feat(metrics): per-app observability dashboard over execution_logs (A2)

Aggregates the existing execution_logs table (no hot-path instrumentation):
ExecutionLogRepository::summarize_for_app returns counts, error rate, latency
avg/p50/p95 (percentile_cont), by-status/by-source breakdowns, and an hourly
series over a trailing window (clamped 1h..90d). New metrics_api router at
GET /api/v1/admin/apps/{id}/metrics?window= (AppLogRead), and a dashboard
Metrics tab (summary cards + an inline-SVG per-hour bar chart, no JS dep).

Pinned by a manager-core test asserting the exact rollup (percentiles included)
and the empty-app zero case.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

CLAUDE.md

@@ -155,6 +155,6 @@ Environment variables consumed by the `picloud` binary:
 
 ## Out of MVP
 
-This section captured the original MVP cut. Most of it has since **shipped in v1.1.x**: queue triggers, cron triggers, inbound email (`email:receive`, HMAC-webhook model), KV / docs / email / users / HTTP SDKs, function-to-function `invoke()`, and secrets are all live (blueprint §12 Phase 4 table). The **Workflows** track (DAG + nested workflows) has since **shipped** (migrations `0071`/`0072`). A **§9.4 service-interceptor** KV allow/deny slice has since shipped (migration `0073`). **Still deferred:** the rest of §9.4 (data transform, non-kv services, after-hooks, chaining), a metrics/observability dashboard, a raw SMTP-listener ingress, and **multi-node cluster mode**. Don't pre-build for them — but don't make decisions that close the door on them either.
+This section captured the original MVP cut. Most of it has since **shipped in v1.1.x**: queue triggers, cron triggers, inbound email (`email:receive`, HMAC-webhook model), KV / docs / email / users / HTTP SDKs, function-to-function `invoke()`, and secrets are all live (blueprint §12 Phase 4 table). The **Workflows** track (DAG + nested workflows) has since **shipped** (migrations `0071`/`0072`). A **§9.4 service-interceptor** KV allow/deny slice has since shipped (migration `0073`). A read-only **metrics/observability dashboard** has since shipped (A2): `GET /api/v1/admin/apps/{id}/metrics?window=` aggregates the existing `execution_logs` table (counts, error rate, latency avg/p50/p95, an hourly series) via `ExecutionLogRepository::summarize_for_app` + `metrics_api`, surfaced as the dashboard's per-app **Metrics** tab — no hot-path instrumentation. **Still deferred:** the rest of §9.4 (data transform, non-kv services, after-hooks, chaining), a raw SMTP-listener ingress, and **multi-node cluster mode**. Don't pre-build for them — but don't make decisions that close the door on them either.
 
 **Pulled forward to Phase 3 (pre-v1.1):** admin auth, multi-app scoping. The general cross-app **export/import** sharing model stays at v1.3+; note that v1.2 §11.6 shipped a narrower form — **group-owned shared collections** (KV/docs/files/topics/queues) let apps in one subtree share data through the owning group, with the ancestor-chain walk as the isolation boundary. See blueprint §11.5 + design-doc §11.6.

crates/manager-core/src/lib.rs

@@ -78,6 +78,7 @@ pub mod kv_service;
 pub mod log_sink;
 pub mod login_rate_limit;
 pub mod materialize;
+pub mod metrics_api;
 pub mod migrations;
 pub mod module_source;
 pub mod outbox_event_emitter;
@@ -231,6 +232,7 @@ pub use kv_api::{kv_admin_router, KvAdminState};
 pub use kv_repo::{KvRepo, KvRepoError, PostgresKvRepo};
 pub use kv_service::KvServiceImpl;
 pub use log_sink::PostgresExecutionLogSink;
+pub use metrics_api::{metrics_router, MetricsState};
 pub use module_source::PostgresModuleSource;
 pub use outbox_event_emitter::OutboxEventEmitter;
 pub use outbox_repo::{

