Log redacted email provider errors and diagnose sender configuration (#57)

backend/notifications/README.md

@@ -80,8 +80,28 @@ remaining status and request fingerprints expire through a seven-day TTL.
 - Callers authorize recipients and preferences. No direct customer-database access,
   bulk campaigns, provider failover, SMS, push or Telegram delivery is included.
 
-Structured logs report notification IDs, driver, state and attempt count; message
+Structured logs report notification IDs, driver, state and attempt count. Provider
-content, template values, addresses and provider response bodies are excluded.
+HTTP outcomes include `provider_status`; failures also include `provider_error`
+and `provider_message` when available. The adapter reads at most 16 KiB plus an
+overflow byte and extracts only the Resend `name`/`message` or first SendGrid
+`errors[].message`. Diagnostic messages are bounded to 512 Unicode characters
+plus an ellipsis. Credentials, idempotency keys, outgoing request string values,
+email addresses, URLs, long token-like strings and numeric verification/account
+values embedded in the request are redacted; controls and formatting characters
+are normalized. Malformed, incomplete and oversized error
+responses produce a diagnostic fallback instead of raw content. Successful
+response bodies are not logged. Diagnostic fields are transient: they are neither
+persisted in notifications nor exposed in gRPC responses. The existing bounded
+failure codes and delivery/retry decisions remain unchanged. Transport failures
+include a sanitized cause. Provider outcome logging happens before the result
+write so a storage failure cannot discard the only provider diagnostic.
+
+Message content, template values, addresses and raw provider bodies must remain
+excluded from logs. Provider text is sanitized external diagnostic data, never
+an instruction. Use `provider_status`, `provider_error` and `provider_message`
+together: the generic `provider_authentication_failed` category combines HTTP
+401 and 403, including sender/domain authorization failures. A sending-only
+Resend key is valid for sending, but cannot read the provider's log API.
 Monitor pending age, UNKNOWN/FAILED states and repeated outbox/worker warnings.
 
 ## Validate and distribute

backend/notifications/email/driver.go

@@ -33,6 +33,10 @@ type Outcome struct {
 	State      model.State
 	Code       string
 	RetryAfter time.Duration
+	// Provider diagnostics are sanitized operational data, not public failures.
+	HTTPStatus      int
+	ProviderError   string
+	ProviderMessage string
 }
 type Sender interface {
 	Send(context.Context, string, model.Email) Outcome

