1inch-U91: add cronos, monad, hyperevm to the 1inch lineage (AR, LO) + fix robinhood mode classification - #9969
1inch-U91: add cronos, monad, hyperevm to the 1inch lineage (AR, LO) + fix robinhood mode classification#9969dodgervl wants to merge 8 commits into
Conversation
The robinhood cfg listed the canonical multichain settlement addresses (SettlementV3 0xabd4..., SettlementV3v1 0x2ad5...), but neither is deployed on robinhood. The chain uses its own settlement deployment at 0xb55ba9617dafae1236313c3cb7806439ceefbd13 (deployed 2026-06-10, same batch as the access tokens), so no LO fill ever matched and all fusion swaps were flagged as mode='limits'.
…oss-chain v1.2)
Adds the august 2026 deployment wave to the 1inch lineage:
- meta cfg: chain configs (cronos uses base_transfers + clone runtime code at offset 11; monad/hyperevm standard) with chain-specific SimpleSettlement as fusion settlement and both cross-chain factories for LO arg detection; access token overrides (non-vanity addresses on these chains)
- AR/LO cfg: per-chain AggregationRouterV6 macros (hyperevm has a chain-specific router address 0x5281602a...)
- CC cfg: v1_2 method set (new selectors, immutables shifted by the dynamic `parameters` field) and a shared v1.2 raw-parsing contracts macro: factory on decoded tables, escrow clones resolved from creation traces, dst escrow address computed with the v1.2 CREATE2 salt (8 fields + keccak(parameters)), validated against all on-chain deployments
- 12 models + schema per chain (sonic/robinhood templates), sources for decoded ARV6/EscrowFactoryV1_2/access-token tables
- parent schema: 12 union model blockchain lists + lop_venue_settled_fills/lop_own_trades entries per chain
- dex_{cronos,monad,hyperevm}_trades: union in the 1inch-LOP passthrough
…+ LO only) The CC escrow decodings will be handled separately, so this PR ships classic/fusion/limits only: - exposed=["ar", "lo"] for the 3 chains (fantom precedent); cc union models drop them via the config loops - delete the 9 per-chain cc models (cc, cc_executions, cc_raw_calls) + their schema entries - transfers models keep ar+lo streams only - revert the v1.2 methods/contracts config (oneinch_cc_contracts_cfg_macro.sql is back to upstream state) - drop the now-unused creations_parent_code_offset from the 3 chain cfgs (robinhood keeps 21) - drop EscrowFactoryV1_2 source declarations escrow_factory_addresses stay in the meta cfg (like fantom): they drive LO factory_in_args detection, so cross-chain fills keep mode='cross-chain' in swaps instead of polluting limits, and the LO-side rows won't change when the CC iteration lands. Also add robinhood to the oneinch_cc_executions schema blockchain list (missed in duneanalytics#9918; the model unions robinhood via the config loop since robinhood is cc-exposed).
…ement Robinhood deployed a second escrow factory line (0x50d26ea1..., live since 2026-07-15, now dominant: 862 src escrows vs 209 on v1). Without it in escrow_factory_addresses all its cross-chain fills classify as limits. Also moves a Jinja comment out of the return([...]) expression in oneinch_blockchains_cfg_macro - comments are invalid inside expressions and broke compilation of every oneinch model.
Clamp every oneinch model's date_from to 2026-08-15 via oneinch_easy_date() so CI scans ~10 days instead of full history (CI rebuilds all oneinch models because shared macros changed). New chains lose nothing (first activity 2026-08-17). Same pattern as duneanalytics#9918.
github-actions
Bot
added
WIP
github-actions
Bot
added
ready-for-review
PR SummaryMedium Risk Overview Robinhood gets config-only fixes: the real fusion SimpleSettlement address and the dominant cross-chain v1.2 escrow factory so LO fills classify as fusion/cross-chain instead of limits; Shared macros clamp Reviewed by Cursor Bugbot for commit 7c14288. Configure here. |
![cursor[bot]](https://avatars.githubusercontent.com/in/1210556?s=60&v=4){kind=small}
There was a problem hiding this comment.
Cursor Bugbot has reviewed your changes using default effort and found 1 potential issue.
Bugbot Autofix is ON, but it could not run because the branch was deleted or merged before autofix could start.