crates/manager-core/src/metrics_api.rs

@@ -0,0 +1,118 @@
+//! `/api/v1/admin/apps/{id}/metrics` — read-only observability rollup (A2).
+//!
+//! Aggregates the existing `execution_logs` table (counts, error rate, latency
+//! percentiles, an hourly series) for one app over a trailing window. No
+//! hot-path instrumentation: the data plane already persists `duration_ms` /
+//! `status` / `source` / `created_at` per run, so this is a pure read model
+//! behind the dashboard's Metrics tab.
+//!
+//! Capability: `AppLogRead` (the same tier as the per-script log view),
+//! resolved against the app loaded from the path.
+
+use std::sync::Arc;
+
+use axum::extract::{Path, Query, State};
+use axum::response::{IntoResponse, Json, Response};
+use axum::routing::get;
+use axum::{Extension, Router};
+use picloud_shared::{AppId, Principal};
+use serde::Deserialize;
+use serde_json::json;
+
+use crate::app_repo::AppRepository;
+use crate::authz::{require, AuthzDenied, AuthzRepo, Capability};
+use crate::repo::{ExecutionLogRepository, MetricsSummary};
+
+#[derive(Clone)]
+pub struct MetricsState {
+    pub logs: Arc<dyn ExecutionLogRepository>,
+    pub apps: Arc<dyn AppRepository>,
+    pub authz: Arc<dyn AuthzRepo>,
+}
+
+pub fn metrics_router(state: MetricsState) -> Router {
+    Router::new()
+        .route("/apps/{app_id}/metrics", get(summary))
+        .with_state(state)
+}
+
+/// Default rollup window when `?window=` is absent (last 24 hours).
+const DEFAULT_WINDOW_HOURS: i32 = 24;
+
+#[derive(Debug, Deserialize)]
+pub struct MetricsQuery {
+    /// Trailing window in hours (clamped to `[1, 2160]` by the repo).
+    #[serde(default)]
+    pub window: Option<i32>,
+}
+
+async fn summary(
+    State(s): State<MetricsState>,
+    Extension(principal): Extension<Principal>,
+    Path(id_or_slug): Path<String>,
+    Query(q): Query<MetricsQuery>,
+) -> Result<Json<MetricsSummary>, MetricsApiError> {
+    let app_id = resolve_app(&*s.apps, &id_or_slug).await?;
+    require(s.authz.as_ref(), &principal, Capability::AppLogRead(app_id)).await?;
+    let window = q.window.unwrap_or(DEFAULT_WINDOW_HOURS);
+    let summary = s
+        .logs
+        .summarize_for_app(app_id, window)
+        .await
+        .map_err(|e| MetricsApiError::Backend(e.to_string()))?;
+    Ok(Json(summary))
+}
+
+async fn resolve_app(apps: &dyn AppRepository, ident: &str) -> Result<AppId, MetricsApiError> {
+    crate::app_repo::resolve_app(apps, ident)
+        .await
+        .map_err(|e| MetricsApiError::Backend(e.to_string()))?
+        .map(|l| l.app.id)
+        .ok_or(MetricsApiError::AppNotFound)
+}
+
+#[derive(Debug, thiserror::Error)]
+pub enum MetricsApiError {
+    #[error("app not found")]
+    AppNotFound,
+    #[error("forbidden")]
+    Forbidden,
+    #[error("authorization repo error: {0}")]
+    AuthzRepo(String),
+    #[error("metrics backend: {0}")]
+    Backend(String),
+}
+
+impl From<AuthzDenied> for MetricsApiError {
+    fn from(d: AuthzDenied) -> Self {
+        match d {
+            AuthzDenied::Denied => Self::Forbidden,
+            AuthzDenied::Repo(e) => Self::AuthzRepo(e.to_string()),
+        }
+    }
+}
+
+impl IntoResponse for MetricsApiError {
+    fn into_response(self) -> Response {
+        use axum::http::StatusCode;
+        let (status, body) = match &self {
+            Self::AppNotFound => (StatusCode::NOT_FOUND, json!({ "error": self.to_string() })),
+            Self::Forbidden => (StatusCode::FORBIDDEN, json!({ "error": self.to_string() })),
+            Self::AuthzRepo(e) => {
+                tracing::error!(error = %e, "metrics admin authz error");
+                (
+                    StatusCode::INTERNAL_SERVER_ERROR,
+                    json!({ "error": "internal error" }),
+                )
+            }
+            Self::Backend(e) => {
+                tracing::error!(error = %e, "metrics admin backend error");
+                (
+                    StatusCode::INTERNAL_SERVER_ERROR,
+                    json!({ "error": "internal error" }),
+                )
+            }
+        };
+        (status, Json(body)).into_response()
+    }
+}

