simulate: lint fix, always emit gh group markers, health tuning

- staticcheck QF1012: fmt.Fprintf(&b, ...) over WriteString(Sprintf)
- drop the gh/plain output split and isGitHubActions gating: group
  markers are a useful delimiter everywhere, and job ids now appear in
  the group titles
- agent health: tolerate 3 not-joined timeouts, require 2 fatal log
  markers (a single one can be an isolated job crash), name the marker
  list pythonFatalMarkers (JS will need its own)
- open the log pane automatically when the agent is detected broken
- prune obvious comments

Author: theomonnom

cmd/lk/perf.go

@@ -276,13 +276,13 @@ func agentLoadTest(ctx context.Context, cmd *cli.Command) error {
 	}
 
 	params := loadtester.AgentLoadTestParams{
-		URL:             pc.URL,
-		APIKey:          pc.APIKey,
-		APISecret:       pc.APISecret,
-		Rooms:           int(cmd.Int("rooms")),
-		AgentName:       cmd.String("agent-name"),
-		EchoSpeechDelay: cmd.Duration("echo-speech-delay"),
-		Duration:        cmd.Duration("duration"),
+		URL:                   pc.URL,
+		APIKey:                pc.APIKey,
+		APISecret:             pc.APISecret,
+		Rooms:                 int(cmd.Int("rooms")),
+		AgentName:             cmd.String("agent-name"),
+		EchoSpeechDelay:       cmd.Duration("echo-speech-delay"),
+		Duration:              cmd.Duration("duration"),
 		ParticipantAttributes: participantAttributes,
 	}
 

cmd/lk/simulate.go

@@ -114,9 +114,8 @@ func writeGeneratedScenariosTemp(run *livekit.SimulationRun) (string, error) {
 	return f.Name(), nil
 }
 
-// scenarioGroupToYAML renders a ScenarioGroup as a scenarios.yaml document — the
-// inverse of loadScenarioGroup, decoding each scenario's JSON userdata string
-// back into a nested mapping.
+// scenarioGroupToYAML renders a ScenarioGroup as a scenarios.yaml document, the
+// inverse of loadScenarioGroup.
 func scenarioGroupToYAML(group *livekit.ScenarioGroup) ([]byte, error) {
 	f := scenariosFile{Name: group.GetName()}
 	for _, s := range group.GetScenarios() {

cmd/lk/simulate_ci.go

@@ -183,12 +183,10 @@ func runSimulateCI(ctx context.Context, config *simulateConfig) error {
 
 	_, _, _, failed := simulationJobCounts(run)
 	if failed > 0 || run.Status == livekit.SimulationRun_STATUS_FAILED {
-		if isGitHubActions() {
-			if failed > 0 {
-				fmt.Fprintf(os.Stdout, "::error::%d simulation(s) failed\n", failed)
-			} else {
-				fmt.Fprintf(os.Stdout, "::error::Simulation run failed: %s\n", run.Error)
-			}
+		if failed > 0 {
+			fmt.Fprintf(os.Stdout, "::error::%d simulation(s) failed\n", failed)
+		} else {
+			fmt.Fprintf(os.Stdout, "::error::Simulation run failed: %s\n", run.Error)
 		}
 		if run.Status == livekit.SimulationRun_STATUS_FAILED && len(run.Jobs) == 0 {
 			return fmt.Errorf("simulation failed: %s", run.Error)
@@ -199,10 +197,9 @@ func runSimulateCI(ctx context.Context, config *simulateConfig) error {
 	return nil
 }
 
-// writeRunResults writes the per-job results and the run summary. With gh set,
-// sections are wrapped in GitHub Actions group markers and failed jobs emit
-// ::error:: annotations on GitHub.
-func writeRunResults(w io.Writer, run *livekit.SimulationRun, ap *AgentProcess, gh bool) {
+// writeRunResults writes the per-job results and the run summary, using GitHub
+// Actions group markers (harmless and a useful delimiter outside GitHub too).
+func writeRunResults(w io.Writer, run *livekit.SimulationRun, ap *AgentProcess) {
 	if run == nil {
 		return
 	}
@@ -226,11 +223,7 @@ func writeRunResults(w io.Writer, run *livekit.SimulationRun, ap *AgentProcess,
 			label = fmt.Sprintf("Job %d", i+1)
 		}
 
-		if gh {
-			fmt.Fprintf(w, "::group::%s %s\n", icon, label)
-		} else {
-			fmt.Fprintf(w, "%s %s (%s)\n", icon, label, job.Id)
-		}
+		fmt.Fprintf(w, "::group::%s %s (%s)\n", icon, label, job.Id)
 
 		if job.Instructions != "" {
 			fmt.Fprintln(w, "Instructions:")
@@ -268,20 +261,16 @@ func writeRunResults(w io.Writer, run *livekit.SimulationRun, ap *AgentProcess,
 			}
 		}
 
-		if gh {
-			fmt.Fprintln(w, "::endgroup::")
-		} else {
-			fmt.Fprintln(w)
-		}
+		fmt.Fprintln(w, "::endgroup::")
 
-		if job.Status == livekit.SimulationRun_Job_STATUS_FAILED && gh && isGitHubActions() {
+		if job.Status == livekit.SimulationRun_Job_STATUS_FAILED {
 			firstLine, _, _ := strings.Cut(job.Error, "\n")
 			fmt.Fprintf(w, "::error::Job %d failed: %s\n", i+1, firstLine)
 		}
 	}
 
 	if run.Summary != nil {
-		writeRunSummary(w, run, gh)
+		writeRunSummary(w, run)
 	} else {
 		msg := "The summary for this run is not available"
 		if run.Error != "" {
@@ -292,16 +281,12 @@ func writeRunResults(w io.Writer, run *livekit.SimulationRun, ap *AgentProcess,
 	}
 }
 
