We are building a resource coordination protocol for AI devs, agents, and physical AI to discover, negotiate, commit, and settle on data, infra, and capital, on-demand.
For the data exchange, we have developed a design for a solver node that anyone can run to satisfy incoming intents termed as “Request for Data” i.e. RFDs. - www.reppo.exchange
The design of the solver node that will operate on the exchange is available here - solver-node/README.md at main · Reppo-Labs/solver-node · GitHub
As you will notice, we are not explicity using juvix here. That’s the next step. We first wanted to show a simple PoC to showcase the functionality.
Currently, the node is designed to call models on Opengradient’s model hub and generate synthetic data to satisfy RFDs.
In the future, we expect anyone to be able to customize the solver node, optimizing for a certain class of RFDs and similarly, we are also planning to build an infra.exchange, which will leverage primciples of intent-centricity to matchmake AI builders and agents with infra providers, on-demand.
I’m curious to know if it makes more sense to contrain the formation of RFDs i.e. the intent in a specific format for them to be standardized for consumption by solver nodes or does it make more sense to leverage Anoma’s tech solely in how solver’s consume and solve for RFDs.