feat: incident report chart

feat: incident report chart

Author: hoalongnatsu

.gitignore

@@ -18,4 +18,5 @@ PROJECT_STRUCTURE.md
 EINO_PATTERNS.md
 runbook-
 tests
+testdata
 runbook-ingest

cmd/main.go

@@ -18,6 +18,7 @@ import (
 	"github.com/VersusControl/versus-incident/pkg/controllers"
 	"github.com/VersusControl/versus-incident/pkg/core"
 	"github.com/VersusControl/versus-incident/pkg/middleware"
+	"github.com/VersusControl/versus-incident/pkg/report"
 	"github.com/VersusControl/versus-incident/pkg/routes"
 	"github.com/VersusControl/versus-incident/pkg/services"
 	"github.com/VersusControl/versus-incident/pkg/storage"
@@ -58,13 +59,11 @@ func main() {
 			MaxIncidents:       cfg.Storage.Redis.MaxIncidents,
 		},
 		Database: storage.DatabaseOptions{
-			Driver:       cfg.Storage.Database.Driver,
-			DSN:          cfg.Storage.Database.DSN,
-			MaxIncidents: cfg.Storage.Database.MaxIncidents,
+			Driver: cfg.Storage.Database.Driver,
+			DSN:    cfg.Storage.Database.DSN,
 		},
 		Postgres: storage.PostgresOptions{
-			DSN:          cfg.Storage.Postgres.DSN,
-			MaxIncidents: cfg.Storage.Postgres.MaxIncidents,
+			DSN: cfg.Storage.Postgres.DSN,
 		},
 	})
 	if err != nil {
@@ -76,6 +75,25 @@ func main() {
 	// alert + record acks).
 	services.SetStorage(store)
 
+	// Wire the incident-report feature (OSS default). The renderer is the
+	// pure-Go in-binary PNG card renderer; an enterprise build may swap it
+	// via services.SetReportRenderer behind the same core.ReportRenderer
+	// seam. Installed unconditionally so report works for webhook incidents
+	// even when the AI agent is disabled.
+	if renderer, rerr := report.NewRenderer(); rerr != nil {
+		log.Printf("report: renderer unavailable: %v", rerr)
+	} else {
+		services.SetReportRenderer(renderer)
+	}
+	// Reuse the operator-configured redaction rules so every text field on
+	// the shared card is scrubbed (defence-in-depth: webhook Content is
+	// attacker-influenced). Built here independent of agent.enable.
+	reportRedactor, redErrs := agent.NewRedactor(cfg.Agent.Redaction.Enable && cfg.Agent.Redaction.RedactIPs, cfg.Agent.Redaction.ExtraPatterns)
+	for _, e := range redErrs {
+		log.Printf("report: redactor warning: %v", e)
+	}
+	services.SetReportRedactor(reportRedactor)
+
 	// Initialize the operator-defined teams / members registry on the
 	// same storage backend. Nil-tolerant: the controller responds with
 	// 503 if construction failed, but a healthy boot wires it in.

cmd/runbook-ingest/main.go

@@ -73,13 +73,11 @@ func main() {
 			MaxIncidents:       cfg.Storage.Redis.MaxIncidents,
 		},
 		Database: storage.DatabaseOptions{
-			Driver:       cfg.Storage.Database.Driver,
-			DSN:          cfg.Storage.Database.DSN,
-			MaxIncidents: cfg.Storage.Database.MaxIncidents,
+			Driver: cfg.Storage.Database.Driver,
+			DSN:    cfg.Storage.Database.DSN,
 		},
 		Postgres: storage.PostgresOptions{
-			DSN:          cfg.Storage.Postgres.DSN,
-			MaxIncidents: cfg.Storage.Postgres.MaxIncidents,
+			DSN: cfg.Storage.Postgres.DSN,
 		},
 	})
 	if err != nil {

config/config.yaml

@@ -164,7 +164,6 @@ storage:
   # database:                    # not yet implemented
   #   driver: postgres           # postgres | mysql | sqlite
   #   dsn: ${DATABASE_DSN}
-  #   max_incidents: 5000
 
 # -----------------------------------------------------------------------------
 # AI agent mode (training | shadow | detect) — opt-in.
@@ -187,7 +186,6 @@ agent:
   batch_max: 5000                 # safety cap per tick
   signal_max_bytes: 65536         # cap on Signal.Raw
 
-
   redaction:
     enable: true
     redact_ips: false             # IPs are usually useful context; opt-in
@@ -258,7 +256,7 @@ agent:
   # New-service grace: when a previously unseen service starts emitting logs
   # in detect mode, it enters an implicit training window so the catalog can
   # build up before AI analysis kicks in. Set to "0" to disable.
-  new_service_grace: 30m            # duration (env: AGENT_NEW_SERVICE_GRACE), ex 10m, 2h, etc.
+  new_service_grace: 24h            # duration (env: AGENT_NEW_SERVICE_GRACE), ex 10m, 2h, etc.
   # Service detection — regexes applied to each log message to extract the
   # service name.
   service_patterns:

go.mod

@@ -20,6 +20,8 @@ require (
 	github.com/jackc/pgx/v5 v5.10.0
 	github.com/slack-go/slack v0.27.0
 	github.com/spf13/viper v1.21.0
+	golang.org/x/image v0.22.0
+	golang.org/x/time v0.6.0
 	google.golang.org/genai v1.62.0
 )
 
