fix(import): consume complete Gemini Apps timestamp zones

Author: rodboev

docs/chat-import.md

@@ -36,9 +36,11 @@ activity kinds are reported as skipped. Each prompt becomes a one-turn session;
 the importer resolves the timestamp's explicit exported zone instead of using
 the host timezone. Mixed Takeout archives may contain other product activity;
 those explicitly identified cells are ignored. The current parser supports the
-observed English rendering for Gemini Apps cells. Declared non-English or
-otherwise unsupported localized Gemini candidates are reported as unsupported
-before any sessions are emitted.
+observed English rendering for Gemini Apps cells, including the named zones it
+currently recognizes and complete `GMT±H`, `GMT±HH`, `GMT±H:MM`, and
+`GMT±HH:MM` zones; omitted minutes mean zero. Declared non-English or otherwise
+unsupported localized Gemini candidates and malformed zone tokens are reported
+as unsupported before any sessions are emitted.
 
 ## Importing via the UI
 

docs/internal/session-format-sources.md

@@ -273,8 +273,11 @@ Grok section and remove the explicit registry exception in the coverage test.
   were searched 2026-08-01. Google does not publish a versioned Gemini Apps
   activity HTML schema, so markup, labels, timestamp zones, and future record
   kinds remain observed compatibility evidence from sanitized exports. No
-  translated label or timestamp vocabulary is claimed; unsupported localized
-  formats return an explicit compatibility error.
+  translated label or timestamp vocabulary is claimed. Timestamp compatibility
+  includes the existing named zones and complete `GMT±H`, `GMT±HH`,
+  `GMT±H:MM`, and `GMT±HH:MM` forms, with omitted minutes treated as zero;
+  unsupported localized formats and malformed zone tokens return an explicit
+  compatibility error.
 - **Usage and cost:** Takeout activity records expose no authoritative token,
   cache, reasoning, credit, or monetary-cost fields to Agentsview.
 - **Agentsview:** `internal/parser/gemini_apps_takeout.go` and

internal/importer/gemini_apps_test.go

@@ -237,6 +237,45 @@ func TestImportGeminiAppsUnknownZoneDoesNotWriteSession(t *testing.T) {
 	assert.Empty(t, page.Sessions)
 }
 
+func TestImportGeminiAppsWholeHourZonePersistsCorrectTimestamp(t *testing.T) {
+	root := t.TempDir()
+	path := filepath.Join(root, "activity.html")
+	valid := geminiAppsImportPromptedCell(
+		"Jan 2, 2025, 3:04:05 PM GMT+8", "prompt",
+	)
+	malformed := geminiAppsImportPromptedCell(
+		"Jan 3, 2025, 3:04:05 PM GMT+8junk", "malformed",
+	)
+	require.NoError(t, os.WriteFile(path, []byte(geminiAppsImportDocument(valid)), 0o644))
+
+	d := testDB(t)
+	stats, err := ImportGeminiApps(context.Background(), d, root, nil)
+	require.NoError(t, err)
+	assert.Equal(t, 1, stats.Imported)
+
+	page, err := d.ListSessions(context.Background(), db.SessionFilter{Agent: "gemini-apps"})
+	require.NoError(t, err)
+	require.Len(t, page.Sessions, 1)
+	stored, err := d.GetSessionFull(context.Background(), page.Sessions[0].ID)
+	require.NoError(t, err)
+	require.NotNil(t, stored.StartedAt)
+	assert.Equal(t, "2025-01-02T07:04:05Z", *stored.StartedAt)
+	messages, err := d.GetAllMessages(context.Background(), page.Sessions[0].ID)
+	require.NoError(t, err)
+	require.Len(t, messages, 1)
+	assert.Equal(t, *stored.StartedAt, messages[0].Timestamp)
+
+	require.NoError(t, os.WriteFile(path, []byte(geminiAppsImportDocument(malformed)), 0o644))
+	stats, err = ImportGeminiApps(context.Background(), d, root, nil)
+	assert.Error(t, err)
+	assert.Zero(t, stats.Imported)
+	assert.Zero(t, stats.Updated)
+	assert.Zero(t, stats.Errors)
+	page, err = d.ListSessions(context.Background(), db.SessionFilter{Agent: "gemini-apps"})
+	require.NoError(t, err)
+	assert.Len(t, page.Sessions, 1)
+}
+
 func TestImportGeminiAppsPersistenceGuards(t *testing.T) {
 	t.Run("unchanged and superset preserve revision", func(t *testing.T) {
 		root := t.TempDir()

internal/parser/gemini_apps_takeout.go

@@ -22,7 +22,7 @@ type GeminiAppsExportParser interface {
 	ParseGeminiAppsExport(string, func(ParseResult) error) (GeminiAppsParseSummary, error)
 }
 