-func writeRunSummary(w io.Writer, run *livekit.SimulationRun, gh bool) {
+func writeRunSummary(w io.Writer, run *livekit.SimulationRun) {
 	summary := run.Summary
 	total, _, passed, failed := simulationJobCounts(run)
 
 	fmt.Fprintln(w)
-	if gh {
-		fmt.Fprintln(w, "::group::Summary")
-	} else {
-		fmt.Fprintln(w, "Summary")
-	}
+	fmt.Fprintln(w, "::group::Summary")
 	fmt.Fprintf(w, "%d total, %d passed, %d failed\n", total, passed, failed)
 
 	if summary.GoingWell != "" {
@@ -331,9 +316,7 @@ func writeRunSummary(w io.Writer, run *livekit.SimulationRun, gh bool) {
 		}
 	}
 
-	if gh {
-		fmt.Fprintln(w, "::endgroup::")
-	}
+	fmt.Fprintln(w, "::endgroup::")
 }
 
 func writeChatHistory(w io.Writer, chatCtx *agent.ChatContext) {
@@ -376,7 +359,3 @@ func writeChatHistory(w io.Writer, chatCtx *agent.ChatContext) {
 		}
 	}
 }
-
-func isGitHubActions() bool {
-	return os.Getenv("GITHUB_ACTIONS") != ""
-}

cmd/lk/simulate_health.go

@@ -27,21 +27,25 @@ import (
 // from transient connection pacing, which can cause an isolated "no agent joined"
 // timeout even when the agent is healthy.
 
-// agentFatalMarkers are framework log lines that only appear on a fatal,
-// non-transient worker error.
-var agentFatalMarkers = []string{
+// pythonFatalMarkers are livekit-agents (Python) log lines that only appear on
+// a fatal, non-transient worker error. The JS framework will need its own set.
+var pythonFatalMarkers = []string{
 	"unhandled exception while running the job task",
 	"error initializing process",
 	"closing due to unrecoverable error",
 }
 
+// minFatalMarkers is how many fatal log lines it takes to call the worker
+// broken; a single one can be an isolated job crash.
+const minFatalMarkers = 2
+
 // maxAgentNotJoined is how many "no agent joined" timeouts to tolerate before
 // treating the worker as broken.
-const maxAgentNotJoined = 1
+const maxAgentNotJoined = 3
 
 // agentBroken reports whether the worker is failing systemically. A completed
 // scenario proves the agent works, so it is never broken in that case; otherwise
-// a fatal log marker, or more than one "no agent joined" timeout, marks it broken.
+// repeated fatal log markers or "no agent joined" timeouts mark it broken.
 func agentBroken(run *livekit.SimulationRun, ap *AgentProcess) bool {
 	completed, notJoined := 0, 0
 	for _, job := range run.GetJobs() {
@@ -57,7 +61,7 @@ func agentBroken(run *livekit.SimulationRun, ap *AgentProcess) bool {
 	switch {
 	case completed > 0:
 		return false
-	case ap != nil && lastFatalMarker(ap.RecentLogs(0)) >= 0:
+	case ap != nil && countFatalMarkers(ap.RecentLogs(0)) >= minFatalMarkers:
 		return true
 	default:
 		return notJoined > maxAgentNotJoined
@@ -81,17 +85,34 @@ func agentErrorContext(ap *AgentProcess) []string {
 	return out
 }
 
-// lastFatalMarker returns the index of the last log line matching an
-// agentFatalMarker, or -1 if none.
+// lastFatalMarker returns the index of the last log line matching a fatal
+// marker, or -1 if none.
 func lastFatalMarker(logs []string) int {
 	last := -1
 	for i, line := range logs {
-		for _, marker := range agentFatalMarkers {
-			if strings.Contains(ansiEscapeRe.ReplaceAllString(line, ""), marker) {
-				last = i
-				break
-			}
+		if isFatalMarker(line) {
+			last = i
 		}
 	}
 	return last
 }
+
+func countFatalMarkers(logs []string) int {
+	n := 0
+	for _, line := range logs {
+		if isFatalMarker(line) {
+			n++
+		}
+	}
+	return n
+}
+
+func isFatalMarker(line string) bool {
+	plain := ansiEscapeRe.ReplaceAllString(line, "")
+	for _, marker := range pythonFatalMarkers {
+		if strings.Contains(plain, marker) {
+			return true
+		}
+	}
+	return false
+}

cmd/lk/simulate_report.go

@@ -117,7 +117,7 @@ func (l *simLog) BrokenAgent() {
 }
 
 func (l *simLog) Results(run *livekit.SimulationRun, ap *AgentProcess) {
-	writeRunResults(l.out, run, ap, true)
+	writeRunResults(l.out, run, ap)
 }
 
 func writeBrokenAgentNote(w io.Writer, ap *AgentProcess) {

cmd/lk/simulate_subprocess.go

@@ -353,7 +353,7 @@ func roomNamePrefix(roomName string) string {
 
 // RecentRoomLogsByPrefix returns log lines for the most recent room matching
 // the prefix of the given room name. When a job is retried, each attempt gets
-// a new room with the same prefix — we show only the latest attempt's logs.
+// a new room with the same prefix; only the latest attempt's logs are shown.
 func (ap *AgentProcess) RecentRoomLogsByPrefix(n int, roomName string) []string {
 	ap.mu.Lock()
 	defer ap.mu.Unlock()