backend/notifications/email/failure.go

@@ -0,0 +1,95 @@
+package email
+
+import (
+	"encoding/json"
+	"regexp"
+	"sort"
+	"strings"
+	"unicode"
+)
+
+const maxErrorBody = 16 << 10
+const maxErrorMessage = 512
+
+var (
+	errorName = regexp.MustCompile(`^[a-zA-Z0-9_.-]{1,64}$`)
+	addresses = regexp.MustCompile(`[^\s<>"'(),;:]+@[^\s<>"'(),;:]+`)
+	urls      = regexp.MustCompile(`(?i)https?://[^\s<>"']+`)
+	tokens    = regexp.MustCompile(`[a-zA-Z0-9_+/-]{32,}={0,2}`)
+	numbers   = regexp.MustCompile(`[0-9]{4,}`)
+)
+
+// Only the documented human-readable error fields cross the adapter boundary.
+// Full bodies, headers and request data must never reach logs or public failures.
+func providerFailure(body, request []byte, secrets ...string) (string, string) {
+	var dto struct {
+		Name    string `json:"name"`
+		Message string `json:"message"`
+		Errors  []struct {
+			Message string `json:"message"`
+		} `json:"errors"`
+	}
+	if err := json.Unmarshal(body, &dto); err != nil {
+		return "", "Provider error details unavailable: invalid JSON response"
+	}
+	if dto.Message == "" && len(dto.Errors) > 0 {
+		dto.Message = dto.Errors[0].Message
+	}
+	// Providers can echo any scalar string from their request DTOs. Strip those
+	// values before truncation, in addition to credentials and embedded tokens.
+	var payload any
+	if json.Unmarshal(request, &payload) == nil {
+		for _, value := range requestStrings(payload) {
+			secrets = append(secrets, value)
+			// Rendered login/verification codes and numeric account references
+			// can be echoed as fragments of an otherwise longer email body.
+			secrets = append(secrets, numbers.FindAllString(value, -1)...)
+		}
+	}
+	sort.Slice(secrets, func(i, j int) bool { return len(secrets[i]) > len(secrets[j]) })
+	name := safeErrorText(dto.Name, secrets)
+	if !errorName.MatchString(name) {
+		name = ""
+	}
+	message := safeErrorText(dto.Message, secrets)
+	if message == "" {
+		message = "Provider did not include an error message"
+	}
+	return name, message
+}
+
+func requestStrings(value any) []string {
+	var values []string
+	switch value := value.(type) {
+	case string:
+		values = append(values, value)
+	case []any:
+		for _, item := range value {
+			values = append(values, requestStrings(item)...)
+		}
+	case map[string]any:
+		for _, item := range value {
+			values = append(values, requestStrings(item)...)
+		}
+	}
+	return values
+}
+
+func safeErrorText(text string, secrets []string) string {
+	for _, secret := range secrets {
+		if secret != "" {
+			text = strings.ReplaceAll(text, secret, "[redacted]")
+		}
+	}
+	text = addresses.ReplaceAllString(text, "[redacted-address]")
+	text = urls.ReplaceAllString(text, "[redacted-url]")
+	text = tokens.ReplaceAllString(text, "[redacted-token]")
+	text = strings.Join(strings.FieldsFunc(text, func(r rune) bool {
+		return unicode.IsSpace(r) || unicode.IsControl(r) || unicode.Is(unicode.Cf, r)
+	}), " ")
+	runes := []rune(text)
+	if len(runes) > maxErrorMessage {
+		text = string(runes[:maxErrorMessage]) + "…"
+	}
+	return text
+}

