Oev 671 txm improvements #294

dimriou · 2025-11-07T18:38:27Z

Each commit corresponds to the following features:

OEV-676 Sync broadcasting and backfilling
OEV-680 Trigger instant retransmission of purgeable transactions
OEV-675 Reduce empty transaction gas limit
OEV-677 Add Meta Error Handler
OEV-668 Add agnostic gas bumping
OEV-679 Add separate pricing for empty transactions

Many of the TODOs of this PR are fixed in:

Implement GetMaxFee method #303
Specifically for bumping Purgeable attempts, you can see the updated method in the above PR.

Update: GetMaxFee PR merged into this

augustbleeds · 2025-11-11T01:58:56Z

pkg/txm/txm.go

-	go t.broadcastLoop(address, triggerCh)
-	go t.backfillLoop(address)
+	t.wg.Add(1)
+	go t.loop(address, triggerCh)


can you remind me what the advantage was of combining these two loops?

It enables easier nonce management via error handling when broadcasting a transaction.

augustbleeds · 2025-11-11T02:03:35Z

pkg/txm/txm.go

+		return err
 	}

 	if tx == nil || *tx.Nonce != latestNonce {


when does this scenario happen? hard for me to think of why the local nonce would be behind the chain nonce if it has to originate from the local txm.

Not behind but in front. Here's an example: you broadcast tx A to the mempool. You restart the TXM, fetch the pending nonce, which is A+1, and start sending more transactions to the mempool. Since you restarted the TXM, transaction A is not being tracked. If it gets dropped by the RPC or the builders, TXM won't be able to recover unless it fills the nonce. That's what this case does.

I noticed TXM uses pending for the block parameter to capture pending transactions.
Does the Flashbots client behave accordingly (i.e., takes into account inflight transactions)?

Yes it does.

augustbleeds · 2025-11-11T02:29:17Z

pkg/txm/attempt_builder.go

+			// TODO: add better handling
+			bumpedAttempt, err := a.NewBumpAttempt(ctx, lggr, tx, *attempt)
+			if err != nil {
+				return attempt, nil


can we select for that specific error so we reduce possibility of halting on a different error?

pkg/txm/clientwrappers/dualbroadcast/meta_client.go

augustbleeds

Changes look good to me, just want to clarify the error handling behavior for empty tx gas bumping.

cll-gg · 2025-11-17T12:28:31Z

pkg/txm/clientwrappers/dualbroadcast/meta_error_handler.go

+}
+
+func (e *errorHandler) HandleError(ctx context.Context, tx *types.Transaction, txErr error, txStore txm.TxStore, setNonce func(common.Address, uint64), isFromBroadcastMethod bool) error {
+	// If this isn't the first broadcast, don't mark the tx as fatal as other txs might be included on-chain.


Suggested change

// If this isn't the first broadcast, don't mark the tx as fatal as other txs might be included on-chain.

// Only if this is the first broadcast, mark the tx as fatal. In other cases, other txs might be included on-chain

cll-gg · 2025-11-17T12:31:01Z

pkg/txm/attempt_builder.go

+
+	// bump purge attempts
+	if tx.IsPurgeable {
+		// TODO: add better handling


Would be good to make a separate ticket with an explanation of what would be better here

PR that permanently handles this is here:

Implement GetMaxFee method #303

lhmerino · 2025-11-18T19:26:19Z

pkg/txm/attempt_builder.go

+		for {
+			bumpedAttempt, err := a.NewBumpAttempt(ctx, lggr, tx, *attempt)
+			if err != nil {
+				if errors.Is(err, fees.ErrConnectivity) {


I'm looking at the error choice. I believe this is due to the code located here

if bumpedMaxPriorityFeePerGas.Cmp(priorityFeeThreshold) > 0 { return ..., ErrConnectivity }

It feels slightly counterintuitive to return ErrConnectivity for a value check, where ErrBumpFeeExceedsLimit might seem more natural.

Is the intention to treat priorityFeeThreshold as a sanity guardrail (implying the network/estimator is broken if fees get this high) rather than a budget limit (implying the tx just needs to pay more)? If so, adding a small comment explaining why this triggers a 'connectivity' error would be helpful for readability.