-var geminiAppsTimestampRE = regexp.MustCompile(`(?i)\b(?:Jan|Feb|Mar|Apr|May|Jun|Jul|Aug|Sep|Oct|Nov|Dec)\s+\d{1,2},\s+\d{4},\s+\d{1,2}:\d{2}:\d{2}[\x{00a0}\x{202f} ]*(?:AM|PM)[\x{00a0}\x{202f} ]+(GMT[+-]\d{1,2}:\d{2}|[A-Za-z]{2,5})\b`)
+var geminiAppsTimestampRE = regexp.MustCompile(`(?i)\b(?:Jan|Feb|Mar|Apr|May|Jun|Jul|Aug|Sep|Oct|Nov|Dec)\s+\d{1,2},\s+\d{4},\s+\d{1,2}:\d{2}:\d{2}[\x{00a0}\x{202f} ]*(?:AM|PM)[\x{00a0}\x{202f} ]+(\S+)`)
 var geminiAppsTimestampLikeRE = regexp.MustCompile(`(?i)(?:\b\d{1,2}\D+\d{4}\b|\b\d{4}\D+\d{1,2}\D+\d{1,2}\b)`)
 
 type geminiAppsFilePlan struct {
@@ -529,16 +529,37 @@ func geminiAppsZoneOffset(zone string) (int, bool) {
 		return -7 * 60 * 60, true
 	}
 	if strings.HasPrefix(zone, "GMT+") || strings.HasPrefix(zone, "GMT-") {
-		parts := strings.Split(strings.TrimPrefix(zone, "GMT"), ":")
-		if len(parts) != 2 {
+		numeric := strings.TrimPrefix(strings.TrimPrefix(zone, "GMT+"), "GMT-")
+		if numeric == "" {
 			return 0, false
 		}
 		sign := 1
-		if strings.HasPrefix(parts[0], "-") {
+		if zone[3] == '-' {
 			sign = -1
 		}
-		hours, e1 := strconv.Atoi(strings.TrimLeft(parts[0], "+-"))
-		minutes, e2 := strconv.Atoi(parts[1])
+		parts := strings.Split(numeric, ":")
+		if len(parts) > 2 || len(parts[0]) < 1 || len(parts[0]) > 2 {
+			return 0, false
+		}
+		for _, part := range parts {
+			if part == "" {
+				return 0, false
+			}
+			for _, char := range part {
+				if char < '0' || char > '9' {
+					return 0, false
+				}
+			}
+		}
+		if len(parts) == 2 && len(parts[1]) != 2 {
+			return 0, false
+		}
+		hours, e1 := strconv.Atoi(parts[0])
+		minutes := 0
+		var e2 error
+		if len(parts) == 2 {
+			minutes, e2 = strconv.Atoi(parts[1])
+		}
 		if e1 != nil || e2 != nil || hours > 23 || minutes > 59 {
 			return 0, false
 		}

internal/parser/gemini_apps_takeout_test.go

@@ -151,6 +151,18 @@ func TestParseGeminiAppsIDsDisambiguateSameTimestampOccurrences(t *testing.T) {
 	assert.Equal(t, ids, repeated)
 }
 
+func TestParseGeminiAppsEquivalentNumericZonesKeepStableIdentity(t *testing.T) {
+	cell := func(zone string) string {
+		return geminiAppsProductCellHTML(
+			"Gemini Apps", "Prompted",
+			"Jan 2, 2025, 3:04:05 PM "+zone, "<p>prompt</p>",
+		)
+	}
+	ids := parseGeminiAppsIDs(t, geminiAppsPromptedDocument(cell("GMT+8")))
+	padded := parseGeminiAppsIDs(t, geminiAppsPromptedDocument(cell("GMT+08:00")))
+	assert.Equal(t, ids["prompt"], padded["prompt"])
+}
+
 func TestParseGeminiAppsRejectsNotPromptedActivityLabel(t *testing.T) {
 	fixture := strings.ReplaceAll(
 		sanitizedGeminiAppsHTML,
@@ -1027,6 +1039,36 @@ func TestParseGeminiAppsPreflightsUnknownZoneFileBeforeCallback(t *testing.T) {
 	assert.Zero(t, callbacks)
 }
 
