feat(config): chain registry for multi-EVM support

[fielding]

Summary

First PR in the multi-EVM rollout (epic RG-7ce12d). Introduces a Record<chainId, ChainConfig> registry at packages/app/src/config/chains.ts seeded with Base (8453) and Base Sepolia (84532), and rewrites config/contracts.ts as a thin compatibility shim that reads every existing singleton export from getChainConfig(DEFAULT_CHAIN_ID).

Follow-up PRs (in order):

• RG-9430aa + RG-42e939 — switch /create and /vaults to useChainId() + getChainConfig(), thread usdcDecimals through the approve/lock flow, delete the singletons.
• RG-b53952 — add the other 6 target chains to the wagmi config, wrong-chain guard on /create.
• RG-213ce7 — landing chain chip row.
• RG-644a80 — per-chain treasury Safes, smoke tests (Arbitrum first for cheap gas, BNB for 18-decimal USDC validation).

Why a compatibility shim instead of refactoring call sites in this PR

Production is live on ripguard.xyz with real users and real broker-fee revenue. Splitting the registry introduction from the call-site refactor means this PR is provably behavior-preserving on Base mainnet — every singleton import resolves to the same address, block number, and chunk size as before. The next PR can then be read purely as a find-and-replace across create/vaults without ambiguity about what data changed.

Design notes

• usdcDecimals is in the ChainConfig shape from day one even though no consumer reads it yet. BNB's USDC being 18 decimals (vs 6 on every other target chain) is the single most likely regression source during the rollout; putting the field in the registry now avoids a second schema churn when the page refactor lands.
• Env overrides (NEXT_PUBLIC_SABLIER_LOCKUP, NEXT_PUBLIC_USDC_ADDRESS, NEXT_PUBLIC_TREASURY_ADDRESS) still work, but only for the DEFAULT_CHAIN_ID (whichever chain the current deployment is targeting). Preserves today's testnet staging flexibility.
• BROKER_FEE stays global (0.5% everywhere). Zero-address treasury still disables the fee, matching Sablier's broker-field semantics.
• PRESETS stays global. Schedules aren't chain-scoped.

Test plan

• pnpm --filter app test — 63 tests pass (was 56; 7 new registry tests)
• pnpm --filter app exec tsc --noEmit — clean
• NODE_ENV=production pnpm --filter app build — clean, 9 routes emit
• Manual smoke: load ripguard.xyz locally against Base mainnet, verify /, /create (connect wallet), /vaults, OG image, share card all render identically to pre-PR
• Vercel preview deploy — walk the same pages against the preview URL before merging

🤖 Generated with Claude Code

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
ripguard	Ready	Preview, Comment	Apr 25, 2026 2:07am
ripguard-testnet	Ready	Preview, Comment	Apr 25, 2026 2:07am

[fielding]

Looks good to me. No blockers.

I checked the registry and compatibility shim path, and it preserves the current Base/Base Sepolia singleton behavior. Keeping call sites unchanged in this PR makes sense, especially with prod already live.

A few small follow-ups, none blocking:

• Add treasury to the registry address sanity test. sablierLockup and usdc are checked today, and treasury matters because it decides whether broker fees work or are disabled. The zero address still passes the same address regex.

• Avoid mutating SUPPORTED_CHAIN_IDS in the test:

  expect([...SUPPORTED_CHAIN_IDS].sort()).toEqual(Object.keys(CHAINS).map(Number).sort());

• In the call-site refactor, make the broker fee derive from the active chain treasury. The current default-chain BROKER_FEE shim is fine for this PR, but the next step should avoid passing a zero treasury with a non-zero fee to Sablier.

Validation I ran locally:

pnpm --filter app test -- src/config/chains.test.ts src/config/contracts.test.ts
pnpm --filter app exec tsc --noEmit
NODE_ENV=production pnpm --filter app build
NEXT_PUBLIC_CHAIN=base-sepolia pnpm --filter app test -- src/config/chains.test.ts src/config/contracts.test.ts

All passed.

[fielding]

Reviewer orientation — multi-EVM stack [1/6]

This PR is the bottom of a 6-PR stack that ships full multi-EVM support. Each PR builds on the prior; review in chronological order.

	PR	Branch	What it does
→ 1	#8	feat/chain-registry	registry foundation (you are here)
2	#11	feat/page-refactor	switch pages to useChainId, thread USDC decimals
3	#12	feat/wagmi-multichain	registry-driven wagmi config + wrong-chain guard
4	#13	feat/expand-chains	add 6 new chains (Eth/Arb/Op/Poly/Avax/BNB) + BNB disclosure
5	#14	feat/landing-chain-chips	landing chain chip row
6	#15	feat/balances-script	`pnpm balances` CLI for treasury reads

Start here. When done, move to #11.

Merge order is bottom → top after smoke testing. Smoke test checklist lives at `.handoff/SMOKE_TEST_MULTI_EVM.md` (in the user's local notes; not committed). Test against #15's Vercel preview since it sits on top of the full stack.

[fielding]

Re-reviewed after the latest update. The two test nits from my earlier pass are fixed:

• treasury is covered by the registry address sanity test.
• SUPPORTED_CHAIN_IDS is copied before sorting, so the exported array is not mutated.

No new findings on this layer. The registry foundation and compatibility shim still look behavior-preserving for Base and Base Sepolia.