Show notes from similar earlier incidents

Each incident gets a signature: the alert name plus the group labels that
say what is broken, minus the ones that only say where it ran (instance,
pod, container, ...). GET /api/incidents/{id}/similar returns resolved
incidents in the same team with the same signature that have notes.

Notes can be marked as the resolution note, "what fixed it", either with a
resolution field on resolve or pinned on a note. Those lead the similar
list, show on the incident page as "Seen before", and the triggered
notification carries the latest one.

Claude-Session: https://claude.ai/code/session_01MMados3BD1oSjevHxbmVqU

internal/api/alertmanager.go

@@ -384,10 +384,10 @@ func openIncident(ctx context.Context, q querier, notify NotifyConfig, teamID in
 
 	var id int64
 	err = q.QueryRowContext(ctx, `
-		INSERT INTO incidents (team_id, group_key, title, group_labels, status, severity, triggered_at, assigned_to)
-		VALUES ($1, $2, $3, $4::jsonb, 'triggered', $5, $6, $7)
+		INSERT INTO incidents (team_id, group_key, title, group_labels, signature, status, severity, triggered_at, assigned_to)
+		VALUES ($1, $2, $3, $4::jsonb, $5, 'triggered', $6, $7, $8)
 		RETURNING id`,
-		teamID, groupKey, title, string(labelsJSON), severity,
+		teamID, groupKey, title, string(labelsJSON), incidentSignature(groupLabels, title), severity,
 		time.Now().Unix(), onCall).Scan(&id)
 	if err != nil {
 		return 0, err

internal/api/incident_store.go

@@ -36,6 +36,10 @@ const (
 	evUnsnoozed      = "unsnoozed"
 	evResolved       = "resolved"
 	evNote           = "note"
+	// evResolutionNote is the note worth finding again: what fixed it. The
+	// similar-incidents lookup and the page lead with these; plain notes are
+	// the working chatter and stay one click away.
+	evResolutionNote = "resolution_note"
 	evDeadmanSilent  = "deadman_silent"
 )
 
@@ -295,6 +299,34 @@ func openIncidentForAlert(ctx context.Context, q querier, alertID int64) (int64,
 	return id, err
 }
 
+// volatileLabels say where a problem ran this time, not what the problem is, so
+// they stay out of the signature. Migration 008's backfill lists the same set.
+var volatileLabels = map[string]bool{
+	"instance": true, "pod": true, "pod_name": true, "pod_ip": true,
+	"container": true, "container_name": true, "endpoint": true,
+}
+
+// incidentSignature identifies "the same problem" across incidents: the alert
+// name plus the stable group labels, sorted. Incidents in one team with equal
+// signatures are what the similar-incidents lookup returns. title stands in for
+// the name when the payload carried no alertname (groupless and dead man's
+// switch incidents).
+func incidentSignature(groupLabels map[string]string, title string) string {
+	name := groupLabels["alertname"]
+	if name == "" {
+		name = title
+	}
+	rest := make([]string, 0, len(groupLabels))
+	for k, v := range groupLabels {
+		if k == "alertname" || volatileLabels[k] {
+			continue
+		}
+		rest = append(rest, k+"="+v)
+	}
+	sort.Strings(rest)
+	return name + "|" + strings.Join(rest, ",")
+}
+
 // incidentTitle renders a human-readable title from Alertmanager's groupLabels,
 // leading with the alert name and appending whatever else the operator grouped
 // by. Falls back to the alert's own name when the payload carried no groupLabels.