backend/notifications/email/failure_test.go

@@ -0,0 +1,90 @@
+package email
+
+import (
+	"context"
+	"encoding/json"
+	"io"
+	"net/http"
+	"net/http/httptest"
+	"strings"
+	"testing"
+	"unicode/utf8"
+
+	"git.sendico.io/tech/sxchnge/notifications/model"
+)
+
+func TestProviderErrorDiagnostics(t *testing.T) {
+	for _, tc := range []struct {
+		name, body, code, message string
+	}{
+		{"resend", `{"name":"validation_error","message":"The sxchnge.com domain is not verified"}`, "validation_error", "The sxchnge.com domain is not verified"},
+		{"sendgrid", `{"errors":[{"message":"The sender identity is not verified","field":"from","help":"private URL"}]}`, "", "The sender identity is not verified"},
+		{"no message", `{"unrecognized":"private response"}`, "", "Provider did not include an error message"},
+		{"malformed", `{"message":"private response"`, "", "Provider error details unavailable: invalid JSON response"},
+		{"wrong type", `{"message":{"secret":"private response"}}`, "", "Provider error details unavailable: invalid JSON response"},
+		{"trailing", `{"message":"private response"} {}`, "", "Provider error details unavailable: invalid JSON response"},
+		{"oversized", `{"message":"` + strings.Repeat("x", maxErrorBody) + `"}`, "", "Provider error details unavailable: incomplete or oversized response"},
+	} {
+		t.Run(tc.name, func(t *testing.T) {
+			server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, _ *http.Request) {
+				w.WriteHeader(http.StatusForbidden)
+				_, _ = io.WriteString(w, tc.body)
+			}))
+			defer server.Close()
+			c := client{http: server.Client(), endpoint: server.URL}
+			out := c.post(context.Background(), "", struct{}{})
+			if out.State != model.Failed || out.Code != "provider_authentication_failed" || out.HTTPStatus != 403 || out.ProviderError != tc.code || out.ProviderMessage != tc.message {
+				t.Fatalf("unexpected provider outcome: %+v", out)
+			}
+		})
+	}
+}
+
+func TestProviderDiagnosticsRedactRequestDataBeforeTruncation(t *testing.T) {
+	key := "re_credential-secret"
+	request := []byte(`{"from":"Sender <sender@example.com>","to":["recipient@example.com"],"personalizations":[{"dynamic_template_data":{"secret":"short-private-value"}}],"subject":"Private subject","text":"Private text","html":"<p>Private HTML</p>","content":[{"value":"Your verification code is 123456; account 9876543210"}]}`)
+	token := strings.Repeat("a", 43)
+	message := "Credential " + key + "; recipient@example.com; Private subject; Private text; <p>Private HTML</p>; short-private-value; https://example.com/verify?token=" + token + "; " + token + "; 123456; 9876543210; job-secret\n\r\x00\u202eforged-log"
+	body, err := json.Marshal(map[string]string{"name": "validation_error", "message": message})
+	if err != nil {
+		t.Fatal(err)
+	}
+	name, result := providerFailure(body, request, key, "job-secret")
+	if name != "validation_error" || !strings.Contains(result, "forged-log") || !strings.Contains(result, "[redacted]") {
+		t.Fatal("lost useful diagnostic", name, result)
+	}
+	for _, forbidden := range []string{key, "recipient@example.com", "Private subject", "Private text", "Private HTML", "short-private-value", "https://", token, "123456", "9876543210", "job-secret", "\n", "\r", "\x00", "\u202e"} {
+		if strings.Contains(result, forbidden) {
+			t.Fatalf("diagnostic contains sensitive value or control %q", forbidden)
+		}
+	}
+	body, err = json.Marshal(map[string]string{"name": "validation_error", "message": strings.Repeat("ą", maxErrorMessage+20) + key})
+	if err != nil {
+		t.Fatal(err)
+	}
+	_, result = providerFailure(body, request, key)
+	if !utf8.ValidString(result) || utf8.RuneCountInString(result) != maxErrorMessage+1 || !strings.HasSuffix(result, "…") || strings.Contains(result, "re_") {
+		t.Fatal("invalid or unbounded diagnostic")
+	}
+}
+
+func TestSuccessfulResponsesDoNotExposeProviderContent(t *testing.T) {
+	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, _ *http.Request) {
+		_, _ = io.WriteString(w, `{"name":"private","message":"private provider response"}`)
+	}))
+	defer server.Close()
+	c := client{http: server.Client(), endpoint: server.URL}
+	if out := c.post(context.Background(), "", struct{}{}); out.State != model.Accepted || out.HTTPStatus != 200 || out.ProviderMessage != "" || out.ProviderError != "" {
+		t.Fatal("success leaked provider content", out)
+	}
+}
+
+func TestTransportFailureHasSanitizedDiagnostic(t *testing.T) {
+	ctx, cancel := context.WithCancel(context.Background())
+	cancel()
+	c := client{http: http.DefaultClient, key: "credential-secret", endpoint: "https://example.com/emails?credential=credential-secret"}
+	out := c.post(ctx, "job-secret", struct{}{})
+	if out.State != model.Unknown || !strings.Contains(out.ProviderMessage, "context canceled") || strings.Contains(out.ProviderMessage, "credential-secret") || strings.Contains(out.ProviderMessage, "https://") {
+		t.Fatal("transport cause missing or unsafe", out)
+	}
+}