crates/manager-core/src/repo.rs

@@ -731,6 +731,40 @@ impl ExecutionLogCursor {
     }
 }
 
+/// One `(key, count)` row of a metrics breakdown (by status or by source).
+#[derive(Debug, Clone, serde::Serialize)]
+pub struct MetricsCount {
+    pub key: String,
+    pub count: i64,
+}
+
+/// One hour-bucket of the metrics time series.
+#[derive(Debug, Clone, serde::Serialize)]
+pub struct MetricsBucket {
+    pub ts: chrono::DateTime<chrono::Utc>,
+    pub total: i64,
+    pub errors: i64,
+}
+
+/// Aggregate execution metrics for one app over a trailing window — the read
+/// model behind the observability dashboard. Sourced entirely from
+/// `execution_logs` (no hot-path instrumentation): counts + error rate +
+/// latency percentiles + an hourly series.
+#[derive(Debug, Clone, serde::Serialize)]
+pub struct MetricsSummary {
+    pub window_hours: i32,
+    pub total: i64,
+    pub errors: i64,
+    /// Errors / total in `[0, 1]` (0 when there are no executions).
+    pub error_rate: f64,
+    pub avg_duration_ms: f64,
+    pub p50_duration_ms: f64,
+    pub p95_duration_ms: f64,
+    pub by_status: Vec<MetricsCount>,
+    pub by_source: Vec<MetricsCount>,
+    pub series: Vec<MetricsBucket>,
+}
+
 #[async_trait]
 pub trait ExecutionLogRepository: Send + Sync {
     /// F-P-005: keyset cursor on `(created_at, id)` — `cursor` is the
@@ -743,6 +777,14 @@ pub trait ExecutionLogRepository: Send + Sync {
         cursor: Option<ExecutionLogCursor>,
         source: Option<ExecutionSource>,
     ) -> Result<Vec<ExecutionLog>, ScriptRepositoryError>;
+
+    /// A2: aggregate metrics for `app_id` over the trailing `window_hours`.
+    /// Read-only rollup over `execution_logs`.
+    async fn summarize_for_app(
+        &self,
+        app_id: AppId,
+        window_hours: i32,
+    ) -> Result<MetricsSummary, ScriptRepositoryError>;
 }
 
 pub struct PostgresExecutionLogRepository {
@@ -813,6 +855,129 @@ impl ExecutionLogRepository for PostgresExecutionLogRepository {
 
         Ok(rows.into_iter().map(Into::into).collect())
     }
+
+    async fn summarize_for_app(
+        &self,
+        app_id: AppId,
+        window_hours: i32,
+    ) -> Result<MetricsSummary, ScriptRepositoryError> {
+        // Clamp the window to something sane (1 hour … 90 days) so a crafted
+        // `?window=` can't ask for an unbounded scan.
+        let window_hours = window_hours.clamp(1, 24 * 90);
+        let app = app_id.into_inner();
+
+        // Headline aggregate: totals, error count, latency mean + percentiles.
+        // `<> 'success'` counts every non-success terminal state as an error.
+        let agg = sqlx::query_as::<_, AggRow>(
+            "SELECT count(*)::int8 AS total, \
+                    count(*) FILTER (WHERE status <> 'success')::int8 AS errors, \
+                    COALESCE(avg(duration_ms), 0)::float8 AS avg_ms, \
+                    COALESCE(percentile_cont(0.5) WITHIN GROUP (ORDER BY duration_ms), 0)::float8 AS p50, \
+                    COALESCE(percentile_cont(0.95) WITHIN GROUP (ORDER BY duration_ms), 0)::float8 AS p95 \
+             FROM execution_logs \
+             WHERE app_id = $1 AND created_at >= now() - ($2 * interval '1 hour')",
+        )
+        .bind(app)
+        .bind(window_hours)
+        .fetch_one(&self.pool)
+        .await?;
+
+        let by_status = sqlx::query_as::<_, CountRow>(
+            "SELECT status AS key, count(*)::int8 AS count \
+             FROM execution_logs \
+             WHERE app_id = $1 AND created_at >= now() - ($2 * interval '1 hour') \
+             GROUP BY status ORDER BY count DESC",
+        )
+        .bind(app)
+        .bind(window_hours)
+        .fetch_all(&self.pool)
+        .await?;
+
+        let by_source = sqlx::query_as::<_, CountRow>(
+            "SELECT source AS key, count(*)::int8 AS count \
+             FROM execution_logs \
+             WHERE app_id = $1 AND created_at >= now() - ($2 * interval '1 hour') \
+             GROUP BY source ORDER BY count DESC",
+        )
+        .bind(app)
+        .bind(window_hours)
+        .fetch_all(&self.pool)
+        .await?;
+
+        let series = sqlx::query_as::<_, BucketRow>(
+            "SELECT date_trunc('hour', created_at) AS ts, count(*)::int8 AS total, \
+                    count(*) FILTER (WHERE status <> 'success')::int8 AS errors \
+             FROM execution_logs \
+             WHERE app_id = $1 AND created_at >= now() - ($2 * interval '1 hour') \
+             GROUP BY ts ORDER BY ts",
+        )
+        .bind(app)
+        .bind(window_hours)
+        .fetch_all(&self.pool)
+        .await?;
+
+        // Counts are execution tallies over a bounded window — always far below
+        // f64's 2^53 exact-integer range, so the cast is lossless in practice.
+        #[allow(clippy::cast_precision_loss)]
+        let error_rate = if agg.total > 0 {
+            agg.errors as f64 / agg.total as f64
+        } else {
+            0.0
+        };
+        Ok(MetricsSummary {
+            window_hours,
+            total: agg.total,
+            errors: agg.errors,
+            error_rate,
+            avg_duration_ms: agg.avg_ms,
+            p50_duration_ms: agg.p50,
+            p95_duration_ms: agg.p95,
+            by_status: by_status
+                .into_iter()
+                .map(|r| MetricsCount {
+                    key: r.key,
+                    count: r.count,
+                })
+                .collect(),
+            by_source: by_source
+                .into_iter()
+                .map(|r| MetricsCount {
+                    key: r.key,
+                    count: r.count,
+                })
+                .collect(),
+            series: series
+                .into_iter()
+                .map(|r| MetricsBucket {
+                    ts: r.ts,
+                    total: r.total,
+                    errors: r.errors,
+                })
+                .collect(),
+        })
+    }
+}
+
+#[derive(sqlx::FromRow)]
+struct AggRow {
+    total: i64,
+    errors: i64,
+    avg_ms: f64,
+    p50: f64,
+    p95: f64,
+}
+
+#[derive(sqlx::FromRow)]
+struct CountRow {
+    key: String,
+    count: i64,
+}
+
+#[derive(sqlx::FromRow)]
+struct BucketRow {
+    ts: chrono::DateTime<chrono::Utc>,
+    total: i64,
+    errors: i64,
 }
 