Want higher recall? High effort reviews run extra passes and find more bugs. A team admin can switch effort levels in the Cursor dashboard.
Comment @cursor review or bugbot run to trigger another review on this PR
Reviewed by Cursor Bugbot for commit 7c14288. Configure here.
| -- Production tables must be built with full history: revert before merge + full refresh as needed. | ||
| {% macro oneinch_easy_date() %} | ||
| {{ return("2026-08-15") }} | ||
| {% endmacro %} |
There was a problem hiding this comment.
CI date clamp left in macros
High Severity
oneinch_easy_date() still returns 2026-08-15 and is folded into every 1inch date_from. The file itself is marked as a temporary CI clamp to revert before merge. If this ships, full refreshes of oneinch_swaps and the other shared models drop all history before that date across every chain.
Additional Locations (2)
Triggered by learned rule: Flag hardcoded date literals in model SQL that look like development filters
Reviewed by Cursor Bugbot for commit 7c14288. Configure here.
… Dune decoding cleanup)
… 1inch-updates-91 # Please enter a commit message to explain why this merge is necessary, # especially if it merges an updated upstream into a topic branch. # # Lines starting with '#' will be ignored, and an empty message aborts # the commit.
1 participant
Description:
Adds cronos, monad and hyperevm to the main 1inch lineage: AggregationRouterV6
(AR) and Limit Order Protocol (LO), following the standard per-chain pattern — 9 models per
chain (raw_calls / stream / executions / transfers / LOP trades), meta config entries
(exposed
["ar", "lo"]like fantom, start2026-08-01), source declarations, and the LOPpassthrough wired into
dex_<chain>_trades. Cross-chain (CC) models are deferred untilthe escrow decodings are sorted out separately.
The intent stack on these chains is deployed at non-canonical addresses shared by all
three (same deployer): settlement
0x65497e56...and the Limits/Fusion/CrossChainaccess tokens — not the canonical
0xacce55...vanity set. The router is the canonical0x1111111254...on cronos and monad; hyperevm has a chain-specific router(
0x5281602a...). Cronos usesbase_transfers(notransfers_from_tracestable),same handling as robinhood.
Also carries two config-only robinhood mode-classification fixes:
not deployed on robinhood — so
oneinch.swapsshows 0 fusion there. The realchain-specific settlement
0xb55ba9617dafae1236313c3cb7806439ceefbd13matches ~36.9kof 48.7k router fill calls currently misflagged as
mode='limits'.(
0x50d26ea1..., 862 src escrows vs 209 on v1) was missing fromescrow_factory_addresses, so its fills also sit inlimits. With the factory addedthey are cc-flagged and excluded from the LO stream (cc-flagged fills never enter
oneinch.swapsfrom the LO side;mode='cross-chain'rows come only from the CCstream, which stays v1-only until the CC iteration). Same story for the new chains'
cross-chain fills: excluded now, surfacing as
mode='cross-chain'when CC lands.Also adds robinhood to the
oneinch_cc_executionsschema blockchain list (missed in #9918).Validation
cross-chain-flagged (209 v1 + 862 v1.2) / 12,948 limits; new chains classify cleanly
(classic/fusion/limits), zero unclassified calls.
oneinch_swapsreconciles exactly againstprod — robinhood prod limits 7,462 = 4,630 fusion misflags + 2,556 plain + 276 v1.2
cc-flagged; cross-chain 43 = 43 (CC stream unchanged); robinhood CC output matches
prod in the same window — all rows trace to the v1 factory, zero garbage rows from
the v1.2 line.
review, so the branch head builds full history.
Post-merge
Full refresh needed (history changes that merge can't repair):
oneinch_swaps,oneinch_intent_accounts,oneinch_robinhood_lo,oneinch_robinhood_lo_executions,oneinch_robinhood_lop_venue_settled_fills,dex_robinhood_trades, and initial buildsof the 27 new models +
dex_{cronos,monad,hyperevm}_trades. Noteoneinch_blockchains,oneinch_intent_resolversandoneinch_evms_ptfciterate the blockchain config withoutbeing graph descendants of the per-chain models, so they need explicit selection (all
cheap). As on #9918,
state:modified.macros+over-selects unrelated chains.