This functionality is based on TXMv1. Both FeeHistory and BlockHistory estimators use the 90th percentile as a maximum threshold to prevent indefinite gas bumping.
The above was used in a hacky way to access that percentile without changing multiple interfaces in the gas package during testing. We'll have to update the PR and change this before it gets merged.

lhmerino · 2025-11-18T22:09:20Z

pkg/txm/txm.go

 	defer cancel()
 	broadcastWithBackoff := newBackoff(1 * time.Second)
 	var broadcastCh <-chan time.Time
+	backfillCh := time.After(utils.WithJitter(t.config.BlockTime))


utils.WithJitter is a deprecated function.
It says to use timeutils package instead.

lhmerino · 2025-11-18T22:39:30Z

pkg/txm/clientwrappers/dualbroadcast/meta_client.go

 	metrics, err := NewMetaMetrics(chainID.String())
 	if err != nil {
 		return nil, fmt.Errorf("failed to create Meta metrics: %w", err)
 	}


Is the MetaClient currently active -- it looks A-specific?
I noticed a potential dependency in the metrics setup. If beholder is unavailable, will this cause the client to error out as well?

Ideally, we want to ensure the client is decoupled so that a beholder outage doesn't block the main flow. For example, on 10/10 I saw many alerts about beholder potentially having issues.

Both prom and beholder packages are safe and if they're not available a default object will be created with NOOPs.

The errors you saw are due to Beholder package not properly exlcuding the bootstrap node from the metrics and sending repeated error messages. This is Beholder-specific.

lhmerino · 2025-11-18T22:45:32Z

pkg/txm/clientwrappers/dualbroadcast/selector.go

-		return NewFlashbotsClient(client, keyStore, url), nil
+		return NewFlashbotsClient(client, keyStore, url), nil, nil
 	default:
-		return NewMetaClient(lggr, client, keyStore, url, chainID)


This code could then be made cleaner if NewMetaClient does not fail.

lhmerino · 2025-11-18T23:10:07Z

pkg/txm/storage/inmemory_store.go

+
+	// TODO: for now do the simple thing and drop the transaction instead of adding it to the fatal queue.
+	delete(m.UnconfirmedTransactions, *txToMark.Nonce)
+	delete(m.Transactions, txToMark.ID)


I have a few safety concerns about deleting these entries:

Nonce Gaps: Could deleting this nonce create a gap in our internal inflight list? I'm concerned that higher-nonce transactions currently inflight might stall indefinitely waiting for the tx associated with this nonce to fill.

Panic Risk: Dereferencing *txToMark.Nonce without checking if txToMark.Nonce is not nil can cause a panic.

Less important but still relevant: could this complicate debugging by deleting this information?

Nonce Gaps: it's in the TXM's responsibility to call MarkTxFatal in the correct place so it doesn't create a nonce gap, the same way it does for other methods as well. In this case, we call MarkTxFata inside the Meta error handler and only if there are no past broadcasts. Finally, if all the above measures for some reason fail, the TXM has a native nonce gap detection mechanism that will fill any untracked nonce gaps and unblock txs with higher nonces.

Panic Risk: we check for a nil nonce value a few lines before here.

Debugging: the fatal transactions are not currently being utilized in some sort of way. We do log the information of this marking here, but for now there is no effort to do something with the fatal queue, hence the TODO.

lhmerino · 2025-11-18T23:26:46Z

pkg/txm/storage/inmemory_store.go

 		FromAddress:       m.address,
 		ToAddress:         common.Address{},
 		Value:             big.NewInt(0),
 		SpecifiedGasLimit: gasLimit,


Suggestion: We could simplify this by hardcoding the gas limit to 21_000.

Since a standard ETH transfer to an EOA (0x0) has a fixed cost (21_000), passing in gasLimit and maintaining a separate EmptyTxLimitDefault config seems like unnecessary overhead. In addition, it would also remove the need to add gasEstimatorConfig parameter to func NewTxmV2.

If we were to instead send ETH to/from a smart contract, only then would the gas amount differ.

Although standard ETH transfer uses 21k we've observed other EVM-based chains using slightly different values. Since there is already a Core config that we can leverage, it's better to future-proof the design than to require a new custom release if something needs to be adjusted.