 #[derive(sqlx::FromRow)]

crates/manager-core/tests/metrics_summary.rs

@@ -0,0 +1,150 @@
+//! A2: `ExecutionLogRepository::summarize_for_app` — the read model behind the
+//! observability dashboard. Seeds a known set of `execution_logs` rows for one
+//! app and asserts the rollup: totals, error rate, latency mean + percentiles,
+//! the by-status / by-source breakdowns, and the hourly series. Scoped to a
+//! unique `app_id`, so it is unaffected by any other rows in a shared DB.
+//! Skips cleanly when `DATABASE_URL` is unset.
+
+use picloud_manager_core::repo::{ExecutionLogRepository, PostgresExecutionLogRepository};
+use picloud_shared::AppId;
+use sqlx::postgres::PgPoolOptions;
+use sqlx::PgPool;
+use uuid::Uuid;
+
+async fn pool_or_skip() -> Option<PgPool> {
+    let Ok(url) = std::env::var("DATABASE_URL") else {
+        picloud_test_support::abort_if_db_required("metrics_summary");
+        eprintln!("metrics_summary: DATABASE_URL unset — skipping");
+        return None;
+    };
+    let pool = PgPoolOptions::new()
+        .max_connections(2)
+        .connect(&url)
+        .await
+        .expect("connect to DATABASE_URL");
+    sqlx::migrate!("./migrations")
+        .run(&pool)
+        .await
+        .expect("apply migrations");
+    Some(pool)
+}
+
+async fn make_app(pool: &PgPool) -> Uuid {
+    // `apps.group_id` is NOT NULL, so anchor the app under a fresh root group.
+    let gslug = format!("metrics-g-{}", Uuid::new_v4());
+    let group: (Uuid,) =
+        sqlx::query_as("INSERT INTO groups (slug, name) VALUES ($1, $1) RETURNING id")
+            .bind(&gslug)
+            .fetch_one(pool)
+            .await
+            .expect("group insert");
+    let slug = format!("metrics-{}", Uuid::new_v4());
+    let row: (Uuid,) =
+        sqlx::query_as("INSERT INTO apps (slug, name, group_id) VALUES ($1, $1, $2) RETURNING id")
+            .bind(&slug)
+            .bind(group.0)
+            .fetch_one(pool)
+            .await
+            .expect("app insert");
+    row.0
+}
+
+async fn seed(pool: &PgPool, app: Uuid, duration_ms: i32, status: &str, source: &str) {
+    sqlx::query(
+        "INSERT INTO execution_logs (app_id, request_id, duration_ms, status, source) \
+         VALUES ($1, gen_random_uuid(), $2, $3, $4)",
+    )
+    .bind(app)
+    .bind(duration_ms)
+    .bind(status)
+    .bind(source)
+    .execute(pool)
+    .await
+    .expect("seed execution_logs row");
+}
+
+fn count_of(counts: &[picloud_manager_core::repo::MetricsCount], key: &str) -> i64 {
+    counts.iter().find(|c| c.key == key).map_or(0, |c| c.count)
+}
+
+#[tokio::test]
+async fn summarize_rolls_up_counts_error_rate_and_percentiles() {
+    let Some(pool) = pool_or_skip().await else {
+        return;
+    };
+    let app = make_app(&pool).await;
+
+    // 3 successes (http), 1 error (kv), 1 timeout (cron). Durations chosen so
+    // the percentiles are exact: sorted [10, 20, 30, 50, 100].
+    seed(&pool, app, 10, "success", "http").await;
+    seed(&pool, app, 20, "success", "http").await;
+    seed(&pool, app, 30, "success", "http").await;
+    seed(&pool, app, 100, "error", "kv").await;
+    seed(&pool, app, 50, "timeout", "cron").await;
+
+    let repo = PostgresExecutionLogRepository::new(pool.clone());
+    let s = repo
+        .summarize_for_app(AppId::from(app), 24)
+        .await
+        .expect("summarize");
+
+    assert_eq!(s.total, 5, "total executions");
+    // Every non-success terminal state counts as an error: error + timeout = 2.
+    assert_eq!(s.errors, 2, "error count");
+    assert!(
+        (s.error_rate - 0.4).abs() < 1e-9,
+        "error rate: {}",
+        s.error_rate
+    );
+    // mean of [10,20,30,100,50] = 42.
+    assert!(
+        (s.avg_duration_ms - 42.0).abs() < 1e-6,
+        "avg: {}",
+        s.avg_duration_ms
+    );
+    // percentile_cont(0.5) of the 5 sorted values = 30.
+    assert!(
+        (s.p50_duration_ms - 30.0).abs() < 1e-6,
+        "p50: {}",
+        s.p50_duration_ms
+    );
+    // percentile_cont(0.95): 0.95*(n-1)=3.8 → interpolate 50..100 → 90.
+    assert!(
+        (s.p95_duration_ms - 90.0).abs() < 1e-6,
+        "p95: {}",
+        s.p95_duration_ms
+    );
+
+    assert_eq!(count_of(&s.by_status, "success"), 3);
+    assert_eq!(count_of(&s.by_status, "error"), 1);
+    assert_eq!(count_of(&s.by_status, "timeout"), 1);
+
+    assert_eq!(count_of(&s.by_source, "http"), 3);
+    assert_eq!(count_of(&s.by_source, "kv"), 1);
+    assert_eq!(count_of(&s.by_source, "cron"), 1);
+
+    // All rows land in the current hour bucket; the series sums back to the total.
+    assert!(!s.series.is_empty(), "series has at least one bucket");
+    assert_eq!(s.series.iter().map(|b| b.total).sum::<i64>(), 5);
+    assert_eq!(s.series.iter().map(|b| b.errors).sum::<i64>(), 2);
+}
+
+#[tokio::test]
+async fn summarize_of_an_empty_app_is_all_zero() {
+    let Some(pool) = pool_or_skip().await else {
+        return;
+    };
+    let app = make_app(&pool).await;
+    let repo = PostgresExecutionLogRepository::new(pool.clone());
+    let s = repo
+        .summarize_for_app(AppId::from(app), 24)
+        .await
+        .expect("summarize");
+    assert_eq!(s.total, 0);
+    assert_eq!(s.errors, 0);
+    assert!(
+        (s.error_rate - 0.0).abs() < 1e-9,
+        "no executions → zero rate"
+    );
+    assert!(s.series.is_empty());
+}