internal/api/incidents.go

@@ -3,6 +3,7 @@ package api
 import (
 	"database/sql"
 	"fmt"
+	"io"
 	"net/http"
 	"strconv"
 	"strings"
@@ -237,6 +238,15 @@ func handleIncidentResolve(db *sql.DB) http.HandlerFunc {
 			return
 		}
 		user, _ := userFromContext(r.Context())
+		// The body is optional: clients that predate resolution notes send none.
+		var req struct {
+			Resolution string `json:"resolution"`
+		}
+		if err := decodeJSON(r, &req); err != nil && err != io.EOF {
+			respond(w, http.StatusBadRequest, errResp("invalid request body"))
+			return
+		}
+		req.Resolution = strings.TrimSpace(req.Resolution)
 		if !updateOpenIncident(w, r, db, id,
 			`UPDATE incidents SET status = 'resolved', resolved_at = $1, resolution_source = $2
 			  WHERE id = $3 AND resolved_at IS NULL`,
@@ -252,6 +262,12 @@ func handleIncidentResolve(db *sql.DB) http.HandlerFunc {
 			respond(w, http.StatusInternalServerError, errResp("internal error"))
 			return
 		}
+		if req.Resolution != "" {
+			if err := logEvent(r.Context(), db, id, evResolutionNote, &user.ID, nil, &req.Resolution); err != nil {
+				respond(w, http.StatusInternalServerError, errResp("internal error"))
+				return
+			}
+		}
 		respondIncident(w, r, db, id)
 	}
 }
@@ -421,11 +437,17 @@ func handleCreateNote(db *sql.DB) http.HandlerFunc {
 		}
 		var req struct {
 			Content string `json:"content"`
+			// Pinned files the note as the resolution note: what fixed it.
+			Pinned bool `json:"pinned"`
 		}
 		if err := decodeJSON(r, &req); err != nil {
 			respond(w, http.StatusBadRequest, errResp("invalid request body"))
 			return
 		}
+		noteType := evNote
+		if req.Pinned {
+			noteType = evResolutionNote
+		}
 		if req.Content == "" {
 			respond(w, http.StatusBadRequest, errResp("content is required"))
 			return
@@ -440,7 +462,7 @@ func handleCreateNote(db *sql.DB) http.HandlerFunc {
 		err := db.QueryRowContext(r.Context(), `
 			INSERT INTO incident_events (incident_id, type, user_id, detail, created_at)
 			VALUES ($1, $2, $3, $4, $5)
-			RETURNING id`, id, evNote, user.ID, req.Content, now.Unix()).Scan(&eventID)
+			RETURNING id`, id, noteType, user.ID, req.Content, now.Unix()).Scan(&eventID)
 		if err != nil {
 			respond(w, http.StatusInternalServerError, errResp("internal error"))
 			return
@@ -449,7 +471,7 @@ func handleCreateNote(db *sql.DB) http.HandlerFunc {
 		respond(w, http.StatusCreated, models.IncidentEvent{
 			ID:         eventID,
 			IncidentID: id,
-			Type:       evNote,
+			Type:       noteType,
 			UserID:     &user.ID,
 			Username:   &user.Username,
 			Detail:     &req.Content,
@@ -475,8 +497,8 @@ func handleDeleteNote(db *sql.DB) http.HandlerFunc {
 		user, _ := userFromContext(r.Context())
 		res, err := db.ExecContext(r.Context(), `
 			DELETE FROM incident_events
-			 WHERE id = $1 AND incident_id = $2 AND type = $3 AND user_id = $4`,
-			eventID, id, evNote, user.ID)
+			 WHERE id = $1 AND incident_id = $2 AND type IN ($3, $4) AND user_id = $5`,
+			eventID, id, evNote, evResolutionNote, user.ID)
 		if err != nil {
 			respond(w, http.StatusInternalServerError, errResp("internal error"))
 			return

internal/api/notifier.go

@@ -330,6 +330,16 @@ func deliver(ctx context.Context, db *sql.DB, cfg NotifyConfig, n outboxRow) err
 
 	msg := renderNotification(inc, n, firing, cfg)
 
+	// The page that opens an incident carries what fixed it last time, so the
+	// person woken up starts from that. Best effort: a failed lookup must not
+	// hold back the page itself.
+	if n.kind == notifyTriggered {
+		if sim, err := similarIncidents(ctx, db, n.incidentID, 1); err == nil && len(sim) > 0 && len(sim[0].ResolutionNotes) > 0 {
+			notes := sim[0].ResolutionNotes
+			msg.Message += "\nLast time: " + shorten(derefString(notes[len(notes)-1].Detail), 160)
+		}
+	}
+
 	// An Acknowledge button needs both a user to attribute the acknowledgement
 	// to and a URL the phone can reach. Minted per delivery, so every push
 	// carries its own short-lived token rather than reusing one.
@@ -573,6 +583,17 @@ func plural(n int) string {
 	return "s"
 }
 
+// shorten cuts s to at most n runes, marking the cut, and flattens newlines so
+// a multi-line note stays one line in a push.
+func shorten(s string, n int) string {
+	s = strings.Join(strings.Fields(s), " ")
+	r := []rune(s)
+	if len(r) <= n {
+		return s
+	}
+	return string(r[:n-1]) + "…"
+}
+
 // derefString reads a nullable text column as a plain string.
 func derefString(s *string) string {
 	if s == nil {

internal/api/router.go

@@ -112,6 +112,7 @@ func NewRouter(db *sql.DB, notify NotifyConfig, cfg config.Config) http.Handler
 		r.Get("/api/incidents/{id}", handleGetIncident(db))
 		r.Get("/api/incidents/{id}/alerts", handleIncidentAlerts(db))
 		r.Get("/api/incidents/{id}/timeline", handleIncidentTimeline(db))
+		r.Get("/api/incidents/{id}/similar", handleIncidentSimilar(db))
 		r.Post("/api/incidents/{id}/acknowledge", handleIncidentAcknowledge(db))
 		r.Delete("/api/incidents/{id}/acknowledge", handleIncidentUnacknowledge(db))
 		r.Post("/api/incidents/{id}/resolve", handleIncidentResolve(db))

