Leonard review — PR #624

Verdict: COMMENT — no blocking findings found on head 80f23665f8aa9ebd291d3ac2e7a5842f9400dfcf. I’m leaving this as a comment rather than approval because approval state was not requested.

Scope reviewed:

• Grouped collateral exposure/open-interest utilities and tests.
• Shared useCollateralOpenInterest() loader, stale-chain handling, and V3 proxy allowlist for /v3/evk/vaults/open-interest/by-collateral.
• Lend vault grouped Collateral exposure UI and borrow-pair Open interest UI on the PR Railway preview.
• Scalability / maintainability hygiene pass across sibling lend/earn/borrow exposure surfaces.
• Existing bot/reviewer feedback. CodeRabbit’s visible top-level note says auto-review was skipped; I did not find a material active bot claim to carry forward.

Validation run:

• npm run test:run -- tests/utils/open-interest.test.ts tests/utils/vault/collateral-exposure.test.ts tests/composables/useCollateralOpenInterest.test.ts tests/server/v3-proxy.test.ts — passed, 30 tests.
• npm run typecheck — passed.
• npm run build — passed, warnings only.
• Browser visual smoke on the PR preview:
  • Desktop /lend/0xba98fC35C9dfd69178AD5dcE9FA29c64554783b5?network=1: expanded Collateral exposure for Sentora PYUSD; live exposure rows rendered with Max LTV / Liquidation LTV and collapsed extra row control.
  • Mobile same route: opened Vault information and expanded Collateral exposure; live exposure rows rendered in the narrow modal layout.
  • Desktop /borrow/0xba98fC35C9dfd69178AD5dcE9FA29c64554783b5/0xAB2726DAf820Aa9270D14Db9B18c8d187cbF2f30?network=1: Open interest surface loaded; by-collateral open-interest API returned 200.

Visual evidence:

• Desktop lend vault Collateral exposure: https://gist.githubusercontent.com/LeonardEulerXYZ/990bfbf0cb45c5fa1aa33938d02fd22d/raw/3df81580cbeafca5944dd505519f47d5f845b564/euler-xyz-euler-lite-pr624-pr624-desktop-lend-pyusd-collateral.png
• Mobile lend vault Collateral exposure: https://gist.githubusercontent.com/LeonardEulerXYZ/990bfbf0cb45c5fa1aa33938d02fd22d/raw/5189c2acb718366e894b4905b403b9bfabcf0099/euler-xyz-euler-lite-pr624-pr624-mobile-lend-pyusd-collateral.png
• Desktop borrow pair Open interest: https://gist.githubusercontent.com/LeonardEulerXYZ/990bfbf0cb45c5fa1aa33938d02fd22d/raw/f308d55366d91714acc91a0be79daf8c2f972ffe/euler-xyz-euler-lite-pr624-pr624-desktop-borrow-pyusd-usdc-open-interest.png

Review notes:

• The two earlier Leonard blockers are resolved in this head: the shared loader now guards stale overlapping chain responses, and the borrow-pair Open interest block uses the shared collateral open-interest path rather than maintaining a duplicate fetch/cache path.
• The grouped exposure abstraction is in shared utilities with focused unit coverage rather than being embedded as one-off component logic. I checked sibling lend-row, lend-vault, earn-row/earn-vault, and borrow-pair exposure/open-interest consumers; the remaining split is intentional presentation shape, not duplicated business logic.
• No wallet/signing path was exercised. This pass covered read-only route/API/visual behavior for the grouped exposure surfaces.

Leonard review — PR #624

Verdict: REQUEST_CHANGES on head 04e01cf3144457e78a01b69fb56a26d228bab828.

Thanks for the latest pass. The two previous blockers appear resolved: useCollateralOpenInterest now ignores stale overlapping chain loads, and the vault overview open-interest section now uses the shared composable rather than its own direct fetch path.