crates/picloud/src/lib.rs

@@ -13,8 +13,8 @@ use picloud_manager_core::{
     admin_router, admins_router, api_keys_router, app_members_router, apply_router, apps_api,
     apps_router, attach_principal_if_present, auth_router, dead_letters_router, dev_emails_router,
     docs_admin_router, email_inbound_router, files_admin_router, groups_router, kv_admin_router,
-    migrations, projects_router, rebuild_route_table, require_authenticated, route_admin_router,
-    secrets_router, topics_router, triggers_router, vars_router, AbandonedRepo,
+    metrics_router, migrations, projects_router, rebuild_route_table, require_authenticated,
+    route_admin_router, secrets_router, topics_router, triggers_router, vars_router, AbandonedRepo,
     AdminPrincipalResolver, AdminSessionRepository, AdminState, AdminUserRepository, AdminsState,
     ApiKeyRepository, ApiKeysState, AppDomainRepository, AppMembersRepository, AppMembersState,
     AppRepository, ApplyService, AppsState, AuthState, AuthzRepo, DeadLetterRepo, DeadLettersState,
@@ -23,22 +23,22 @@ use picloud_manager_core::{
     FsGroupFilesRepo, GroupDocsServiceImpl, GroupFilesServiceImpl, GroupKvServiceImpl,
     GroupMembersRepository, GroupPubsubServiceImpl, GroupQueueServiceImpl, GroupRepository,
     GroupsState, HttpConfig, HttpServiceImpl, InboundNonceDedup, KvAdminState, KvServiceImpl,
-    OutboxEventEmitter, OutboxRepo, PostgresAbandonedRepo, PostgresAdminSessionRepository,
-    PostgresAdminUserRepository, PostgresApiKeyRepository, PostgresAppDomainRepository,
-    PostgresAppMembersRepository, PostgresAppRepository, PostgresAppSecretsRepo,
-    PostgresAppUserInvitationRepo, PostgresAppUserPasswordResetRepo, PostgresAppUserRepository,
-    PostgresAppUserRoleRepo, PostgresAppUserSessionRepository, PostgresAppUserVerificationRepo,
-    PostgresDeadLetterRepo, PostgresDeadLetterService, PostgresDocsRepo,
-    PostgresExecutionLogRepository, PostgresExecutionLogSink, PostgresGroupCollectionResolver,
-    PostgresGroupDocsRepo, PostgresGroupKvRepo, PostgresGroupMembersRepository,
-    PostgresGroupQueueRepo, PostgresGroupRepository, PostgresKvRepo, PostgresOutboxRepo,
-    PostgresProjectRepository, PostgresPubsubRepo, PostgresRouteRepository,
-    PostgresScriptRepository, PostgresSecretsRepo, PostgresTopicRepo, PostgresTriggerRepo,
-    PostgresVarsRepo, PrincipalResolver, ProjectRepository, ProjectsState, PubsubServiceImpl,
-    RealtimeAuthorityImpl, RepoResolver, RouteAdminState, SandboxCeiling, ScriptRepository,
-    SecretsConfig, SecretsServiceImpl, SecretsState, SubscriberTokenConfig, TopicRepo, TopicsState,
-    TriggerConfig, TriggerRepo, TriggersState, UsersServiceConfig, UsersServiceImpl, VarsApiState,
-    VarsServiceImpl,
+    MetricsState, OutboxEventEmitter, OutboxRepo, PostgresAbandonedRepo,
+    PostgresAdminSessionRepository, PostgresAdminUserRepository, PostgresApiKeyRepository,
+    PostgresAppDomainRepository, PostgresAppMembersRepository, PostgresAppRepository,
+    PostgresAppSecretsRepo, PostgresAppUserInvitationRepo, PostgresAppUserPasswordResetRepo,
+    PostgresAppUserRepository, PostgresAppUserRoleRepo, PostgresAppUserSessionRepository,
+    PostgresAppUserVerificationRepo, PostgresDeadLetterRepo, PostgresDeadLetterService,
+    PostgresDocsRepo, PostgresExecutionLogRepository, PostgresExecutionLogSink,
+    PostgresGroupCollectionResolver, PostgresGroupDocsRepo, PostgresGroupKvRepo,
+    PostgresGroupMembersRepository, PostgresGroupQueueRepo, PostgresGroupRepository,
+    PostgresKvRepo, PostgresOutboxRepo, PostgresProjectRepository, PostgresPubsubRepo,
+    PostgresRouteRepository, PostgresScriptRepository, PostgresSecretsRepo, PostgresTopicRepo,
+    PostgresTriggerRepo, PostgresVarsRepo, PrincipalResolver, ProjectRepository, ProjectsState,
+    PubsubServiceImpl, RealtimeAuthorityImpl, RepoResolver, RouteAdminState, SandboxCeiling,
+    ScriptRepository, SecretsConfig, SecretsServiceImpl, SecretsState, SubscriberTokenConfig,
+    TopicRepo, TopicsState, TriggerConfig, TriggerRepo, TriggersState, UsersServiceConfig,
+    UsersServiceImpl, VarsApiState, VarsServiceImpl,
 };
 use picloud_orchestrator_core::realtime::DEFAULT_GC_INTERVAL_SECS;
 use picloud_orchestrator_core::routing::{AppDomainTable, RouteTable};
@@ -553,7 +553,7 @@ pub async fn build_app(
 
     let admin = AdminState {
         repo: script_repo.clone(),
-        logs: log_repo,
+        logs: log_repo.clone(),
         apps: apps_repo.clone(),
         authz: authz.clone(),
         validator: engine.clone(),
@@ -810,6 +810,12 @@ pub async fn build_app(
             apps: apps_repo.clone(),
             authz: authz.clone(),
         }))
+        // A2: read-only observability rollup over execution_logs.
+        .merge(metrics_router(MetricsState {
+            logs: log_repo.clone(),
+            apps: apps_repo.clone(),
+            authz: authz.clone(),
+        }))
         // §11.6 M4: read-only operator inspection of a group's shared collections.
         .merge(picloud_manager_core::group_blobs_api::group_blobs_router(
             picloud_manager_core::group_blobs_api::GroupBlobsState {

dashboard/src/lib/AppTabBar.svelte

@@ -39,6 +39,7 @@
 			href: `${base}/apps/${slug}/workflows`,
 			adminOnly: true
 		},
+		{ id: 'metrics', label: 'Metrics', href: `${base}/apps/${slug}/metrics`, adminOnly: true },
 		{
 			id: 'dead-letters',
 			label: 'Dead letters',