backend/notifications/email/http.go

@@ -28,14 +28,15 @@ func (c *client) post(ctx context.Context, key string, payload any) Outcome {
 	}
 	resp, err := c.http.Do(req)
 	if err != nil {
-		return Outcome{State: model.Unknown, Code: "provider_outcome_unknown"}
+		return Outcome{State: model.Unknown, Code: "provider_outcome_unknown", ProviderMessage: safeErrorText(err.Error(), []string{c.key, key})}
 	}
 	defer resp.Body.Close()
-	// Discard a bounded response without logging provider-controlled content.
+	// Never pass raw provider bodies through the outcome or log them directly.
-	_, _ = io.Copy(io.Discard, io.LimitReader(resp.Body, 64<<10))
+	details, readErr := io.ReadAll(io.LimitReader(resp.Body, maxErrorBody+1))
+	outcome := Outcome{HTTPStatus: resp.StatusCode}
 	switch {
 	case resp.StatusCode >= 200 && resp.StatusCode < 300:
-		return Outcome{State: model.Accepted}
+		outcome.State = model.Accepted
 	case resp.StatusCode == 429:
 		delay := time.Minute
 		if n, e := strconv.Atoi(resp.Header.Get("Retry-After")); e == nil && n > 0 && n <= 86400 {
@@ -43,12 +44,20 @@ func (c *client) post(ctx context.Context, key string, payload any) Outcome {
 		} else if date, e := http.ParseTime(resp.Header.Get("Retry-After")); e == nil && time.Until(date) > 0 {
 			delay = time.Until(date)
 		}
-		return Outcome{State: model.RetryWait, Code: "provider_rate_limited", RetryAfter: delay}
+		outcome.State, outcome.Code, outcome.RetryAfter = model.RetryWait, "provider_rate_limited", delay
 	case resp.StatusCode >= 500 || resp.StatusCode == 408:
-		return Outcome{State: model.Unknown, Code: "provider_outcome_unknown"}
+		outcome.State, outcome.Code = model.Unknown, "provider_outcome_unknown"
 	case resp.StatusCode == 401 || resp.StatusCode == 403:
-		return Outcome{State: model.Failed, Code: "provider_authentication_failed"}
+		outcome.State, outcome.Code = model.Failed, "provider_authentication_failed"
 	default:
-		return Outcome{State: model.Failed, Code: "provider_rejected"}
+		outcome.State, outcome.Code = model.Failed, "provider_rejected"
+	}
+	if outcome.State != model.Accepted {
+		if readErr != nil || len(details) > maxErrorBody {
+			outcome.ProviderMessage = "Provider error details unavailable: incomplete or oversized response"
+		} else {
+			outcome.ProviderError, outcome.ProviderMessage = providerFailure(details, body, c.key, key)
 		}
 	}
+	return outcome
+}

backend/notifications/service/worker.go

@@ -5,6 +5,7 @@ import (
 	"errors"
 	"time"
 
+	"git.sendico.io/tech/sxchnge/notifications/email"
 	"git.sendico.io/tech/sxchnge/notifications/model"
 	merrors "git.sendico.io/tech/sxchnge/shared/errors"
 	"go.uber.org/zap"
@@ -23,6 +24,7 @@ func (s *Service) Dispatch(ctx context.Context, id string) (bool, error) {
 	if err != nil {
 		return false, err
 	}
+	var outcome email.Outcome
 	switch {
 	case !n.ExpiresAt.After(now):
 		n.State = model.Expired
@@ -36,7 +38,7 @@ func (s *Service) Dispatch(ctx context.Context, id string) (bool, error) {
 			deadline = n.ExpiresAt
 		}
 		sendCtx, cancel := context.WithDeadline(ctx, deadline)
-		outcome := s.sender.Send(sendCtx, n.ID.Hex(), n.Email)
+		outcome = s.sender.Send(sendCtx, n.ID.Hex(), n.Email)
 		cancel()
 		n.State = outcome.State
 		n.FailureCode = outcome.Code
@@ -62,14 +64,24 @@ func (s *Service) Dispatch(ctx context.Context, id string) (bool, error) {
 	// Persist the result even when shutdown canceled the provider request.
 	finishCtx, cancel := context.WithTimeout(context.WithoutCancel(ctx), 5*time.Second)
 	defer cancel()
-	if err = s.repo.Finish(finishCtx, n); err != nil {
-		return true, err
-	}
 	fields := []zap.Field{zap.String("notification_id", n.ID.Hex()), zap.String("driver", n.Driver), zap.String("state", string(n.State)), zap.Uint32("attempts", n.Attempts), zap.String("failure_code", n.FailureCode)}
+	if outcome.HTTPStatus != 0 {
+		fields = append(fields, zap.Int("provider_status", outcome.HTTPStatus))
+	}
+	if outcome.ProviderError != "" {
+		fields = append(fields, zap.String("provider_error", outcome.ProviderError))
+	}
+	if outcome.ProviderMessage != "" {
+		fields = append(fields, zap.String("provider_message", outcome.ProviderMessage))
+	}
+	// Keep the provider outcome observable even if its local result write fails.
 	if n.State == model.Accepted {
 		s.logger.Info("Email accepted by provider", fields...)
 	} else {
 		s.logger.Warn("Email delivery attempt finished", fields...)
 	}
+	if err = s.repo.Finish(finishCtx, n); err != nil {
+		return true, err
+	}
 	return true, nil
 }

