[danielgorgonha]

What & why

Closes #949.

On the web, getProposalDetail(id) reads the vault's whole proposal list and picks one proposal from it, with no cache. The Ledger called it once per payroll in three places: the book by beneficiary, opening a payroll row, and the export. On a vault with many payrolls that is N reads of the whole list in sequence.

readPayrollLines(ids) (new, ui/src/payroll-lines.ts) gives the Ledger the lines of the payrolls it asks for:

• on the web, from ONE read through the existing getBookWithLines (no second "book with lines" reader);
• on the local bridge, one detail read per payroll, as before, and with no cap (getBookWithLines stops at the newest 12 settled payrolls on the bridge, which the Ledger cannot afford).

What the Ledger shows is unchanged: the book by beneficiary keeps only payrolls with lines and does not ask again until reopened; a payroll row that answered no lines shows nothing (not "loading"); the export uses an empty list for a payroll that did not answer. One difference on the bridge: the book by beneficiary now stores the payrolls' lines when all are read, rather than one by one; the final state is the same.

Tests:

• ui/src/payroll-lines.dom.test.ts: through the real reader with a counted fetch: 20 payrolls on the web = 1 fetch; 15 on the bridge = 15 detail reads; an empty payroll answers [], an unreadable book answers nothing; a source pin that the Ledger no longer calls getProposalDetail.
• ui/src/ledger-reads.dom.test.tsx: the real Ledger: on the web, the book by beneficiary and the export each add exactly one read of the proposals (against the previous Ledger: 21); on the bridge, 20 detail reads each, past the cap.

Follow-up worth an issue: the bridge's per-id loop now exists in getBookWithLines (capped, for the proposal screen's note) and in readPayrollLines (uncapped); moving readPayrollLines into api.ts and having getBookWithLines use it would leave one loop.

Verify-before-finishing checklist

• cargo fmt --check + cargo clippy -- -D warnings: not applicable (no Rust change)
• cargo test: not applicable
• ui/: lint + vitest (263 files, 3034 tests) + tsc pass
• No secret, key share, seed, or passphrase in code, logs, URLs, or query strings
• Every fund-moving path keeps preview + explicit confirmation (none touched)
• FROST signer, PCZT, loopback bridge: not touched
• User-facing strings: none added
• Docs/comments in English; commits are type(scope): description
• AI disclosure: written with Claude Code (the change, the tests, and the review and test runs), reviewed by a code audit and a test audit; the commits carry Co-Authored-By: Claude <noreply@anthropic.com>, no session link

[vercel]

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

Project	Deployment	Actions	Updated
konclave-demo	Ready	Preview	Nov 13, 2026 11:38am UTC