dashboard/src/lib/api.ts

@@ -218,6 +218,33 @@ export interface ExecutionLog {
 	created_at: string;
 }
 
+/** A2: one `(key, count)` row of a metrics breakdown. */
+export interface MetricsCount {
+	key: string;
+	count: number;
+}
+
+/** A2: one hour-bucket of the metrics time series. */
+export interface MetricsBucket {
+	ts: string;
+	total: number;
+	errors: number;
+}
+
+/** A2: the observability rollup for one app over a trailing window. */
+export interface MetricsSummary {
+	window_hours: number;
+	total: number;
+	errors: number;
+	error_rate: number;
+	avg_duration_ms: number;
+	p50_duration_ms: number;
+	p95_duration_ms: number;
+	by_status: MetricsCount[];
+	by_source: MetricsCount[];
+	series: MetricsBucket[];
+}
+
 export interface CreateScriptInput {
 	app_id: string;
 	name: string;
@@ -1169,6 +1196,14 @@ export const api = {
 			)
 	},
 
+	metrics: {
+		/** A2: aggregate execution metrics for an app over the trailing window (hours). */
+		summary: (idOrSlug: string, windowHours = 24) =>
+			adminRequest<MetricsSummary>(
+				`/api/v1/admin/apps/${encodeURIComponent(idOrSlug)}/metrics?window=${windowHours}`
+			)
+	},
+
 	deadLetters: {
 		count: (idOrSlug: string) =>
 			adminRequest<{ unresolved: number }>(

dashboard/src/routes/apps/[slug]/+layout.svelte

@@ -72,6 +72,8 @@
 				return 'queues';
 			case 'workflows':
 				return 'workflows';
+			case 'metrics':
+				return 'metrics';
 			case 'dead-letters':
 				return 'dead-letters';
 			default:

dashboard/src/routes/apps/[slug]/metrics/+page.svelte

@@ -0,0 +1,275 @@
+<script lang="ts">
+	import { page } from '$app/state';
+	import { api, ApiError, type MetricsSummary } from '$lib/api';
+
+	let slug = $derived(page.params.slug ?? '');
+	let summary = $state<MetricsSummary | null>(null);
+	let loading = $state(true);
+	let error = $state<string | null>(null);
+	let windowHours = $state(24);
+
+	const windows = [
+		{ label: '24 h', hours: 24 },
+		{ label: '7 d', hours: 24 * 7 },
+		{ label: '30 d', hours: 24 * 30 }
+	];
+
+	async function load() {
+		loading = true;
+		error = null;
+		try {
+			summary = await api.metrics.summary(slug, windowHours);
+		} catch (e) {
+			error = e instanceof ApiError ? e.message : String(e);
+		} finally {
+			loading = false;
+		}
+	}
+
+	$effect(() => {
+		// Re-load whenever the slug or window changes.
+		void slug;
+		void windowHours;
+		void load();
+	});
+
+	function pct(x: number): string {
+		return `${(x * 100).toFixed(1)}%`;
+	}
+	function ms(x: number): string {
+		return `${Math.round(x)} ms`;
+	}
+
+	// Chart geometry. Bars are scaled to the busiest bucket; the error portion
+	// is drawn on top of the success portion so the split reads at a glance.
+	const chartW = 720;
+	const chartH = 160;
+	let maxTotal = $derived(Math.max(1, ...(summary?.series ?? []).map((b) => b.total)));
+	let barStep = $derived(
+		summary && summary.series.length > 0 ? chartW / summary.series.length : chartW
+	);
+	function barW(): number {
+		return Math.max(1, barStep * 0.8);
+	}
+	function fmtBucket(iso: string): string {
+		return new Date(iso).toLocaleString();
+	}
+</script>
+
+<svelte:head>
+	<title>Metrics · {slug} · PiCloud</title>
+</svelte:head>
+
+<div class="panel">
+	<header class="panel-head">
+		<div>
+			<h2>Metrics</h2>
+			<p class="subtitle">Execution rollup over the selected window (from the request log).</p>
+		</div>
+		<div class="controls" role="group" aria-label="Time window">
+			{#each windows as w (w.hours)}
+				<button
+					class:active={windowHours === w.hours}
+					onclick={() => (windowHours = w.hours)}
+					type="button">{w.label}</button
+				>
+			{/each}
+		</div>
+	</header>
+
+	{#if error}
+		<p class="error" role="alert">{error}</p>
+	{:else if loading && !summary}
+		<p class="muted">Loading…</p>
+	{:else if summary}
+		{#if summary.total === 0}
+			<p class="muted">No executions recorded in this window.</p>
+		{:else}
+			<div class="cards">
+				<div class="card">
+					<span class="k">Executions</span>
+					<span class="v">{summary.total}</span>
+				</div>
+				<div class="card" class:bad={summary.errors > 0}>
+					<span class="k">Error rate</span>
+					<span class="v">{pct(summary.error_rate)}</span>
+					<span class="sub">{summary.errors} errors</span>
+				</div>
+				<div class="card">
+					<span class="k">Avg latency</span>
+					<span class="v">{ms(summary.avg_duration_ms)}</span>
+				</div>
+				<div class="card">
+					<span class="k">p50</span>
+					<span class="v">{ms(summary.p50_duration_ms)}</span>
+				</div>
+				<div class="card">
+					<span class="k">p95</span>
+					<span class="v">{ms(summary.p95_duration_ms)}</span>
+				</div>
+			</div>
+
+			<h3>Executions per hour</h3>
+			<div class="chart-wrap">
+				<svg
+					viewBox="0 0 {chartW} {chartH}"
