Changelog

0.4.0: May 2026

Cross-chain SwapVM swap with multi-LP partition repayment proven end-to-end.

Cross-chain SwapVM ship can now partition each token across multiple chain-A source LPs. The brain's prepareCrossChainSwapVMCommitment flow does a greedy fill in caller-supplied LP order, falling through when capped by per-LP availability. When sourceLps is omitted, the brain auto-picks chain-A LPs by available capital.
Live E2E (2026-05-07): 1 mUSDC traded for 0.995 mDAI on Base Sepolia against a strategy shipped on Base, with the repayment leg bridged from chain A to the destination filler on chain B via LayerZero V2 OFT. Both bridge legs settled on chain (200.2 mUSDC + 200.2 mDAI delivered to the filler).
Filler fee (JIT_FILLER_FEE_BPS, default 10 bps) is computed per partition rather than per strategy, so a 200-unit ship split 100 / 100 across two source LPs produces two 100.1-unit obligations.
Fixed a load-bearing arg-order bug in resolveEquivalentToken (cross-chain asset mapping) that surfaced when the 2026-05-01 testnet redeploy left mUSDC.84532 and mWBTC.1301 at the same address.

Dumb-vault rollback. SLP simplified to custody and signature plumbing only.

Removed on-chain per-LP balance tracking, per-strategy LP commitments, aggregate committed-balance accounting, and bridge delivery attribution from the SLP. All of that now lives in the brain DB.
Deleted the SwapVM attribution hook (Aqua0SwapVMHook). SwapVM PnL is now distributed off-chain by the brain reading SwapVMRouter.Swapped against swap_strategy_commitments.
Deleted the intent stack (IntentRouter, SettlementContract, EscrowModule, SourceLock, FillerRegistry) and the legacy LP smart account (lp/Account*.sol). LPs now interact with the SLP directly using their EOA or 4337 wallet; the SLP uses msg.sender as the LP identity.
Deleted the pre-LayerZero bridge adapters (Composer, CCTPAdapter, CCTPComposer, StargateAdapter, BridgeRegistry). Cross-chain repayment now goes through slp.withdrawForRepayment plus a LayerZero OFT bridge, with tokens landing directly in the destination SLP's balance.
SLP authorization model collapsed to three single-slot privileged identities: backendSigner, repaymentWorker, operator. No AccessControl role mapping.
shipStrategy / dockStrategy / reshipStrategy are now operator-gated and commitment-less on chain (the brain pre-validates aggregate commitments off-chain before signing).

V4 JIT cross-chain swap loop landing. LayerZero V2 OFT adapters deployed.

Backend signs an EIP-712 JITPayload. Trader submits a Uniswap V4 swap with that payload as hookData. The hook validates against the SLP's backendSigner, injects a temporary range, the swap executes, the hook removes the range and emits SwapSettled.
Repayment-worker drains the source LP via slp.withdrawForRepayment against a backend-signed RepaymentWithdrawPayload, then dispatches the LayerZero OFT message; tokens land directly in the destination SLP via OFT delivery (no Composer hop).
LayerZero V2 OFT adapters deployed and peered between Base Sepolia (84532) and Unichain Sepolia (1301) for mUSDC, mDAI, mWBTC, mWETH (MockTokenOFTAdapter on testnet).
Brain refactor: read source data from Ponder rather than running its own RPC getLogs cursor. The SwapSettled poller still drives PnL overlay updates, but reads from slp_swap_settled via a 4-second cached read-through.

Single-chain SwapVM swap proven end-to-end on Base Sepolia.

LP deposit, operator-shipped strategy via slp.shipStrategy, 1inch Aqua registers the strategy in its 4D virtual ledger, trader hits AquaSwapVMRouter.swap, router emits Swapped.
Brain attribution (swapvm-attribution.service.ts) consumes SwapVMRouter.Swapped and pro-rata-distributes deltas across swap_strategy_commitments rows.
E2E driver (scripts/e2e-aqua-swap.ts) passes: 1 mUSDC traded for 0.99 mDAI with attribution credited to LP overlays.
Initial Base Sepolia and Unichain Sepolia deployments of the SLP, V4 hook, and 1inch Aqua infrastructure. Public docs scaffolding shipped.