feat: streaming pricing strategy

[tnrdd]

Superfluid streaming pricing strategy for Markee. Backers stream ETHx into a SuperApp leaderboard ranked by stream rate; existing v1.3 lump-sum boards migrate onto it in place.

• StreamingLeaderboard SuperApp + factory
• Top markee streams 62% to the beneficiary, losers refunded ~net-zero, 38% to per-funder RevNet settlement
• Batchable single-tx open, jail-safe callbacks, v1.3-compat reads
• Integration tests on a Base fork

[vercel]

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

Project	Deployment	Actions	Updated (UTC)
markee	Ready	Preview, Comment	Jul 10, 2026 5:28am

[gaston-review]

Gaston Review

Verdict: Approved

Score: ███████░░░ 7/10

Pull Request Summary

This PR introduces a Superfluid-based streaming pricing strategy for the Markee leaderboard. Instead of lump-sum payments, backers open CFA streams into a StreamingLeaderboard contract. The highest-aggregate-rate Markee holds #1 (62% to beneficiary, 38% accrues for per-backer RevNet settlement via a MasterChef accumulator). Losers are refunded via GDA pools at their aggregate rate (net ~zero). There's a legacy grandfather floor system so migrated lump-sum Markees hold #1 until a stream overtakes their decaying floor. The whole thing is deployed as EIP-1167 clones from a factory that handles SuperApp registration and global RevNet/fee config. Also includes a Markee.sol fix replacing tx.origin with msg.sender in events, a comprehensive fork test suite against Base mainnet, a deploy script, and Foundry build config.

Review Summary

🔵 Suggestion — minimumMonthlyRate not enforced on flow updates
onFlowCreated enforces rate * SECONDS_IN_MONTH >= minimumMonthlyRate, but onFlowUpdated does not. A backer can lower their rate below the minimum after creation. If this is intentional (allowing graceful wind-down without forcing an all-or-nothing close), the asymmetry deserves a comment. If it's an oversight, a sub-minimum update could let dust streams linger on the leaderboard indefinitely.

⚪ Nitpick — _clone helper duplicated between StreamingLeaderboard and StreamingLeaderboardFactory
The EIP-1167 _clone function is copy-pasted verbatim in both contracts. Not a blocker (each contract needs its own copy since clones can't share internal functions from a different contract), but worth noting if a shared library is ever introduced.

🔵 Suggestion — Contract pushing EIP-170 size limit
The via_ir = true in foundry.toml is needed to keep StreamingLeaderboard under the 24,576-byte deployment limit. This leaves very little room for future additions. Any new feature or fix will need to watch bytecode size carefully — consider whether some view functions (getTopMarkees, totalLeaderboardFunds) could be moved to a separate reader contract to free up space.

This is a well-crafted, complex piece of DeFi infrastructure. The Superfluid integration is thorough — jail-safety is carefully handled in termination callbacks, the MasterChef settlement accumulator is correctly implemented, and the deposit-based buffer collateral model is sound. The fork test suite is excellent: it covers auto-flip promotions, decay-driven claimTop pokes, liquidation scenarios, migration in/out, batch settlement with reverting recipients, and buffer solvency reconciliation. I have a couple of minor findings but nothing that blocks shipping.

📌 4 inline comments
🔵 Suggestion: 2 · ⚪ Nitpick: 2

---

_{🔍 Reviewed by Gaston}

[gaston-review]

Gaston Review

Verdict: Approved

Score: ████████░░ 8/10

Pull Request Summary

This PR introduces a Superfluid-based streaming pricing strategy for the Markee leaderboard. Backers open CFA streams into a StreamingLeaderboard SuperApp; the highest-aggregate-rate Markee holds #1 (62% forwarded to beneficiary, 38% accrues for per-backer RevNet settlement via a MasterChef accumulator). Losers are refunded via GDA pools. A legacy grandfather floor system protects migrated lump-sum Markees until stream rates overtake their decaying floor. The whole thing deploys as EIP-1167 clones from a factory that handles SuperApp registration and global RevNet/fee config. The latest push applies review fixes: minimumMonthlyRate enforcement on flow updates, implementation constructor locks, deposit surplus withdrawal, and comprehensive regression tests.

Review Summary

🔵 Suggestion — Aggregated platform fee pay is not try/catch-isolated
In _routeRevNetSettlement, per-backer RevNet pays are individually try/catch-isolated (a reverting recipient only restores its own claimable). But the final aggregated _revNetPay(terminal, projectId, feeReceiver, totalFee) is a bare call — if it reverts, the entire settle transaction reverts, undoing all the successful per-backer pays too. In practice this is unlikely (same terminal/project as the successful backer pays, and the fee receiver is governance-set), and the admin can recover by zeroing platformFeeReceiver. Still, wrapping the fee pay in a try/catch with a fallback (e.g., hold the fee as contract ETHx for a later admin sweep) would make settlement fully brick-proof.

The review-fix push cleanly addresses the prior round's suggestions — minimumMonthlyRate is now enforced symmetrically on creates and updates, both implementations are constructor-locked against hijacking, and deposit surplus is withdrawable while streaming. The regression test coverage is excellent, with targeted tests for every fix. The one minor concern I have is the un-isolated aggregated fee pay in settlement, but it's mitigated by admin control over the fee receiver. Production-ready.

📌 3 inline comments
🔵 Suggestion: 1 · ⚪ Nitpick: 2

---

_{🔍 Reviewed by Gaston}

[gaston-review]

Gaston Review

Verdict: Approved

Score: ████████░░ 8/10

Pull Request Summary

This PR introduces a Superfluid-based streaming pricing strategy for the Markee leaderboard. The latest push (2be5d72) addresses the one remaining suggestion from my prior review: the aggregated platform-fee pay in settle is now wrapped in try/catch with _rewrapAndCredit fallback, so a reverting fee receiver never bricks the settlement batch. The fee is credited as claimable[feeReceiver] and recoverable via a later settle([feeReceiver]) once the terminal is healthy.

All prior review items are now addressed. The fee isolation fix is clean — same pattern as the per-backer isolation, same _rewrapAndCredit fallback, same self-call try/catch frame. The test coverage for the new path (test_settle_feePayFailure_doesNotBrickBatch) verifies both the fallback credit and the recovery settle. The minor quirk that recovering the fee receiver's credit runs through _settleViaRevNet again (taking a fee-on-fee) is harmless — both portions route to the same feeReceiver via RevNet, so total value is preserved. This is production-ready.

---

_{🔍 Reviewed by Gaston}