This fork of CloudX's OpenAuction adds optional embedding-space scoring to the core auction:
score = log(price) - distance² / σ²
Bids carry three optional fields — embedding, embedding_model, sigma — passed to RunAuction as a variadic query embedding. Bids without embeddings fall back to pure price ranking. Existing callers compile unchanged.
For context: An Open Letter to CloudX · Power Diagrams for Ad Auctions (series start)
cmd/simulate/ runs a multi-agent market simulation validating that relocation fees (λ · ||c_new - c_old||²) stabilize advertiser positions in embedding space. See It Costs Money to Move.
go run ./cmd/simulate/See the upstream README for build, test, and enclave documentation.
go test ./...