@@ -50,7 +52,6 @@ require (
 	go.opentelemetry.io/otel/metric v1.29.0 // indirect
 	go.opentelemetry.io/otel/trace v1.29.0 // indirect
 	golang.org/x/oauth2 v0.30.0 // indirect
-	golang.org/x/time v0.6.0 // indirect
 	google.golang.org/api v0.197.0 // indirect
 	google.golang.org/genproto/googleapis/rpc v0.0.0-20240903143218-8af14fe29dc1 // indirect
 	google.golang.org/grpc v1.66.2 // indirect

go.sum

@@ -386,6 +386,8 @@ golang.org/x/crypto v0.50.0/go.mod h1:3muZ7vA7PBCE6xgPX7nkzzjiUq87kRItoJQM1Yo8S+
 golang.org/x/exp v0.0.0-20190121172915-509febef88a4/go.mod h1:CJ0aWSM057203Lf6IL+f9T1iT9GByDxfZKAQTCR3kQA=
 golang.org/x/exp v0.0.0-20260410095643-746e56fc9e2f h1:W3F4c+6OLc6H2lb//N1q4WpJkhzJCK5J6kUi1NTVXfM=
 golang.org/x/exp v0.0.0-20260410095643-746e56fc9e2f/go.mod h1:J1xhfL/vlindoeF/aINzNzt2Bket5bjo9sdOYzOsU80=
+golang.org/x/image v0.22.0 h1:UtK5yLUzilVrkjMAZAZ34DXGpASN8i8pj8g+O+yd10g=
+golang.org/x/image v0.22.0/go.mod h1:9hPFhljd4zZ1GNSIZJ49sqbp45GKK9t6w+iXvGqZUz4=
 golang.org/x/lint v0.0.0-20181026193005-c67002cb31c3/go.mod h1:UVdnD1Gm6xHRNCYTkRU2/jEulfH38KcIWyp/GAMgvoE=
 golang.org/x/lint v0.0.0-20190227174305-5b3e6a55c961/go.mod h1:wehouNa3lNwaWXcvxsM5YxQ5yQlVC4a0KAMCusXpPoU=
 golang.org/x/lint v0.0.0-20190313153728-d0100b6bd8b3/go.mod h1:6SW0HCj/g11FgYtHlgUYUwCkIfeOF89ocIRzGO/8vkc=

helm/versus-incident/templates/configmap.yaml

@@ -32,7 +32,6 @@ data:
       {{- else if eq (include "versus-incident.storageType" .) "postgres" }}
       postgres:
         dsn: ${POSTGRES_DSN}
-        max_incidents: {{ .Values.storage.postgres.maxIncidents | default 1000 }}
       {{- end }}
 
     alert:

helm/versus-incident/values-ha.yaml

@@ -38,7 +38,6 @@ storage:
     #   --set storage.postgres.dsnExistingSecret=...    (your own Secret)
     dsn: ""
     dsnExistingSecret: ""
-    maxIncidents: 1000
 
 # Enterprise license (REQUIRED — HA is enterprise). Provide one of these:
 #   --set enterprise.licenseKey=...        (inline → chart Secret)

helm/versus-incident/values.yaml

@@ -199,7 +199,6 @@ storage:
     # Reference an existing Secret instead (keeps the DSN out of the release).
     dsnExistingSecret: ""
     dsnExistingSecretKey: "postgres_dsn"
-    maxIncidents: 1000
 
 # AI SRE Agent (opt-in). When agent.enable=false (default) nothing
 # changes: no goroutines start, no AI calls are made, no extra config

pkg/agent/ai/analyze/tools/describe_service.go

@@ -53,6 +53,9 @@ type describePatternEntry struct {
 	Count    int     `json:"count"`
 	Baseline float64 `json:"baseline"`
 	Verdict  string  `json:"verdict,omitempty"`
+	// Sample is the latest REDACTED example log line for the pattern (only the
+	// most recent one, to save tokens in the compact per-service listing).
+	Sample string `json:"sample,omitempty"`
 }
 
 // Invoke implements core.AnalyzeTool.
@@ -91,13 +94,17 @@ func (d DescribeService) Invoke(_ context.Context, args json.RawMessage) (*core.
 		if !strings.EqualFold(p.Service, a.Service) {
 			continue
 		}
-		matches = append(matches, describePatternEntry{
+		entry := describePatternEntry{
 			ID:       p.ID,
 			Template: p.Template,
 			Count:    p.Count,
 			Baseline: p.Baseline,
 			Verdict:  p.Verdict,
-		})
+		}
+		if s := latestSamples(p.Samples, 1); len(s) > 0 {
+			entry.Sample = s[0]
+		}
+		matches = append(matches, entry)
 	}
 	sort.SliceStable(matches, func(i, j int) bool {
 		return matches[i].Count > matches[j].Count