• Resolve metric name collisions by adding consumer/provider prefixes
  - Rename lava_latest_block → lava_consumer_latest_block / lava_provider_latest_block
  - Rename virtual_epoch → lava_consumer_virtual_epoch / lava_provider_virtual_epoch
  - Add missing lava_consumer_ prefix to 4 metrics
• Fix label typo: dissconectReason → disconnectReason
• Implement StartSelectionStatsUpdater in SmartRouterMetricsManager
  - Periodically writes optimizer QoS reports to metrics
  - Prevents silent no-op that dropped optimizer-sampled scores
• Remove unused provider selection score gauges from consumer metrics
• Wire OptimizerQoSClient through SmartRouterMetricsManagerOptions

Add metric prefixes, fix typo, remove unused gauges

• Rename lava_latest_block to lava_consumer_latest_block and virtual_epoch to
 lava_consumer_virtual_epoch
• Add missing lava_consumer_ prefix to relay processing latency, requests per provider, and
 protocol errors metrics
• Fix typo in websocket disconnect metric label: dissconectReason → disconnectReason
• Remove unused provider selection score gauges (availability, latency, sync, stake, composite)
• Simplify SetProviderSelected method to only extract selected provider's composite score

protocol/metrics/consumer_metrics_manager.go

Add provider prefix to block and epoch metrics

• Rename lava_latest_block to lava_provider_latest_block
• Rename virtual_epoch to lava_provider_virtual_epoch

protocol/metrics/provider_metrics_manager.go

Implement optimizer QoS periodic stats updater

• Add optimizerQoSClient field to SmartRouterMetricsManager struct
• Add OptimizerQoSClient parameter to SmartRouterMetricsManagerOptions
• Implement StartSelectionStatsUpdater method with periodic ticker
• Periodically fetch optimizer QoS reports and update endpoint selection score metrics

protocol/metrics/smartrouter_metrics_manager.go

Description

SmartRouterMetricsManager.StartSelectionStatsUpdater calls time.NewTicker(updateInterval) without
validating updateInterval, so a 0/negative OptimizerQosServerSamplingInterval will panic and crash
the smart router. The interval is user-configurable via a CLI flag, so this is a runtime footgun
rather than a theoretical edge case.

Code

protocol/metrics/smartrouter_metrics_manager.go[R822-828]

+func (m *SmartRouterMetricsManager) StartSelectionStatsUpdater(ctx context.Context, updateInterval time.Duration) {
+	if m == nil || m.optimizerQoSClient == nil {
+		return
+	}
+	go func() {
+		ticker := time.NewTicker(updateInterval)
+		defer ticker.Stop()

Evidence

StartSelectionStatsUpdater unconditionally creates a ticker from the passed duration; Go tickers
panic on non-positive durations. The duration comes from a DurationVar flag (user input) and is
passed into StartSelectionStatsUpdater at startup, so misconfiguration leads to an immediate crash
loop.

protocol/metrics/smartrouter_metrics_manager.go[822-829]
protocol/rpcsmartrouter/rpcsmartrouter.go[244-246]
protocol/rpcsmartrouter/rpcsmartrouter.go[1458-1462]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

### Issue description
`SmartRouterMetricsManager.StartSelectionStatsUpdater` creates `time.NewTicker(updateInterval)` without checking that `updateInterval > 0`. Since `OptimizerQosServerSamplingInterval` is user-configurable via a flag, setting it to `0` or a negative duration will panic and crash the smart router.

### Issue Context
The smart router calls `StartSelectionStatsUpdater(ctx, metrics.OptimizerQosServerSamplingInterval)` during startup. The interval is set via a `DurationVar` flag, so it can be misconfigured.

### Fix Focus Areas
- protocol/metrics/smartrouter_metrics_manager.go[822-845]
- protocol/rpcsmartrouter/rpcsmartrouter.go[244-246]
- protocol/rpcsmartrouter/rpcsmartrouter.go[1458-1462]

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

Description

SmartRouterMetricsManagerOptions exposes a concrete *ConsumerOptimizerQoSClient dependency,
expanding the exported API surface and preventing interface-based injection. This conflicts with the
guideline to depend on interfaces and minimize exported APIs.