Copilot

Pull request overview

This PR implements several TXMv2 improvements including synchronized broadcasting and backfilling, instant retransmission of purgeable transactions, reduced gas limits for empty transactions, a Meta error handler, agnostic gas bumping, and separate pricing for empty transactions.

Reviewed changes

Copilot reviewed 16 out of 16 changed files in this pull request and generated 8 comments.

Show a summary per file

File	Description
pkg/txmgr/builder.go	Adds gasEstimatorConfig parameter, temporary validation check, and error handler support to NewTxmV2
pkg/txm/txm_test.go	Changes package to txm_test, updates test function names to use public methods, adds comprehensive flow tests for retransmission and error handling
pkg/txm/txm.go	Merges broadcast and backfill loops into single loop, adds NewAgnosticBumpAttempt interface, updates ErrorHandler signature, exports Metrics and nonce methods
pkg/txm/storage/inmemory_store_test.go	Adds test for MarkTxFatal functionality
pkg/txm/storage/inmemory_store.go	Implements MarkTxFatal to remove fatal transactions from store
pkg/txm/mock_tx_store_test.go	Renames mock types from lowercase to uppercase for consistent exported names
pkg/txm/mock_client_test.go	Renames mock types from lowercase to uppercase for consistent exported names
pkg/txm/mock_attempt_builder_test.go	Removes NewBumpAttempt mock, renames to NewAgnosticBumpAttempt, updates mock naming
pkg/txm/clientwrappers/dualbroadcast/selector.go	Updates SelectClient to return error handler alongside client
pkg/txm/clientwrappers/dualbroadcast/meta_error_handler_test.go	Adds tests for Meta error handler behavior with no bids errors
pkg/txm/clientwrappers/dualbroadcast/meta_error_handler.go	Implements error handler to mark transactions fatal on first attempt with no bids error
pkg/txm/clientwrappers/dualbroadcast/meta_client.go	Returns error for no bids, adds error constants, fixes logger name, improves logging
pkg/txm/attempt_builder_test.go	Adds comprehensive tests for NewAttempt and NewAgnosticBumpAttempt methods
pkg/txm/attempt_builder.go	Implements NewAgnosticBumpAttempt with configurable bumping logic and purgeable transaction handling
pkg/chains/legacyevm/evm_txm.go	Passes gasEstimatorConfig to NewTxmV2
pkg/.mockery.yaml	Updates mock naming convention to use uppercase prefix

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2025-11-26T15:08:06Z

pkg/txm/txm.go

 	t.nonceMapMu.Lock()
-	t.nonceMap[address] = nonce
 	defer t.nonceMapMu.Unlock()
+	t.nonceMap[address] = nonce


The defer statement for unlocking the mutex is placed before the actual write operation, which could cause the lock to be released before the write completes in certain execution scenarios. Move the defer statement to immediately follow the Lock() call on line 206.

Copilot · 2025-11-26T15:08:06Z

pkg/txm/txm.go

 		}

-		if tx.LastBroadcastAt == nil || time.Since(*tx.LastBroadcastAt) > (t.config.BlockTime*time.Duration(t.config.RetryBlockThreshold)) {
+		if tx.LastBroadcastAt == nil || time.Since(*tx.LastBroadcastAt) > (t.config.BlockTime*time.Duration(t.config.RetryBlockThreshold)) || tx.IsPurgeable {


The complex conditional combining time calculations and boolean checks reduces readability. Consider extracting the time-based retry condition into a named variable like shouldRetryBasedOnTime := tx.LastBroadcastAt == nil || time.Since(*tx.LastBroadcastAt) > (t.config.BlockTime*time.Duration(t.config.RetryBlockThreshold)) to improve clarity.