internal/api/similar.go

@@ -0,0 +1,113 @@
+package api
+
+import (
+	"context"
+	"database/sql"
+	"net/http"
+	"strconv"
+	"time"
+
+	"git.ryuvia.com/niklas/terdut-server/internal/models"
+)
+
+const (
+	similarDefaultLimit = 5
+	similarMaxLimit     = 20
+)
+
+// handleIncidentSimilar lists earlier, resolved incidents in the same team with
+// the same signature that someone left notes on, incidents with a resolution
+// note first. This is the "have we seen this before" answer for a responder
+// looking at a fresh incident; the plain notes are one timeline fetch away.
+func handleIncidentSimilar(db *sql.DB) http.HandlerFunc {
+	return func(w http.ResponseWriter, r *http.Request) {
+		id, ok := incidentIDParam(w, r, db)
+		if !ok {
+			return
+		}
+		limit := similarDefaultLimit
+		if v := r.URL.Query().Get("limit"); v != "" {
+			n, err := strconv.Atoi(v)
+			if err != nil || n < 1 {
+				respond(w, http.StatusBadRequest, errResp("invalid limit"))
+				return
+			}
+			limit = min(n, similarMaxLimit)
+		}
+
+		out, err := similarIncidents(r.Context(), db, id, limit)
+		if err != nil {
+			respond(w, http.StatusInternalServerError, errResp("internal error"))
+			return
+		}
+		respond(w, http.StatusOK, out)
+	}
+}
+
+func similarIncidents(ctx context.Context, q querier, id int64, limit int) ([]models.SimilarIncident, error) {
+	rows, err := q.QueryContext(ctx, `
+		SELECT o.id, o.title, o.triggered_at, o.resolved_at,
+		       (SELECT COUNT(*) FROM incident_events e
+		         WHERE e.incident_id = o.id AND e.type = $3)
+		FROM   incidents i
+		JOIN   incidents o ON o.team_id = i.team_id AND o.signature = i.signature
+		WHERE  i.id = $1 AND o.id <> i.id AND o.resolved_at IS NOT NULL
+		  AND  EXISTS (SELECT 1 FROM incident_events e
+		                WHERE e.incident_id = o.id AND e.type IN ($3, $4))
+		ORDER  BY EXISTS (SELECT 1 FROM incident_events e
+		                   WHERE e.incident_id = o.id AND e.type = $4) DESC,
+		          o.triggered_at DESC
+		LIMIT  $2`, id, limit, evNote, evResolutionNote)
+	if err != nil {
+		return nil, err
+	}
+	defer rows.Close()
+
+	out := []models.SimilarIncident{}
+	ids := []int64{}
+	for rows.Next() {
+		var s models.SimilarIncident
+		var triggered, resolved int64
+		if err := rows.Scan(&s.ID, &s.Title, &triggered, &resolved, &s.NoteCount); err != nil {
+			return nil, err
+		}
+		s.TriggeredAt = time.Unix(triggered, 0).UTC()
+		s.ResolvedAt = time.Unix(resolved, 0).UTC()
+		s.ResolutionNotes = []models.IncidentEvent{}
+		out = append(out, s)
+		ids = append(ids, s.ID)
+	}
+	if err := rows.Err(); err != nil {
+		return nil, err
+	}
+	if len(out) == 0 {
+		return out, nil
+	}
+
+	nrows, err := q.QueryContext(ctx, `
+		SELECT e.id, e.incident_id, e.type, e.user_id, u.username, e.detail, e.created_at
+		FROM   incident_events e
+		LEFT   JOIN users u ON u.id = e.user_id
+		WHERE  e.incident_id = ANY($1) AND e.type = $2
+		ORDER  BY e.created_at, e.id`, ids, evResolutionNote)
+	if err != nil {
+		return nil, err
+	}
+	defer nrows.Close()
+
+	byID := make(map[int64]*models.SimilarIncident, len(out))
+	for i := range out {
+		byID[out[i].ID] = &out[i]
+	}
+	for nrows.Next() {
+		var e models.IncidentEvent
+		var ts int64
+		if err := nrows.Scan(&e.ID, &e.IncidentID, &e.Type, &e.UserID, &e.Username, &e.Detail, &ts); err != nil {
+			return nil, err
+		}
+		e.CreatedAt = time.Unix(ts, 0).UTC()
+		s := byID[e.IncidentID]
+		s.ResolutionNotes = append(s.ResolutionNotes, e)
+	}
+	return out, nrows.Err()
+}

