Source types
- Exchanges, dealers and benchmark providers
- Normalization
- Common clock, units and quote conventions, with quality tagging
- Aggregation
- Robust median with liquidity and reliability weighting
- Confidence inputs
- Freshness of inputs and dispersion across sources
- Output
- Signed market state: value, confidence score and provenance
- Failure response
- Confidence-driven circuit breakers configured per market
Terminology
Terminology
- Robust median
- A central estimate that stays stable when a minority of inputs are extreme or wrong.
- Dispersion
- How widely sources disagree about the same value at the same moment.
- Freshness
- The age of the observations behind a published value.
- Confidence score
- A published measure of how far a value can be relied upon, derived from freshness and dispersion.
- Provenance
- The record linking a published value to the sources and processing steps that produced it.
- Mark price
- The reference used to value open positions and margin, taken from Prism rather than from the last trade alone.