Exposure Notification API

[adamfowleruk]

Built on top of the Analysis API, in order to provide and hit thresholds, and perform some action on hitting them.

[adamfowleruk]

Implement Risk Scoring

• Provide aggregation of a SampleList by a time interval (1 Day by default)
• Maintain a History of these Risk aggregates by date
• Allow retrieval of risk aggregates by date
• Allow merging of multiple risk aggregate details (E.g. one local, one from a remote paired device)
• Allow a Risk summary to execute over these histories as they are updated
• Link a Risk Summary calculator to a particular Disease
• A Risk Summary calculator could have one or more named Diseases associated with them
• Provide a Risk score updated callback
• Provide the ability to save risk scores to disk, and read them again later

Implement Exposure Notification

• For each disease, provide a set of risk score thresholds, and callbacks for increasedTo and decreasedTo
• Ensure callbacks are not fired whilst history merging events are happening
• Allow specification of rules such that the status cannot downgrade until a particular date time (prevents multiple notifications per day)
• For each threshold, link the level to a description, summary, and current advice summary text and URLs
• Ensure multiple languages are supported, with a set of language defaults
• Allow these thresholds for different diseases to share the same impact score, so a person's advisories can be aggregated across diseases/strains

[adamfowleruk]

Need to split the current ExposureManager into ExposureRecorder (live RSSIMinute data for 'today') and ExposureStatusManager for confirming exposure sources (SampleIDs) as ill or not, and calculating the aggregated risk score for each source, and then calling the ExposureManager with updates for the risk score for each Agent(Disease). At the moment we're assuming everyone encountered is ill!

[adamfowleruk]

v2.1.0-beta2 released with the latest changes from this branch merged into develop.

De-prioritising non-required functionality for this release. Remaining TODOs therefore:-

Risk and Exposure stores:-

• Add aggregate() function to RiskStore as exists inside ExposureStore (to allow retrieval by date, for example)
• Add Exposure Manager addExposure() function call with different source but same everything else to allow merging of risk scores shared from another device (E.g. wearable to phone, or vice versa)
• Add Zephyr external Flash disc Exposure store and Risk store
• Add removeOlderThan function to both exposure manager and risk manager stores
• Move index position based functions to iterator based functions in risk and exposure managers
• Support saving and reading of more static Parameters to/from flash too (E.g. age, phylogenic sex, weight et al)

Exposure Notification:-

• Add convert function to ExposureManager allowing conversion from humanProximity to (for example) Covid19Exposure (Derived Exposure) -> Consider doing this ephemerally based on stored source exposure types (i.e. 1:1 mapping)
• Ensure this also works for rescind or modification (E.g. changing Confidence value, or lowering/increasing exposure score based on some external status notification change) - Note: Network/protocol mechanisms out of scope (will be part of OpenTraceV2 in a future release)

Risk Threshold Notification:-

• Apply concept of Levels, states, alerts, and rules from old ExposureManager to a new RiskThresholdManager class
• Allow the config of this to be stored, and loaded at runtime from flash storage

Out of scope:-

• X Risk threshold / notification status messages, urls (more of a UI thing in the Java and iOS API)
• X Storing to/from csv on windows/mac/linux (for now)
• X CLIs for running risk models over csv stored data
• X Any network comms or bluetooth comms for sync / notification at this stage (part of v2.2 of C++ and OpenTraceV2)