internal/api/similar_test.go

@@ -0,0 +1,99 @@
+package api_test
+
+import (
+	"bytes"
+	"encoding/json"
+	"net/http"
+	"strconv"
+	"testing"
+)
+
+// postGrouped posts a firing webhook whose group labels are exactly the given
+// map, unlike postWebhook, which only ever groups by alertname.
+func postGrouped(t *testing.T, s *ts, fingerprint, startsAt string, groupLabels map[string]string) {
+	t.Helper()
+	labels := map[string]string{}
+	for k, v := range groupLabels {
+		labels[k] = v
+	}
+	payload := map[string]any{
+		"version": "4", "status": "firing",
+		"groupKey":    fingerprint,
+		"groupLabels": groupLabels,
+		"alerts": []map[string]any{
+			amAlert(fingerprint, groupLabels["alertname"], "firing", startsAt, zeroTime, labels),
+		},
+	}
+	data, _ := json.Marshal(payload)
+	resp, err := http.Post(s.URL+"/api/integrations/"+s.ingestKey+"/alertmanager",
+		"application/json", bytes.NewReader(data))
+	if err != nil {
+		t.Fatalf("post webhook: %v", err)
+	}
+	resp.Body.Close()
+}
+
+func similar(t *testing.T, s *ts, id int) []map[string]any {
+	t.Helper()
+	var out []map[string]any
+	decode(t, s.req(t, http.MethodGet, "/api/incidents/"+strconv.Itoa(id)+"/similar", nil), &out)
+	return out
+}
+
+// Same alert on another instance is the same problem; a resolution note left on
+// the first one is what the second one should be shown.
+func TestSimilar_IgnoresVolatileLabelsAndLeadsWithResolutionNote(t *testing.T) {
+	s := newTS(t)
+	postGrouped(t, s, "fp-a", "2026-05-20T10:00:00Z",
+		map[string]string{"alertname": "DiskFull", "instance": "web-1", "job": "node"})
+	s.req(t, http.MethodPost, "/api/incidents/1/resolve",
+		map[string]string{"resolution": "rotated the logs"}).Body.Close()
+
+	postGrouped(t, s, "fp-b", "2026-05-21T10:00:00Z",
+		map[string]string{"alertname": "DiskFull", "instance": "web-2", "job": "node"})
+
+	got := similar(t, s, 2)
+	if len(got) != 1 || int(got[0]["id"].(float64)) != 1 {
+		t.Fatalf("expected incident 1 as the only similar one, got %v", got)
+	}
+	notes := got[0]["resolution_notes"].([]any)
+	if len(notes) != 1 || notes[0].(map[string]any)["detail"] != "rotated the logs" {
+		t.Fatalf("expected the resolution note, got %v", notes)
+	}
+}
+
+// A different stable label (job) is a different problem, and an incident nobody
+// wrote a note on has nothing to show.
+func TestSimilar_DifferentSignatureOrNoNotesIsExcluded(t *testing.T) {
+	s := newTS(t)
+	postGrouped(t, s, "fp-1", "2026-05-20T10:00:00Z",
+		map[string]string{"alertname": "DiskFull", "job": "node"})
+	s.req(t, http.MethodPost, "/api/incidents/1/notes", map[string]string{"content": "checked"}).Body.Close()
+	s.req(t, http.MethodPost, "/api/incidents/1/resolve", nil).Body.Close()
+
+	postGrouped(t, s, "fp-2", "2026-05-20T11:00:00Z",
+		map[string]string{"alertname": "DiskFull", "job": "db"})
+	s.req(t, http.MethodPost, "/api/incidents/2/resolve", nil).Body.Close()
+
+	postGrouped(t, s, "fp-3", "2026-05-21T10:00:00Z",
+		map[string]string{"alertname": "DiskFull", "job": "db"})
+
+	// Incident 3 matches 2 by signature, but 2 has no notes.
+	if got := similar(t, s, 3); len(got) != 0 {
+		t.Fatalf("expected nothing similar to incident 3, got %v", got)
+	}
+}
+
+// An open incident is not "earlier experience" yet, and the incident itself is
+// never its own match.
+func TestSimilar_OpenIncidentsAreNotListed(t *testing.T) {
+	s := newTS(t)
+	postGrouped(t, s, "fp-o1", "2026-05-20T10:00:00Z", map[string]string{"alertname": "Flap"})
+	s.req(t, http.MethodPost, "/api/incidents/1/notes",
+		map[string]any{"content": "still open", "pinned": true}).Body.Close()
+	postGrouped(t, s, "fp-o2", "2026-05-21T10:00:00Z", map[string]string{"alertname": "Flap"})
+
+	if got := similar(t, s, 2); len(got) != 0 {
+		t.Fatalf("expected an open incident not to be listed, got %v", got)
+	}
+}