Suggested change

if tx.LastBroadcastAt == nil || time.Since(*tx.LastBroadcastAt) > (t.config.BlockTime*time.Duration(t.config.RetryBlockThreshold)) || tx.IsPurgeable {

shouldRetryBasedOnTime := tx.LastBroadcastAt == nil || time.Since(*tx.LastBroadcastAt) > (t.config.BlockTime*time.Duration(t.config.RetryBlockThreshold))

if shouldRetryBasedOnTime || tx.IsPurgeable {

Copilot · 2025-11-26T15:08:07Z

pkg/txm/clientwrappers/dualbroadcast/meta_error_handler.go

+
+func (e *errorHandler) HandleError(ctx context.Context, tx *types.Transaction, txErr error, txStore txm.TxStore, setNonce func(common.Address, uint64), isFromBroadcastMethod bool) error {
+	// Mark the tx as fatal only if this is the first broadcast. In any other case, other txs might be included on-chain.
+	if strings.Contains(txErr.Error(), NoBidsError) && tx.AttemptCount == 1 {


Using string comparison to detect error types is fragile and error-prone. If the error message changes or wraps additional context, this check may fail. Consider using errors.Is() with a sentinel error or implement a custom error type for NoBidsError.

Copilot · 2025-11-26T15:08:07Z

pkg/txm/storage/inmemory_store.go

 		Value:             big.NewInt(0),
 		SpecifiedGasLimit: gasLimit,
 		CreatedAt:         time.Now(),
 		State:             txmgr.TxUnconfirmed,


Empty transactions are automatically marked as purgeable without clear documentation explaining this design decision. Add a comment explaining why empty transactions should always be purgeable to help future maintainers understand the intent.

Suggested change

State: txmgr.TxUnconfirmed,

State: txmgr.TxUnconfirmed,

// Empty transactions are always marked as purgeable because they do not represent real on-chain activity.

// This allows the system to safely remove them to free up resources or maintain nonce continuity.

Copilot · 2025-11-26T15:08:07Z

pkg/txm/attempt_builder.go

+		for {
+			bumpedAttempt, err := a.NewBumpAttempt(ctx, lggr, tx, *attempt)
+			if err != nil {
+				if errors.Is(err, fees.ErrConnectivity) {
+					return attempt, nil
+				}
+				return nil, fmt.Errorf("error bumping attempt for txID: %v, err: %w", tx.ID, err)
+			}
+			attempt = bumpedAttempt
+		}


The infinite loop for purgeable attempts has no bounds or circuit breaker. If fees.ErrConnectivity is never encountered, this will loop indefinitely, potentially causing performance issues or resource exhaustion. Add a maximum iteration count or timeout mechanism.

Copilot · 2025-11-26T15:08:08Z

pkg/txm/txm_test.go

 			GasLimit: 22000,
 		}
-		ab.On("NewAttempt", mock.Anything, mock.Anything, mock.Anything, mock.Anything).Return(attempt, nil).Once()
+		ab.On("NewAgnosticBumpAttempt", mock.Anything, mock.Anything, mock.Anything, mock.Anything).Return(attempt, nil).Once()


[nitpick] Using mock.Anything for all parameters reduces the value of the test by not verifying the actual arguments passed. Consider using more specific matchers to validate that the correct transaction, context, and logger are being passed to NewAgnosticBumpAttempt.

Suggested change

ab.On("NewAgnosticBumpAttempt", mock.Anything, mock.Anything, mock.Anything, mock.Anything).Return(attempt, nil).Once()

ab.On(

"NewAgnosticBumpAttempt",

mock.MatchedBy(func(txArg *types.Tx) bool { return txArg != nil && txArg.ID == tx.ID }),

mock.Anything, // context.Context, hard to match exactly

mock.MatchedBy(func(l logger.Logger) bool { return l == lggr }),

mock.Anything, // EvmFeeEstimator

).Return(attempt, nil).Once()

Copilot · 2025-11-26T15:08:08Z

pkg/txm/attempt_builder_test.go

+		tx := &types.Transaction{ID: 10, FromAddress: address}
+		mockEstimator.On("GetFee", mock.Anything, mock.Anything, mock.Anything, mock.Anything, mock.Anything, mock.Anything).
+			Return(gas.EvmFee{}, uint64(0), errors.New("estimator error")).Once()


The test verifies that GetFee errors are propagated but doesn't test NewBumpAttempt error scenarios. Add test cases to verify error handling when NewBumpAttempt fails during the bumping process in NewAgnosticBumpAttempt.