backend/notifications/service/worker_test.go

@@ -2,18 +2,23 @@ package service
 
 import (
 	"context"
+	"errors"
+	"fmt"
 	"testing"
 	"time"
 
 	"git.sendico.io/tech/sxchnge/notifications/email"
 	"git.sendico.io/tech/sxchnge/notifications/model"
+	merrors "git.sendico.io/tech/sxchnge/shared/errors"
 	"go.uber.org/zap"
+	"go.uber.org/zap/zaptest/observer"
 )
 
 type workerRepo struct {
 	Repository
 	job       model.Notification
 	saved     *model.Notification
+	finishErr error
 }
 
 func (r *workerRepo) Recover(context.Context, time.Time) error { return nil }
@@ -21,7 +26,10 @@ func (r *workerRepo) Claim(context.Context, string, time.Time) (*model.Notificat
 	n := r.job
 	return &n, nil
 }
-func (r *workerRepo) Finish(_ context.Context, n *model.Notification) error { r.saved = n; return nil }
+func (r *workerRepo) Finish(_ context.Context, n *model.Notification) error {
+	r.saved = n
+	return r.finishErr
+}
 
 type workerSender struct {
 	calls   int
@@ -62,3 +70,36 @@ func TestWorkerTerminalAndRetryOutcomes(t *testing.T) {
 		})
 	}
 }
+
+func TestWorkerLogsProviderFailureOnce(t *testing.T) {
+	for _, failPersistence := range []bool{false, true} {
+		t.Run(fmt.Sprint(failPersistence), func(t *testing.T) {
+			core, logs := observer.New(zap.WarnLevel)
+			repo := &workerRepo{job: model.Notification{Driver: "resend", ExpiresAt: time.Now().Add(time.Hour), Attempts: 1, Email: model.Email{Address: "private@example.com", Subject: "Private subject", Text: "Private text"}}}
+			if failPersistence {
+				repo.finishErr = merrors.ServiceUnavailable("storage offline")
+			}
+			sender := &workerSender{outcome: email.Outcome{State: model.Failed, Code: "provider_authentication_failed", HTTPStatus: 403, ProviderError: "validation_error", ProviderMessage: "The sxchnge.com domain is not verified"}}
+			svc := New(repo, nil, sender, "resend", zap.New(core))
+			if _, err := svc.Dispatch(context.Background(), "id"); !errors.Is(err, repo.finishErr) {
+				t.Fatal("unexpected persistence error", err)
+			}
+			entries := logs.All()
+			if len(entries) != 1 || entries[0].Message != "Email delivery attempt finished" {
+				t.Fatal("expected one provider outcome log", entries)
+			}
+			fields := entries[0].ContextMap()
+			if fields["provider_status"] != int64(403) || fields["provider_error"] != "validation_error" || fields["provider_message"] != "The sxchnge.com domain is not verified" {
+				t.Fatal("provider diagnostic missing", fields)
+			}
+			if repo.saved.FailureCode != "provider_authentication_failed" {
+				t.Fatal("public failure code changed", repo.saved.FailureCode)
+			}
+			for _, field := range []string{"address", "email", "subject", "text", "html", "parameters"} {
+				if _, ok := fields[field]; ok {
+					t.Fatalf("private notification field %s logged", field)
+				}
+			}
+		})
+	}
+}