I found one remaining consistency blocker in the newly-added stats-matrix exposure surface: when a vault is utilized but the API lacks the collateral split, the matrix silently drops that vault from exposureItemsByVault while the global exposure state remains ready. That renders the stats cell as - rather than Unavailable, which makes “data missing” look like “no exposure” for exactly the high-utilization vaults where this feature is most useful. I left the concrete note inline.

Validation performed:

• npm run test:run -- tests/composables/useCollateralOpenInterest.test.ts tests/utils/vault/exposure-display.test.ts tests/utils/vault/collateral-exposure.test.ts ✅
• npm run typecheck ✅
• npm run build ✅
• Targeted ESLint over PR-changed source/test files ✅
• Full npm run lint is still blocked by existing repository-wide lint debt outside this PR; targeted lint for the changed files passed.
• Railway preview browser visual smoke at /lend?network=1 on desktop and mobile ✅: the route loaded real content, the new Exposure column/rows were visible, and no material page error state was observed.

Screenshot evidence:

• Desktop lend list, exposure column visible: https://gist.githubusercontent.com/LeonardEulerXYZ/423903624e3a1db461124cb483167a88/raw/a2beec46b7c9f4f45ed71d8d9b9c69dc49d47016/euler-xyz-euler-lite-pr624-pr624-lend-desktop.png
• Mobile lend list, exposure row visible: https://gist.githubusercontent.com/LeonardEulerXYZ/423903624e3a1db461124cb483167a88/raw/3a70f7816b4f1d2e69718119ee9fead994c46036/euler-xyz-euler-lite-pr624-pr624-lend-mobile.png

Scalability / maintainability hygiene pass:

• Good: the open-interest loader and exposure display/grouping logic are now shared across the card, overview, earn, and modal surfaces, with focused tests around the reusable utilities and stale-load behavior.
• Remaining issue: the stats matrix adds a sibling consumer with its own per-vault omission behavior for missing collateral splits. That should share the same “missing utilized split => unavailable” display invariant as the other grouped-exposure surfaces, rather than degrading to an empty ready state.

Bot/reviewer feedback pass:

• I did not find active CodeRabbit findings to verify on the current head.
• Prior Leonard blockers were rechecked against the current diff and are no longer applicable; no Leonard-owned inline comments were present via the review-comments API to delete.

Smoke coverage: browser visual smoke + mobile smoke on the Railway preview; no wallet/signing smoke, since this PR changes read-only exposure presentation and proxy allowlisting, not transaction construction.

Leonard review — PR #624

Verdict: COMMENT — no blocking findings found on head 792e9e748f701863359466eeba03ce0f5d8ea8ef. I’m leaving this as a comment rather than approval because this reviewer request did not explicitly ask for approval state.

Compared against parent branch feature/lite-215-convert-vault-page-sections-to-accordion-style / PR #617. This pass was an incremental re-review of the child delta since my previous blocking review on 04e01cf3144457e78a01b69fb56a26d228bab828; the only new diff is components/entities/vault/discovery/DiscoveryMarketAttributeMatrix.vue.

What changed in this head

• The stats matrix now carries a per-vault exposure entry with both items and valueState.
• Missing USD input or missing utilized collateral split now records valueState: 'unavailable' for that vault instead of omitting it from the map.
• VaultExposureSummary now receives the per-vault state, and matrix cell data-value returns loading / unavailable when the exposure cell is not ready.

This resolves my previous blocker: a utilized vault with a missing collateral split should no longer degrade to an empty ready value (-) merely because the global open-interest loader is ready.

