feat(monitor): add alert monitoring page with KPI dashboard, alert list, and incident management

- Add Tauri plugins for file dialog and filesystem access
- Create Rust backend for processing Grafana alerts (pairing, KPI calculation)
- Add virtualized alert list with invalid toggle and incident attachment
- Add incident manager with create, attach, and detach functionality
- Add KPI dashboard showing coverage ratio, downtime, and invalid rate
- Add date-fns for date formatting

apps/web/src-tauri/src/lib.rs

@@ -1,16 +1,250 @@
+mod commands;
+
+use serde::{Deserialize, Serialize};
+use std::collections::HashMap;
+use thiserror::Error;
+
+#[derive(Debug, Error)]
+pub enum AppError {
+    #[error("IO error: {0}")]
+    Io(#[from] std::io::Error),
+    #[error("JSON parse error: {0}")]
+    Json(#[from] serde_json::Error),
+    #[error("Not found: {0}")]
+    NotFound(String),
+}
+
+impl Serialize for AppError {
+    fn serialize<S>(&self, serializer: S) -> Result<S::Ok, S::Error>
+    where
+        S: serde::ser::Serializer,
+    {
+        serializer.serialize_str(self.to_string().as_ref())
+    }
+}
+
+#[derive(Debug, Clone, Serialize, Deserialize)]
+#[serde(rename_all = "camelCase")]
+pub struct GrafanaAlert {
+    pub id: u64,
+    pub alert_id: u32,
+    #[serde(default)]
+    pub alert_name: Option<String>,
+    #[serde(default)]
+    pub dashboard_uid: Option<String>,
+    pub new_state: String,
+    pub prev_state: String,
+    pub created: u64,
+    pub updated: u64,
+    pub time: u64,
+    pub time_end: u64,
+    pub text: String,
+    #[serde(default)]
+    pub data: Option<AlertData>,
+}
+
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct AlertData {
+    #[serde(default, deserialize_with = "deserialize_metric_values")]
+    pub values: HashMap<String, f64>,
+}
+
+/// Deserialize metric values that may contain "+Inf", "-Inf", "NaN" as strings
+fn deserialize_metric_values<'de, D>(deserializer: D) -> Result<HashMap<String, f64>, D::Error>
+where
+    D: serde::Deserializer<'de>,
+{
+    use serde::de::{MapAccess, Visitor};
+    use std::fmt;
+
+    struct MetricValuesVisitor;
+
+    impl<'de> Visitor<'de> for MetricValuesVisitor {
+        type Value = HashMap<String, f64>;
+
+        fn expecting(&self, formatter: &mut fmt::Formatter) -> fmt::Result {
+            formatter.write_str("a map of string to number or special float strings")
+        }
+
+        fn visit_map<M>(self, mut access: M) -> Result<Self::Value, M::Error>
+        where
+            M: MapAccess<'de>,
+        {
+            let mut map = HashMap::new();
+
+            while let Some((key, value)) = access.next_entry::<String, serde_json::Value>()? {
+                let num = match value {
+                    serde_json::Value::Number(n) => n.as_f64().unwrap_or(0.0),
+                    serde_json::Value::String(s) => match s.as_str() {
+                        "+Inf" | "Inf" | "infinity" | "-Inf" | "-infinity" | "NaN" | "nan" => 0.0,
+                        other => other.parse().unwrap_or(0.0),
+                    },
+                    _ => 0.0,
+                };
+                map.insert(key, num);
+            }
+
+            Ok(map)
+        }
+    }
+
+    deserializer.deserialize_map(MetricValuesVisitor)
+}
+
+#[derive(Debug, Clone, Serialize, Deserialize)]
+#[serde(rename_all = "camelCase")]
+pub struct ProcessedAlert {
+    pub id: u64,
+    pub alert_id: u32,
+    #[serde(default)]
+    pub alert_name: Option<String>,
+    pub new_state: String,
+    pub prev_state: String,
+    pub time: u64,
+    pub time_end: u64,
+    pub text: String,
+    pub duration_ms: u64,
+    pub is_firing: bool,
+    pub is_invalid: bool,
+    pub attached_incident_id: Option<String>,
+    pub values: HashMap<String, f64>,
+}
+
+impl From<GrafanaAlert> for ProcessedAlert {
+    fn from(alert: GrafanaAlert) -> Self {
+        let duration_ms = if alert.time_end > alert.time {
+            alert.time_end - alert.time
+        } else {
+            0
+        };
+
+        let is_firing = alert.new_state == "Alerting" || alert.new_state == "Firing";
+
+        ProcessedAlert {
+            id: alert.id,
+            alert_id: alert.alert_id,