docs/signup-facts.md

@@ -38,12 +38,12 @@ and service cancellation still ends it.
 
 The rebuilt worker processed the queued activation message. The live Resend
 provider rejected it with `provider_authentication_failed` (HTTP 401 or 403),
-leaving delivery in `failed` after one attempt. The configured API key/sender
+leaving delivery in `failed` after one attempt. Later diagnostics below identify
-authorization must be corrected before sending another activation email. Rotate
+the unverified sender domain as the actual blocker. Correct domain authorization
-the credential securely with `make email`, wait for Vault Agent to refresh it,
+before activation resend with a new idempotency key. Credential rotation is not
-restart notifications so it reloads its environment, then use activation resend
+required by that error; if a separate rotation is needed, use `make email` and
-with a new idempotency key. Do not replay the failed notification as a success or
+allow Vault Agent to refresh/restart the notification process. Do not replay the
-mark the customer active on their behalf.
+failed notification as a success or mark the customer active on their behalf.
 
 Validation passed: `GOFLAGS=-p=2 make test` (all backend race tests and canonical
 frontend suites), `make lint-backend`, focused provider/identity race tests and
@@ -59,6 +59,39 @@ is blocked by provider authorization as described above. No security scans were
 requested or run. Unrelated notifications runtime settings are excluded from
 the task commits.
 
+### Provider diagnostics follow-up
+
+The same task/PR adds transient provider diagnostics to the worker's existing
+structured outcome log: `provider_status`, `provider_error` and
+`provider_message`. The adapter extracts bounded Resend/SendGrid error fields and
+redacts credentials, request values, addresses, URLs and token/code fragments.
+Transport errors have a sanitized cause; invalid or oversized responses have a
+safe fallback. Diagnostics remain outside stored notification models and public
+gRPC responses. The outcome is logged even when its result write fails. See
+the [notification logging contract](../backend/notifications/README.md).
+
+Runtime inspection confirmed `email.driver: resend`, an API key in Vault, and
+sender `noreply@sxchnge.com`. A live diagnostic notification used Resend's
+documented `delivered+sxchnge-config@resend.dev` test mailbox, without retrying
+the customer's email. The rebuilt worker logged HTTP 403,
+`provider_error: validation_error`, and the sanitized message:
+`The sxchnge.com domain is not verified. Please, add and verify your domain on [redacted-url]`.
+The precise blocker is sender-domain verification, not missing local Resend
+configuration. A sending-only key is appropriate for sending; its inability to
+read Resend's logs API does not explain the original send failure. Verify the
+configured domain in Resend (and its required DNS records), or explicitly choose
+an already verified sender domain. No sender/domain/DNS setting was changed by
+this follow-up; the customer's account remains pending email confirmation.
+
+Follow-up validation passed: `GOFLAGS=-p=2 make test`, `make lint-backend`, and
+`go -C backend/notifications test -race ./email ./service -count=1`. Coverage
+includes Resend/SendGrid messages, malformed/trailing/oversized response bodies,
+credential/content/code redaction, Unicode-safe truncation, sanitized transport
+errors, and exactly one provider outcome log even when persistence fails. The
+notifications image was rebuilt from fresh stable digest pins and deployed
+healthy; the live test verified both the diagnostic log and the unchanged generic
+gRPC failure response. No security scan or real customer email retry was run.
+
 ## Behavior and boundaries
 
 With `api.sendico` configured, the product BFF creates the local pending customer