fix: sre agent

cmd/deploy.go

@@ -124,7 +124,7 @@ Examples:
 			fmt.Fprintf(os.Stderr, format+"\n", args...)
 		}
 		if sreMode {
-			logf("[deploy] --sre requested; Docker is the default SRE runtime. After deployment, run: clanker sre install --target docker --apply")
+			logf("[deploy] --sre requested; planning a long-running Clanker SRE agent with heartbeat verification")
 		}
 
 		// 3. Resolve AWS profile/region early so intelligence pipeline can scan infra
@@ -323,7 +323,7 @@ Examples:
 		if planProvider == "" {
 			planProvider = "aws"
 		}
-		deployObjectiveContext := withOneClickDeployContext(baseQuestion, planProvider, intel.Architecture.Method, enforceImageDeploy)
+		deployObjectiveContext := withOneClickDeployContext(baseQuestion, planProvider, intel.Architecture.Method, enforceImageDeploy, sreMode)
 		planningContext := compactPlanningContext(deployObjectiveContext, planProvider)
 		projectSummaryForLLM := strings.TrimSpace(rp.Summary)
 		if intel.DeepAnalysis != nil && strings.TrimSpace(intel.DeepAnalysis.AppDescription) != "" {
@@ -1773,20 +1773,27 @@ func compactOneLine(raw string, limit int) string {
 
 var shellStylePlaceholderRe = regexp.MustCompile(`\$\{([A-Z0-9_]+)\}`)
 
-func withOneClickDeployContext(base, provider, method string, enforceImageDeploy bool) string {
-	context := buildOneClickDeployObjective(provider, method, enforceImageDeploy)
+func withOneClickDeployContext(base, provider, method string, enforceImageDeploy bool, sreMode bool) string {
+	context := buildOneClickDeployObjective(provider, method, enforceImageDeploy, sreMode)
 	base = strings.TrimSpace(base)
 	if base == "" {
 		return context
 	}
 	return context + "\n\n" + base
 }
 
-func buildOneClickDeployObjective(provider, method string, enforceImageDeploy bool) string {
+func buildOneClickDeployObjective(provider, method string, enforceImageDeploy bool, sreMode bool) string {
 	prov := strings.ToLower(strings.TrimSpace(provider))
 	if prov == "" {
 		prov = "aws"
 	}
+	if sreMode {
+		deployID := strings.TrimSpace(os.Getenv("CLANKER_SRE_DEPLOY_ID"))
+		if deployID == "" {
+			deployID = "$CLANKER_SRE_DEPLOY_ID"
+		}
+		return fmt.Sprintf("[one-click SRE deploy objective]\nGenerate command plan steps that deploy only the long-running Clanker SRE agent, not the analyzed app. Use provider=%s and the smallest practical always-on runtime for that provider. Prefer ghcr.io/bgdnvk/clanker:latest and run: clanker sre run --sre --target cloud-vm --provider %s --deploy-id %s. The runtime MUST set CLANKER_CEREBRO_URL, CLANKER_CEREBRO_INGEST_TOKEN, CLANKER_SRE_DEPLOY_ID, and CLANKER_SRE_PROVIDER. Tag or label every created resource with clanker-sre=true and clanker-sre-deploy-id=%s. Keep commands idempotent, preserve created resource IDs, and include a final non-secret health/verification command that checks the service/container is still running. Never print token values.", prov, prov, deployID, deployID)
+	}
 	m := strings.ToLower(strings.TrimSpace(method))
 	if m == "" {
 		m = "ec2"

cmd/sre.go

@@ -96,6 +96,7 @@ var sreRunCmd = &cobra.Command{
 		cerebroURL, _ := cmd.Flags().GetString("cerebro-url")
 		ingestToken, _ := cmd.Flags().GetString("ingest-token")
 		provider, _ := cmd.Flags().GetString("provider")
+		deployID, _ := cmd.Flags().GetString("deploy-id")
 		interval, err := durationFlag(cmd, "interval", sre.DefaultInterval)
 		if err != nil {
 			return err
@@ -114,6 +115,7 @@ var sreRunCmd = &cobra.Command{
 			Once:        once,
 			Writer:      os.Stdout,
 			Provider:    provider,
+			DeployID:    deployID,
 		})
 		if errors.Is(err, context.Canceled) {
 			return nil
@@ -184,6 +186,8 @@ func addSREPlanFlags(cmd *cobra.Command) {
 	cmd.Flags().String("name", sre.DefaultAgentName, "SRE bot name/container/service name")
 	cmd.Flags().String("cerebro-url", "", "Cerebro API base URL, e.g. http://127.0.0.1:8080/api")
 	cmd.Flags().String("ingest-token-env", sre.DefaultIngestTokenEnv, "Environment variable name that holds the Cerebro ingest token")
+	cmd.Flags().String("provider", "", "Cloud provider name for heartbeat identification (aws, gcp, azure, etc.)")
+	cmd.Flags().String("deploy-id", "", "Stable SRE deployment ID for heartbeat verification")
 	cmd.Flags().String("interval", sre.DefaultInterval.String(), "Heartbeat/discovery interval")
 	cmd.Flags().String("format", "text", "Output format: text or json")
 }
@@ -195,6 +199,7 @@ func addSRERunFlags(cmd *cobra.Command) {
 	cmd.Flags().String("cerebro-url", "", "Cerebro API base URL, e.g. http://127.0.0.1:8080/api")
 	cmd.Flags().String("ingest-token", "", "Cerebro ingest token (or set CLANKER_CEREBRO_INGEST_TOKEN)")
 	cmd.Flags().String("provider", "", "Cloud provider name for heartbeat identification (aws, gcp, azure, etc.)")
+	cmd.Flags().String("deploy-id", "", "Stable SRE deployment ID for heartbeat verification")
 	cmd.Flags().String("interval", sre.DefaultInterval.String(), "Heartbeat/discovery interval")
 	cmd.Flags().Bool("once", false, "Send one heartbeat and exit")
 }
@@ -209,8 +214,10 @@ func buildSREPlan(cmd *cobra.Command) (sre.InstallPlan, error) {
 	name, _ := cmd.Flags().GetString("name")
 	cerebroURL, _ := cmd.Flags().GetString("cerebro-url")
 	ingestTokenEnv, _ := cmd.Flags().GetString("ingest-token-env")
+	provider, _ := cmd.Flags().GetString("provider")
+	deployID, _ := cmd.Flags().GetString("deploy-id")
 	discovery := sre.Discover(cmd.Context())
-	return sre.BuildPlan(discovery, sre.PlanOptions{Target: target, Image: image, Name: name, BackendURL: cerebroURL, IngestTokenEnv: ingestTokenEnv, Interval: interval}), nil
+	return sre.BuildPlan(discovery, sre.PlanOptions{Target: target, Image: image, Name: name, BackendURL: cerebroURL, IngestTokenEnv: ingestTokenEnv, Provider: provider, DeployID: deployID, Interval: interval}), nil
 }
 
 func durationFlag(cmd *cobra.Command, name string, fallback time.Duration) (time.Duration, error) {