Code

protocol/metrics/smartrouter_metrics_manager.go[R107-112]

// SmartRouterMetricsManagerOptions contains configuration for the metrics manager
type SmartRouterMetricsManagerOptions struct {
-	NetworkAddress  string
-	StartHTTPServer bool // If false, only register metrics (for use alongside ConsumerMetricsManager)
+	NetworkAddress      string
+	StartHTTPServer     bool // If false, only register metrics (for use alongside ConsumerMetricsManager)
+	OptimizerQoSClient *ConsumerOptimizerQoSClient
}

Evidence

Compliance ID 2 requires depending on interfaces and minimizing exports. The PR adds an exported
options field OptimizerQoSClient typed as the concrete *ConsumerOptimizerQoSClient, even though
the manager only needs the GetReportsToSend() behavior.

AGENTS.md
protocol/metrics/smartrouter_metrics_manager.go[107-112]
protocol/metrics/smartrouter_metrics_manager.go[822-841]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

## Issue description
`SmartRouterMetricsManagerOptions` adds an exported field `OptimizerQoSClient` typed as `*ConsumerOptimizerQoSClient`. This hard-codes a concrete dependency into the exported API surface, making testing and alternate implementations harder.

## Issue Context
`StartSelectionStatsUpdater` only needs `GetReportsToSend()`, so this dependency can be expressed as a narrow interface (or a function callback) to comply with interface-based dependency guidance and minimize exported API coupling.

## Fix Focus Areas
- protocol/metrics/smartrouter_metrics_manager.go[91-112]
- protocol/metrics/smartrouter_metrics_manager.go[822-845]
- protocol/rpcsmartrouter/rpcsmartrouter.go[233-246]

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

Description

ConsumerMetricsManager no longer registers/updates the lava_consumer_provider_*_score gauges, but
the repository monitoring docs still instruct users to query lava_consumer_provider_composite_score.
This will break the documented PromQL queries (and likely dashboards/alerts) immediately after
deployment.

Code

protocol/metrics/consumer_metrics_manager.go[L879-883]

-		pme.providerAvailabilityScoreGauge.WithLabelValues(chainId, scores.ProviderAddress).Set(scores.Availability)
-		pme.providerLatencyScoreGauge.WithLabelValues(chainId, scores.ProviderAddress).Set(scores.Latency)
-		pme.providerSyncScoreGauge.WithLabelValues(chainId, scores.ProviderAddress).Set(scores.Sync)
-		pme.providerStakeScoreGauge.WithLabelValues(chainId, scores.ProviderAddress).Set(scores.Stake)
-		pme.providerCompositeScoreGauge.WithLabelValues(chainId, scores.ProviderAddress).Set(scores.Composite)

Evidence

The metrics manager no longer registers any of the provider score gauges, and SetProviderSelected no
longer emits score-per-provider gauges. Meanwhile, scripts/monitoring/README.md explicitly
references the removed metric name in example queries, creating a guaranteed mismatch between docs
and actual exported metrics.

protocol/metrics/consumer_metrics_manager.go[360-391]
protocol/metrics/consumer_metrics_manager.go[822-886]
scripts/monitoring/README.md[69-76]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

### Issue description
Consumer provider score gauge metrics (`lava_consumer_provider_*_score`) have been removed from the consumer metrics manager, but in-repo monitoring documentation still references at least `lava_consumer_provider_composite_score`. This breaks documented PromQL queries and likely any dashboards/alerts based on those metrics.

### Issue Context
The PR removes creation/registration/updating of provider score gauges and changes `SetProviderSelected` behavior accordingly. The monitoring README still recommends querying the removed metric names.

### Fix Focus Areas
- protocol/metrics/consumer_metrics_manager.go[360-437]
- protocol/metrics/consumer_metrics_manager.go[822-886]
- scripts/monitoring/README.md[59-100]

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.