Copilot · 2025-11-26T15:08:08Z

pkg/txmgr/builder.go

-	attemptBuilder := txm.NewAttemptBuilder(fCfg.PriceMaxKey, estimator, keyStore)
+	// TODO: temporary check until we implement the required methods on the estimator interface
+	if gasEstimatorConfig.Mode() != "BlockHistory" || gasEstimatorConfig.BlockHistory().CheckInclusionBlocks() == 0 {
+		return nil, errors.New("only BlockHistory mode with CheckInclusionBlocks > 0 is supported for TXMv2")


The error message doesn't provide actionable guidance on how to fix the configuration. Consider enhancing it to: return nil, fmt.Errorf(\"TXMv2 requires GasEstimator Mode to be 'BlockHistory' with CheckInclusionBlocks > 0, got Mode=%s, CheckInclusionBlocks=%d\", gasEstimatorConfig.Mode(), gasEstimatorConfig.BlockHistory().CheckInclusionBlocks())

Suggested change

return nil, errors.New("only BlockHistory mode with CheckInclusionBlocks > 0 is supported for TXMv2")

return nil, fmt.Errorf("TXMv2 requires GasEstimator Mode to be 'BlockHistory' with CheckInclusionBlocks > 0, got Mode=%s, CheckInclusionBlocks=%d", gasEstimatorConfig.Mode(), gasEstimatorConfig.BlockHistory().CheckInclusionBlocks())

pavel-raykov · 2025-11-26T17:07:10Z

pkg/chains/legacyevm/evm_txm.go

 				logPoller,
 				opts.KeyStore,
 				estimator,
+				cfg.GasEstimator(),


why do you need to pass the GasEstimator separately, isn't it already passed in cfg?

The entire cfg is a superset of all configs, but the expected param interfaces are a small subset of it. Historically, we did that to split the responsibilities of configs and not pass one large EVM config, so I replicated the same behaviour here, but if you ask me it's not a great design choice.

Thanks for explaining, let's keep it as is.

pavel-raykov · 2025-11-26T19:49:35Z

pkg/txm/clientwrappers/dualbroadcast/meta_client.go

-	timeout = time.Second * 5
-	metaABI = `[
+	timeout                    = time.Second * 5
+	NoBidsError                = "no bids"


why not errors.New("no bids") ? this way you can use errors.Is for easy error testing.

pavel-raykov · 2025-11-26T19:57:49Z

pkg/txm/txm.go

-		if tx.LastBroadcastAt == nil || time.Since(*tx.LastBroadcastAt) > (t.config.BlockTime*time.Duration(t.config.RetryBlockThreshold)) {
+		if tx.LastBroadcastAt == nil || time.Since(*tx.LastBroadcastAt) > (t.config.BlockTime*time.Duration(t.config.RetryBlockThreshold)) || tx.IsPurgeable {
 			// TODO: add optional graceful bumping strategy
 			t.lggr.Info("Rebroadcasting attempt for txID: ", tx.ID)


you can add more information here to show why rebroadcasting happened: it seems that there are 3 conditions above and it would be useful to understand which triggered.

* Implement GetMaxFee method * Update empty transaction fee building

dimriou added 5 commits November 5, 2025 17:53

OEV-676 Sync broadcasting and backfilling

976ba44

OEV-680 Trigger instant retransmission of purgeable transaction

e533378

OEV-675 Reduce empty transaction gas limit

a822f17

OEV-677 Add Meta Error Handler

91c5a2d

OEV-668 Add gas bumping

81ef8e6

dimriou added the do-not-merge label Nov 7, 2025

OEV-679 Add separate pricing for empty txs

64f4d2f

dimriou force-pushed the oev-671_txm_improvements branch from 8276ab6 to 64f4d2f Compare November 10, 2025 10:51

Add tests

7ef98a2

augustbleeds reviewed Nov 11, 2025

View reviewed changes

pkg/txm/clientwrappers/dualbroadcast/meta_client.go Show resolved Hide resolved

Add more tests

c3d4502

augustbleeds reviewed Nov 11, 2025

View reviewed changes

Add txm test cases for purgeable retransmission and error handler

115e871

dimriou force-pushed the oev-671_txm_improvements branch from 29a1813 to 115e871 Compare November 11, 2025 15:51

cll-gg reviewed Nov 17, 2025

View reviewed changes

lhmerino reviewed Nov 18, 2025

View reviewed changes

Address feedback and extend tests

9008093

dimriou marked this pull request as ready for review November 26, 2025 15:06

dimriou requested a review from a team as a code owner November 26, 2025 15:06

Copilot AI review requested due to automatic review settings November 26, 2025 15:06

dimriou requested review from a team as code owners November 26, 2025 15:06

product-security-plaid-production bot requested a review from fernandezlautaro November 26, 2025 15:06

product-security-plaid-production bot requested review from DylanTinianov and pavel-raykov November 26, 2025 15:06

