Remove CoW AMMs by fleupold · Pull Request #4624 · cowprotocol/services

fleupold · 2026-07-13T14:59:41Z

Description

As we are stopping support for CoW AMMs, this PR removes all occurrences of the concept in our backend code base. It leaves the underlying protocol concept of "surplus capturing jit orders" (the idea that the protocol can define certain accounts, for which the solver can create arbitrary limit orders during the competition and execute them with surplus that counts towards their score), as this is relatively little overhead to maintain and potentially a useful concept down the line.

Note, I'm not intending to merge this PR right away (mainly wanted to get feedback and a feel if I'm missing anything). We still need to talk to solvers and the one remaining LP before we can execute on this.

Changes

Remove all CoW AMM related contract artefacts
Remove cow amm indexing logic in the autopilot (replaced with an infra configurable list of surplus capturing jit order owners)
Remove cow amm order creation logic in the driver
Drop cow amm DB table

How to test

CI

Related Issues

https://linear.app/cowswap/issue/BE-99/remove-cow-amm-support

github-actions · 2026-07-13T15:00:02Z

Reminder: Please update the DB Readme and comment whether migrations are reversible (include rollback scripts if applicable).

If creating new tables, update the tables list.
When adding a new index, consider using CREATE INDEX CONCURRENTLY for tables involved in the critical execution path.
For breaking changes, remember that during rollout k8s starts the new autopilot, runs the Flyway migration, and only then shuts down the old pod. That overlap means the previous version can still be processing requests on the migrated schema, so make it compatible first and ship the breaking DB change in the following release.

Caused by:

database/sql/V113__drop_cow_amms.sql

claude · 2026-07-13T15:00:29Z

Claude encountered an error after 4m 1s —— View job

PR Review: Remove CoW AMMs

Read the PR diff against origin/main
Analyze changes for correctness & completeness
Post inline review comments
Post summary

The removal looks clean and consistent: the surplus-capturing JIT order owner concept is correctly preserved (autopilot still forwards the config-driven allowlist to solvers, driver still accepts it in the auction), and the dead code / artifacts / DB table / crate are all removed together. The TABLES list and log filters were updated too. No dangling cow_amm references remain in compiled code, and Maintaining is still needed for the other indexers.

I flagged two rollout-coordination concerns as inline comments — neither is a logic bug, but both can break a deploy:

V113__drop_cow_amms.sql — dropping the table in the same release that removes the indexer means the still-running old autopilot pod (its cow-amm indexer sits in run_essential_maintenance) will error on the migrated schema during the k8s overlap window. Suggest shipping the DROP in a follow-up release.
Config deny_unknown_fields — removing the cow-amm (autopilot) / cow-amms (driver) fields means any environment whose config still has those sections will fail to start. Infra configs need cleaning up as part of the rollout.

One behavioral note (not flagged, appears intentional): surplus-capturing JIT owners are no longer filtered by native-price availability in solvable_orders.rs, since token info isn't known for a plain owner allowlist — responsibility now sits with the solver creating the JIT order.

Given the PR is explicitly not meant to merge yet, these are mainly things to sequence correctly when you do execute.

claude · 2026-07-13T15:04:23Z

@@ -0,0 +1 @@
+DROP TABLE IF EXISTS cow_amms;


Rollout ordering: dropping cow_amms in the same release that removes the indexer is risky.

Per the DB rollout note, during a deploy k8s starts the new autopilot, runs the Flyway migration (this DROP TABLE), and only then shuts down the old pod. During that overlap the old autopilot is still running its cow-amm indexer, which reads/writes cow_amms. In the current (old) code that indexer runs inside run_essential_maintenance's tokio::try_join! — so once the table is gone, every essential-maintenance cycle on the old pod will error, which can stall auction building until the old pod is finally terminated.

Suggest splitting into two releases: this PR removes all code that touches cow_amms (keep the table), then ship DROP TABLE cow_amms in a follow-up release once no running version references it.

Also worth noting for the DB README/checklist: DROP TABLE is not reversible (data loss) — intentional here, but should be called out.

Claude is correct and if the drop takes a while it might break deployment too

The drop itself would be super fast but the new autopilot takes a while to become ready. During that time the maintenance of the cow amms would fail.
When dropping the table, please also delete its documentation in the readme file.

claude · 2026-07-13T15:04:27Z

+    /// Allowlist of owner addresses whose JIT orders are allowed to capture
+    /// surplus in the auction.
    #[serde(default)]
-    pub cow_amm: CowAmmGroupConfig,
+    pub surplus_capturing_jit_order_owners: Vec<Address>,


Configuration is annotated with #[serde(rename_all = "kebab-case", deny_unknown_fields)], and this PR removes the cow-amm field. Any deployed autopilot config that still contains a cow-amm = { ... } section will now fail to deserialize and the process won't start.