+func TestParseGeminiAppsPreflightsMalformedZoneBeforeCallback(t *testing.T) {
+	supported := geminiAppsSingleCellHTML(
+		"", "My Activity History", "Prompted",
+		"Jan 2, 2025, 3:04:05 PM EDT", "<p>prompt</p>",
+	)
+	for _, zone := range []string{"GMT+8:3", "GMT+8junk", "GMT+8:30junk", "GMT+24", "GMT+8:60"} {
+		t.Run(zone, func(t *testing.T) {
+			unsupported := geminiAppsSingleCellHTML(
+				"", "My Activity History", "Prompted",
+				"Jan 3, 2025, 3:04:05 PM "+zone, "<p>prompt</p>",
+			)
+			path := filepath.Join(t.TempDir(), "malformed-zone.html")
+			require.NoError(t, os.WriteFile(path, []byte(strings.Replace(
+				supported, "</body>", unsupported+"</body>", 1),
+			), 0o644))
+
+			provider, ok := NewProvider(AgentGeminiApps, ProviderConfig{})
+			require.True(t, ok)
+			exporter := provider.(GeminiAppsExportParser)
+			callbacks := 0
+			_, err := exporter.ParseGeminiAppsExport(path, func(ParseResult) error {
+				callbacks++
+				return nil
+			})
+			assert.Error(t, err)
+			assert.Zero(t, callbacks)
+		})
+	}
+}
+
 func TestParseGeminiAppsSkipsUnknownCompatibleActivityLabels(t *testing.T) {
 	for _, label := range []string{"Nicht Prompted", "Unknown Ereignis", "Prompted extra"} {
 		t.Run(label, func(t *testing.T) {
@@ -1123,6 +1165,9 @@ func TestParseGeminiAppsTimestampUsesExplicitZones(t *testing.T) {
 	}{
 		{"edt", "Jan 2, 2025, 3:04:05 PM EDT", "2025-01-02T19:04:05Z"},
 		{"pst", "Jan 2, 2025, 3:04:05 PM PST", "2025-01-02T23:04:05Z"},
+		{"whole-hour", "Jan 2, 2025, 3:04:05 PM GMT+8", "2025-01-02T07:04:05Z"},
+		{"whole-hour-padded", "Jan 2, 2025, 3:04:05 PM GMT+08:00", "2025-01-02T07:04:05Z"},
+		{"whole-hour-negative", "Jan 2, 2025, 3:04:05 PM GMT-8", "2025-01-02T23:04:05Z"},
 		{"numeric", "Jan 2, 2025, 3:04:05 PM GMT+05:30", "2025-01-02T09:34:05Z"},
 		{"negative-zero", "Jan 2, 2025, 3:04:05 PM GMT-00:30", "2025-01-02T15:34:05Z"},
 	}
@@ -1136,12 +1181,18 @@ func TestParseGeminiAppsTimestampUsesExplicitZones(t *testing.T) {
 		})
 	}
 
-	match := geminiAppsTimestampRE.FindStringSubmatch(
-		"Jan 2, 2025, 3:04:05 PM XYZ",
-	)
-	require.Len(t, match, 2)
-	_, err := parseGeminiAppsTimestamp(match[0], match[1])
-	assert.ErrorContains(t, err, "unsupported")
+	for _, zone := range []string{
+		"GMT+8:3", "GMT+8junk", "GMT+8:30junk", "GMT+24", "GMT+8:60", "XYZ",
+	} {
+		t.Run("reject-"+zone, func(t *testing.T) {
+			match := geminiAppsTimestampRE.FindStringSubmatch(
+				"Jan 2, 2025, 3:04:05 PM " + zone,
+			)
+			require.Len(t, match, 2)
+			_, err := parseGeminiAppsTimestamp(match[0], match[1])
+			assert.ErrorContains(t, err, "unsupported")
+		})
+	}
 }
 
 func TestParseGeminiAppsMissingContentCellIsCountedError(t *testing.T) {