Copilot AI reviewed Nov 26, 2025

View reviewed changes

pavel-raykov reviewed Nov 26, 2025

View reviewed changes

dimriou removed the do-not-merge label Nov 27, 2025

Update NoBidsErr string to ErrNoBids

c8b5fd4

dimriou force-pushed the oev-671_txm_improvements branch from aa97fa9 to c8b5fd4 Compare November 27, 2025 11:59

dimriou added 2 commits December 1, 2025 16:57

Implement GetMaxFee method (#303)

5509c16

* Implement GetMaxFee method * Update empty transaction fee building

Change RequestResponse name

a2968b2

pavel-raykov approved these changes Dec 1, 2025

View reviewed changes

dimriou enabled auto-merge (squash) December 1, 2025 17:09

amit-momin approved these changes Dec 1, 2025

View reviewed changes

dimriou merged commit af04e91 into develop Dec 1, 2025
33 checks passed

dimriou deleted the oev-671_txm_improvements branch December 1, 2025 17:55

dimriou mentioned this pull request Dec 1, 2025

Bump chainlink-evm dep to add SVR-related changes smartcontractkit/chainlink#20495

Merged

	// If this isn't the first broadcast, don't mark the tx as fatal as other txs might be included on-chain.
	// Only if this is the first broadcast, mark the tx as fatal. In other cases, other txs might be included on-chain

	if tx.LastBroadcastAt == nil \|\| time.Since(tx.LastBroadcastAt) > (t.config.BlockTimetime.Duration(t.config.RetryBlockThreshold)) \|\| tx.IsPurgeable {
	shouldRetryBasedOnTime := tx.LastBroadcastAt == nil \|\| time.Since(tx.LastBroadcastAt) > (t.config.BlockTimetime.Duration(t.config.RetryBlockThreshold))
	if shouldRetryBasedOnTime \|\| tx.IsPurgeable {

-		ab.On("NewAgnosticBumpAttempt", mock.Anything, mock.Anything, mock.Anything, mock.Anything).Return(attempt, nil).Once()
+		ab.On(
+			"NewAgnosticBumpAttempt",
+			mock.MatchedBy(func(txArg *types.Tx) bool { return txArg != nil && txArg.ID == tx.ID }),
+			mock.Anything, // context.Context, hard to match exactly
+			mock.MatchedBy(func(l logger.Logger) bool { return l == lggr }),
+			mock.Anything, // EvmFeeEstimator
+		).Return(attempt, nil).Once()

	return nil, errors.New("only BlockHistory mode with CheckInclusionBlocks > 0 is supported for TXMv2")
	return nil, fmt.Errorf("TXMv2 requires GasEstimator Mode to be 'BlockHistory' with CheckInclusionBlocks > 0, got Mode=%s, CheckInclusionBlocks=%d", gasEstimatorConfig.Mode(), gasEstimatorConfig.BlockHistory().CheckInclusionBlocks())

Oev 671 txm improvements #294

Oev 671 txm improvements #294

Uh oh!

Conversation

dimriou commented Nov 7, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

lhmerino Nov 18, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

augustbleeds left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

lhmerino Nov 18, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

lhmerino Nov 18, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

lhmerino Nov 18, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

lhmerino Nov 18, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Nov 26, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Nov 26, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Nov 26, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Nov 26, 2025

Choose a reason for hiding this comment

Uh oh!

dimriou commented Nov 7, 2025 •

edited

Loading

lhmerino Nov 18, 2025 •

edited

Loading

lhmerino Nov 18, 2025 •

edited

Loading

lhmerino Nov 18, 2025 •

edited

Loading

lhmerino Nov 18, 2025 •

edited

Loading

lhmerino Nov 18, 2025 •

edited

Loading