scoring
Weighted average over a fixed-length score vector.
- input
- ScoringInput { scores[64], weights[64], count }
- output
- ScoringOutput { weighted_avg_fp, total_weight }
- use
- Credit / reputation aggregation.
assembling
assembling
Massive computation, compressed into a single proof. Verified on Solana mainnet.
BPF compute units cap each transaction. A model inference, an oracle aggregation, a leaderboard sort — all of them already overshoot. The chain refuses the work, so the dApp pretends it never asked.
SP1 generates a proof per sub-computation. Nova folds them — recursively, instance by instance — into a single SNARK. The Anchor verifier on mainnet checks the folded proof in under 200K CU.
Drop-in templates. Bring your own host or use ours. The folding adapter is identical across all five.
Weighted average over a fixed-length score vector.
SUM, AVG, MIN, MAX in one pass, all four committed.
Median with a sortedness witness over the input multiset.
Permutation proof — output is monotonic, multiset matches input.
Two-layer MLP forward pass, ReLU, fixed-point i32.
Add a guest program under packages/circuits/<name>, then register the builder entry. SDK + verifier auto-pick it up.
Network-native unit of account. Used for compute submission, marketplace settlement, and verifier prioritization.
The SDK is one npm install. The CLI is one Homebrew tap. The circuits are open from day one.