+					width="100%"
+					height={chartH}
+					role="img"
+					aria-label="Executions per hour, error portion highlighted"
+				>
+					{#each summary.series as b, i (b.ts)}
+						{@const total = (b.total / maxTotal) * (chartH - 4)}
+						{@const errs = (b.errors / maxTotal) * (chartH - 4)}
+						{@const x = i * barStep + (barStep - barW()) / 2}
+						<rect
+							{x}
+							y={chartH - total}
+							width={barW()}
+							height={total - errs}
+							class="bar-ok"
+						>
+							<title>{fmtBucket(b.ts)}: {b.total} ({b.errors} errors)</title>
+						</rect>
+						<rect {x} y={chartH - errs} width={barW()} height={errs} class="bar-err">
+							<title>{fmtBucket(b.ts)}: {b.errors} errors</title>
+						</rect>
+					{/each}
+				</svg>
+			</div>
+
+			<div class="breakdowns">
+				<div>
+					<h3>By status</h3>
+					<table>
+						<tbody>
+							{#each summary.by_status as row (row.key)}
+								<tr>
+									<td>{row.key}</td>
+									<td class="num">{row.count}</td>
+								</tr>
+							{/each}
+						</tbody>
+					</table>
+				</div>
+				<div>
+					<h3>By source</h3>
+					<table>
+						<tbody>
+							{#each summary.by_source as row (row.key)}
+								<tr>
+									<td>{row.key}</td>
+									<td class="num">{row.count}</td>
+								</tr>
+							{/each}
+						</tbody>
+					</table>
+				</div>
+			</div>
+		{/if}
+	{/if}
+</div>
+
+<style>
+	.panel-head {
+		display: flex;
+		justify-content: space-between;
+		align-items: flex-start;
+		gap: 1rem;
+		flex-wrap: wrap;
+	}
+	.subtitle {
+		color: var(--text-muted);
+		margin: 0.25rem 0 0;
+		font-size: 0.875rem;
+	}
+	.controls {
+		display: inline-flex;
+		gap: 0.25rem;
+	}
+	.controls button {
+		padding: 0.35rem 0.7rem;
+		border: 1px solid var(--border);
+		background: transparent;
+		color: var(--text-muted);
+		border-radius: var(--radius-sm);
+		font: inherit;
+		font-size: 0.8rem;
+		cursor: pointer;
+	}
+	.controls button.active {
+		color: var(--color-link);
+		border-color: var(--color-link);
+	}
+	.cards {
+		display: flex;
+		flex-wrap: wrap;
+		gap: 0.75rem;
+		margin: 1rem 0 1.5rem;
+	}
+	.card {
+		display: flex;
+		flex-direction: column;
+		gap: 0.15rem;
+		min-width: 8rem;
+		padding: 0.75rem 1rem;
+		border: 1px solid var(--border);
+		border-radius: var(--radius-md);
+	}
+	.card.bad .v {
+		color: var(--color-danger-fg);
+	}
+	.card .k {
+		font-size: 0.72rem;
+		text-transform: uppercase;
+		letter-spacing: 0.04em;
+		color: var(--text-muted);
+	}
+	.card .v {
+		font-size: 1.4rem;
+		font-weight: 600;
+	}
+	.card .sub {
+		font-size: 0.72rem;
+		color: var(--text-muted);
+	}
+	.chart-wrap {
+		overflow-x: auto;
+		border: 1px solid var(--border);
+		border-radius: var(--radius-md);
+		padding: 0.5rem;
+	}
+	.bar-ok {
+		fill: var(--color-link);
+	}
+	.bar-err {
+		fill: var(--color-danger-fg);
+	}
+	.breakdowns {
+		display: flex;
+		flex-wrap: wrap;
+		gap: 2rem;
+		margin-top: 1.5rem;
+	}
+	table {
+		border-collapse: collapse;
+	}
+	td {
+		padding: 0.25rem 0.75rem 0.25rem 0;
+		font-size: 0.875rem;
+	}
+	td.num {
+		text-align: right;
+		font-variant-numeric: tabular-nums;
+	}
+	.muted {
+		color: var(--text-muted);
+	}
+	.error {
+		color: var(--color-danger-fg);
+	}
+	h3 {
+		margin: 1rem 0 0.5rem;
+		font-size: 0.95rem;
+	}
+</style>