Last quarter 2023 update HQB
The user experience of the toolkit is well under transition (see video below) with the code located at BentoBoxDS. The DIYhsToolkit was the first itteration of the total user experience and allowed for putting together all the data flows from the HOP (health oracle protocol). Much learning was gained. Forming queries to HOP from the BentoBoxDS or any third party API call needs to become much simplier. A new library the HOP-Query-Builder (HQB) is seen as a solution.
This library is used by beebee (the AI help agent) in BentoBoxDS. The goal is to use natural language parsing (even a local LLM) to extract numbers, source of data, upload files etc. and then the HQB will take the context input and prepare the Network Library Contracts, modules and reference contracts that are the inputs to SafeFlow. There are four Module Contracts required for a query into SafeFlow, A question, data (packaging), compute and visualisation. These Module contracts include reference contracts e.g. the charting library being used, the computation being applied and location of the data file/source and how it is structured. It is surprising how much information is required to identify a specific set of data. This is the role the network library contracts provide and when replicated across a network of peers we can provide guarantees on data interoperabilty. These contracts do not exist when setting up a new BentoBoxDS (or in the old toolkit), the first task is to replicate the Module and Reference Contracts the first peer on the network has authored or for a peer to build a new set of Module and Reference Contracts using the Network Library Application (note this is not integrated into BentoBoxDS as of time of writing UPDATE: in the future branch of the repo the library has been added and this video shows its use https://youtu.be/Ok_xzelSqWQ), it is DIYhsToolkit). The HQB will need to implement a call to the network to get the latest Module and Reference Contracts so that it can use those as templates to build new queries from or to join an existing set of contracts authored by another peers. A bundle of module and reference contracts is called a Network Experiment. A such a bundle is the query structure requried to be fed into SafeFlow.
SafeFlow is the ‘compute engine’ for the protocol and uses the information contained with the contracts to understand what work is required. All these proof of works are recorded in a hash and then a hash of all the stages is hashed to create a ledger entry along with the hash of the results output by SafeFlow. This means we can gurantee reproducing of results, either to use them again or as a basis for trust when performing machine learning peer to peer.