+            alert_name: alert.alert_name,
+            new_state: alert.new_state,
+            prev_state: alert.prev_state,
+            time: alert.time,
+            time_end: alert.time_end,
+            text: alert.text,
+            duration_ms,
+            is_firing,
+            is_invalid: false,
+            attached_incident_id: None,
+            values: alert.data.map(|d| d.values).unwrap_or_default(),
+        }
+    }
+}
+
+#[derive(Debug, Clone, Serialize, Deserialize)]
+#[serde(rename_all = "camelCase")]
+pub struct Incident {
+    pub id: String,
+    pub title: String,
+    pub description: String,
+    pub start_time: u64,
+    pub end_time: u64,
+    pub attached_alert_ids: Vec<u64>,
+}
+
+#[derive(Debug, Clone, Serialize, Deserialize)]
+#[serde(rename_all = "camelCase")]
+pub struct KpiMetrics {
+    pub error_coverage_ratio: f64,
+    pub total_incidents: usize,
+    pub covered_incidents: usize,
+    pub overall_downtime_ms: u64,
+    pub overall_downtime_formatted: String,
+    pub invalid_alert_ratio: f64,
+    pub total_firing_alerts: usize,
+    pub invalid_alerts: usize,
+}
+
+pub fn calculate_kpis(alerts: &[ProcessedAlert], incidents: &[Incident]) -> KpiMetrics {
+    let total_incidents = incidents.len();
+    let covered_incidents = incidents
+        .iter()
+        .filter(|incident| !incident.attached_alert_ids.is_empty())
+        .count();
+    let error_coverage_ratio = if total_incidents > 0 {
+        (covered_incidents as f64 / total_incidents as f64) * 100.0
+    } else {
+        0.0
+    };
+
+    let overall_downtime_ms: u64 = alerts
+        .iter()
+        .filter(|a| a.is_firing && !a.is_invalid)
+        .map(|a| a.duration_ms)
+        .sum();
+
+    let overall_downtime_formatted = format_duration(overall_downtime_ms);
+
+    let total_firing_alerts = alerts.iter().filter(|a| a.is_firing).count();
+    let invalid_alerts = alerts
+        .iter()
+        .filter(|a| a.is_firing && a.is_invalid)
+        .count();
+    let invalid_alert_ratio = if total_firing_alerts > 0 {
+        (invalid_alerts as f64 / total_firing_alerts as f64) * 100.0
+    } else {
+        0.0
+    };
+
+    KpiMetrics {
+        error_coverage_ratio,
+        total_incidents,
+        covered_incidents,
+        overall_downtime_ms,
+        overall_downtime_formatted,
+        invalid_alert_ratio,
+        total_firing_alerts,
+        invalid_alerts,
+    }
+}
+
+fn format_duration(ms: u64) -> String {
+    let seconds = ms / 1000;
+    let minutes = seconds / 60;
+    let hours = minutes / 60;
+    let days = hours / 24;
+
+    if days > 0 {
+        format!("{}d {}h {}m", days, hours % 24, minutes % 60)
+    } else if hours > 0 {
+        format!("{}h {}m {}s", hours, minutes % 60, seconds % 60)
+    } else if minutes > 0 {
+        format!("{}m {}s", minutes, seconds % 60)
+    } else {
+        format!("{}s", seconds)
+    }
+}
+
 #[cfg_attr(mobile, tauri::mobile_entry_point)]
 pub fn run() {
-  tauri::Builder::default()
-    .setup(|app| {
-      if cfg!(debug_assertions) {
-        app.handle().plugin(
-          tauri_plugin_log::Builder::default()
-            .level(log::LevelFilter::Info)
-            .build(),
-        )?;
-      }
-      Ok(())
-    })
-    .run(tauri::generate_context!())
-    .expect("error while running tauri application");
+    tauri::Builder::default()
+        .plugin(tauri_plugin_fs::init())
+        .plugin(tauri_plugin_dialog::init())
+        .setup(|app| {
+            if cfg!(debug_assertions) {
+                app.handle().plugin(
+                    tauri_plugin_log::Builder::default()
+                        .level(log::LevelFilter::Info)
+                        .build(),
+                )?;
+            }
+            Ok(())
+        })
+        .invoke_handler(tauri::generate_handler![
+            commands::load_alerts_from_file,
+            commands::process_alerts_json,
+            commands::calculate_kpis_command,
+            commands::set_alert_invalid,
+            commands::attach_alert_to_incident,
+            commands::detach_alert_from_incident,
+        ])
+        .run(tauri::generate_context!())
+        .expect("error while running tauri application");
 }