Validation performed

• git diff --check origin/feature/lite-215-convert-vault-page-sections-to-accordion-style...HEAD — passed.
• npm run test:run -- tests/composables/useCollateralOpenInterest.test.ts tests/utils/vault/exposure-display.test.ts tests/utils/vault/collateral-exposure.test.ts tests/server/v3-proxy.test.ts — passed, 32 tests.
• npm run typecheck — passed.
• Targeted ESLint over the changed stats-matrix/exposure files and related tests — passed.
• npm run build — passed; existing Vite/Rollup chunk/comment warnings only.
• Headed Chromium/Xvfb route smoke on the current PR Railway preview:
  • /explore?network=1, KPK/Securitize market expanded, Matrix → Stats selected.
  • Stats matrix rendered with the new Exposure column and exposure cells populated for live assets; no material page errors observed.
• Malicious/supply-chain scan over the incremental diff: no executable script/workflow/dependency/secret/exfiltration indicators found.

Review notes

• Prior Leonard blockers were rechecked against the current diff and are resolved.
• CodeRabbit’s visible comment is still a skipped-review status for the non-default base branch; I found no active CodeRabbit finding to verify.
• Wallet/signing coverage was not run; this PR remains read-only display/data aggregation work and does not modify transaction construction.

Smoke coverage: browser route/visual smoke on the Railway preview; no wallet/signing smoke.

Leonard review — PR #624

Verdict: COMMENT — no blocking findings found on head 243c4e456a47ced60ecfba3a8187f7f9c915aa0b. I’m leaving this as a comment rather than approval because approval state was not requested.

Compared against parent branch feature/lite-215-convert-vault-page-sections-to-accordion-style / PR #617. This pass focused on the current grouped exposure views, the latest Earn/exposure source behavior, and the previously-reviewed shared loader/matrix invariants.

What I checked

• PR metadata, current diff, surrounding code, and previous Leonard review artifacts.
• Existing Leonard comments/reviews and current review-comment API state.
• Bot/reviewer feedback: CodeRabbit’s visible status remains a skipped-review note for this non-default target; I found no active bot code finding to verify or carry forward.
• Scalability / maintainability hygiene pass across the sibling grouped-exposure consumers: Lend rows, Lend vault overview, Earn rows, Earn vault overview/stats/exposure, Borrow pair open-interest, and Explore matrix stats.

Validation performed

• npm run test:run -- tests/components/base/BaseExposureSourcesModal.test.ts tests/components/entities/vault/overview/VaultOverviewEarnBlockStats.test.ts tests/utils/vault/exposure-display.test.ts — passed for the matching focused suite that Vitest selected, 6 tests.
• npm run typecheck — passed.
• npm run build — passed, warnings only.
• Headed Chromium/Xvfb browser visual smoke on the current PR build:
  • Desktop /earn?network=1: expanded Euler Prime and opened the Exposure modal; grouped exposure rows and Markets listed links rendered.
  • Mobile /earn?network=1 at 390px width: opened the same Exposure modal; narrow layout kept exposure rows and source links visible without blank/loading/error state.
  • Desktop /explore?network=1: expanded the Securitize/KPK market, selected Matrix + Stats, and verified the Exposure column/cells render.

Visual evidence

I updated the existing Leonard screenshot-evidence comment with verified raw PNG links for this head:
#624 (comment)

Review notes

• Prior Leonard blockers were rechecked against the current diff and remain resolved: the shared collateral-open-interest loader has stale-load protection, and sibling consumers now use the shared exposure/open-interest path rather than the earlier duplicate fetch/cache shape.
• The grouped exposure abstraction remains mostly in shared utilities/composables with focused tests, which is the right direction for this surface.
• I left one non-blocking inline maintainability note on the repeated Earn strategy source helper. The current behavior appears safe enough for merge, but the helper is now duplicated across three Earn exposure components and the !marketKey branch silently drops an otherwise displayable source label. If that is intentional, fine; otherwise it is a tidy candidate for a shared helper/test.

Smoke coverage: browser visual smoke + mobile smoke; no wallet/signing smoke, since this PR changes read-only exposure presentation and proxy/data plumbing, not transaction construction.

Comment lifecycle: updated the existing Leonard screenshot-evidence comment for the current head; no Leonard-owned inline review comments were present to delete; submitted this fresh formal review for the current head.