The same applies to the driver's ContractsConfig (also deny_unknown_fields), which drops the cow-amms list.

Make sure the infra config for every environment has the cow-amm / cow-amms sections removed as part of (or before) this rollout so the new binaries can boot.

jmg-duarte · 2026-07-13T17:18:09Z

@@ -0,0 +1 @@
+DROP TABLE IF EXISTS cow_amms;


Claude is correct and if the drop takes a while it might break deployment too

jmg-duarte · 2026-07-13T17:20:07Z

+        let base_auction = tasks.auction.clone().await;

        let auction = Auction {
            id: base_auction.id,
-            orders: base_auction
-                .orders
-                .iter()
-                .cloned()
-                .chain(cow_amm_orders.iter().cloned())
-                .collect(),
+            orders: base_auction.orders.clone(),


unless im missing something can't we do

let BaseAuction { elements go here } = tasks.auction.clone().await;

and remove the later clones?

No, the auction returned from the future is Arced. But since orders are now Arced as well we can consider unArcing the auction to directly return an owned auction.
That should not be addressed in this PR, though. Might make sense to look into this during the 1 block per auction speedup investigations.

jmg-duarte · 2026-07-13T17:22:18Z

+    /// surplus in the auction.
    #[serde(default)]
-    pub cow_amm: CowAmmGroupConfig,
+    pub surplus_capturing_jit_order_owners: Vec<Address>,


the infra pr removes the fields and doesnt add any, do we actually need this or could we also remove this field and remove even more things?

MartinquaXD

🧹
Not sure if we actually need to keep the surplus capturing jit order owner concept but removing this might also be a bit more involved. Starting by ripping out only the cow amm specific stuff is definitely the right call.

linear-code · 2026-07-15T12:48:09Z

BE-99

fleupold added 2 commits July 13, 2026 16:31

Remove CoW AMMs

824e795

recreate inflight settlement logic

6b5f04a

fleupold requested a review from a team as a code owner July 13, 2026 14:59

claude Bot reviewed Jul 13, 2026

View reviewed changes

jmg-duarte reviewed Jul 13, 2026

View reviewed changes

MartinquaXD reviewed Jul 15, 2026

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Remove CoW AMMs#4624

Remove CoW AMMs#4624
fleupold wants to merge 2 commits into
mainfrom
remove-cow-amm

fleupold commented Jul 13, 2026

Uh oh!

github-actions Bot commented Jul 13, 2026

Uh oh!

claude Bot commented Jul 13, 2026 •

edited

Loading

Uh oh!

claude Bot Jul 13, 2026

Uh oh!

jmg-duarte Jul 13, 2026

Uh oh!

MartinquaXD Jul 15, 2026

Uh oh!

claude Bot Jul 13, 2026

Uh oh!

jmg-duarte Jul 13, 2026

Uh oh!

jmg-duarte Jul 13, 2026

Uh oh!

MartinquaXD Jul 15, 2026

Uh oh!

jmg-duarte Jul 13, 2026

Uh oh!

MartinquaXD left a comment

Uh oh!

linear-code Bot commented Jul 15, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Uh oh!

Conversation

fleupold commented Jul 13, 2026

Description

Changes

How to test

Related Issues

Uh oh!

github-actions Bot commented Jul 13, 2026

Uh oh!

claude Bot commented Jul 13, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

PR Review: Remove CoW AMMs

Uh oh!

claude Bot Jul 13, 2026

Choose a reason for hiding this comment

Uh oh!

jmg-duarte Jul 13, 2026

Choose a reason for hiding this comment

Uh oh!

MartinquaXD Jul 15, 2026

Choose a reason for hiding this comment

Uh oh!

claude Bot Jul 13, 2026

Choose a reason for hiding this comment

Uh oh!

jmg-duarte Jul 13, 2026

Choose a reason for hiding this comment

Uh oh!

jmg-duarte Jul 13, 2026

Choose a reason for hiding this comment

Uh oh!

MartinquaXD Jul 15, 2026

Choose a reason for hiding this comment

Uh oh!

jmg-duarte Jul 13, 2026

Choose a reason for hiding this comment

Uh oh!

MartinquaXD left a comment

Choose a reason for hiding this comment

Uh oh!

linear-code Bot commented Jul 15, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

claude Bot commented Jul 13, 2026 •

edited

Loading