@agent-native/core 0.141.2 → 0.141.4

package/corpus/templates/factory/AGENTS.md

@@ -3,7 +3,7 @@
 Factory is the visual workspace for building agent factories from incoming work
 to governed delivery. The map is the source of truth; Dispatch owns the shared
 inbox and routing, while Factory owns graph versions, queue state, rules,
-decisions, feedback, agent runs, and provider-action audit records.
+decisions, feedback, agent runs, and provider audit records.
 
 Before building common workspace or agent UI, read `agent-native-toolkit`; use
 `customizing-agent-native` for the configure → compose → eject → propose
@@ -12,7 +12,7 @@ ladder.
 ## Core rules
 
 - Keep app state in SQL via Drizzle and scope every read/write by org and
-  authenticated member. Use actions as the UI, agent, CLI, MCP, and A2A surface.
+  member. Use actions as the UI, agent, CLI, MCP, and A2A surface.
 - Keep migrations additive and portable. These tables intentionally use explicit
   `ownerEmail`/`orgId` columns for org-visible data, not `ownableColumns()`;
   do not call `accessFilter` on them without adding deliberate visibility data.
@@ -31,23 +31,23 @@ ladder.
   automations add 👀 and tag `@builderio` in the source thread; GitHub and Sentry
   clear bugs use the Builder run API. Clips, Design, and Content are always
   owner-managed and never enter autonomous dispatch or PR governance.
-- PR governance requires verified internal BuilderIO membership, a clear bug,
-  passing CI, and handled review feedback. Product or UX implications stay
+- PR governance requires verified BuilderIO membership, a clear bug, passing
+  CI, and handled review feedback. Product or UX implications stay
   manual. Auto-merge additionally requires a verified Factory Builder run.
 - Reuse the existing ai-services GitHub read and Builder execution APIs. Do not
   duplicate GitHub installation/webhook infrastructure in this template.
 - Do not add CRUD routes under `server/routes/api/`; actions are the domain
   surface. Provider callbacks are the only exception and must verify signatures.
-- Factory graph edits create immutable blueprint versions. AI proposes a
-  complete graph with `source=ai`; a person reviews and publishes it through
+- Factory graph edits create immutable blueprint versions. AI proposes a graph
+  with `source=ai`; a person reviews and publishes it through
   the same action surface as manual edits.
 
 ## Application state
 
-- `navigation.view`: `factory` when the visual workspace is open.
+- `navigation.view`: `factory` when the workspace is open.
 - `navigation.factoryId`: selected Factory id when present.
 - `navigation.factoryTab`: `map` | `inbox` | `rules` | `automations` | `settings`.
-- `navigation.factoryNodeId` / `navigation.factoryEdgeId`: selected graph target.
+- `navigation.factoryNodeId` / `navigation.factoryEdgeId`: selected graph item.
 - A selected graph node or edge is part of `navigation` context. Read
   `view-screen` before answering why a route exists or changing the selected
   Factory.
@@ -56,18 +56,19 @@ ladder.
 
 | Action | Purpose |
 | --- | --- |
-| `list-triage-items` / `get-triage-item` | Inspect the queue and evidence. |
+| `list-triage-items` / `get-triage-item` | Inspect queue and evidence. |
 | `poll-slack-channel` | Observe Slack history; never writes to Slack. |
 | `get-slack-feedback-context` | Read the bounded full Slack thread before classification. |
 | `poll-github-sources` / `poll-sentry-errors` | Observe bounded GitHub and Sentry source queues. |
 | `ingest-github-observation` | Store read-only PR evidence. |
 | `list-triage-rules` / `save-triage-rule` | Tune prompt rules and guards. |
-| `evaluate-triage-item` | Append a structured decision. |
+| `evaluate-triage-item` | Append a decision. |
 | `record-triage-feedback` | Capture human correction for learning. |
 | `approve-factory-item` | Explicitly authorize one bounded run. |
 | `start-builder-for-item` | Govern clear-bug dispatch; Slack tags Builder in-thread, other sources use Builder API. |
 | `govern-agent-native-pull-request` | Apply CI, review, internal-author, product, and owner gates to PR approval/merge. |
-| `list-factory-automations` / `save-factory-automation` / `run-factory-automation` | Inspect and edit organization-owned Factory prompts, models, schedules, and runs. |
+| `list-factory-automations` / `save-factory-automation` / `run-factory-automation` | Inspect and edit org-owned Factory prompts, models, schedules, and runs. |
+| `get-factory-automation-health` | Inspect the durable scheduler heartbeat and last scheduler error when runs appear stale. |
 | `suggest-factory-rules` | Mine feedback and fast approvals into proposals. |
 | `reconcile-triage-run` | Persist callback/provider reconciliation. |
 | `list-factories` / `get-factory-graph` | Inspect Factory definitions, graph versions, and live evidence metrics. |
@@ -77,7 +78,7 @@ ladder.
 Rules start in shadow mode; hard guards always apply. Editable organization
 automations execute stored prompts; every external mutation needs a durable run,
 idempotency key, and provider confirmation. The legacy observer is disabled
-when organization automations are seeded.
+once organization automations are seeded.
 
 Use the visual editor for direct blueprint changes. Use the agent chat for
 natural language design, explanations, and proposals; it must preserve a
@@ -87,14 +88,14 @@ actions, never through graph JSON.
 
 ## Scheduler identity
 
-`WORKSPACE_OWNER_EMAIL` is read only at startup to find the deployment
-organization and stamp seeded automation `createdBy`; it is never caller
-identity and must not enter request authorization or credential resolution.
+`WORKSPACE_OWNER_EMAIL` is read only at startup to find the deployment org and
+stamp seeded automation `createdBy`; it is never caller identity and must not
+enter request authorization or credential resolution.
 
 ## Hosting
 
 Production needs `DATABASE_URL`, `WORKSPACE_OWNER_EMAIL`, and
-`FACTORY_PUBLIC_URL`. Builder execution additionally needs the service URL and
-project ID plus workspace-resolved Builder credentials. GitHub and Sentry
+`FACTORY_PUBLIC_URL`. Builder execution additionally needs the service URL,
+project ID, and workspace-resolved Builder credentials. GitHub and Sentry
 polling use workspace-resolved provider credentials. Provider callbacks and
 external writes must remain auditable and fail closed when evidence is partial.

package/corpus/templates/factory/actions/get-factory-automation-health.ts

@@ -0,0 +1,50 @@
+import { defineAction } from "@agent-native/core/action";
+import { getAutomationSchedulerHealth } from "@agent-native/core/jobs";
+import { z } from "zod";
+
+import {
+  requireWorkspaceMember,
+  workspaceMemberIdentityFromContext,
+} from "../server/lib/require-workspace-member.js";
+
+const STALE_AFTER_MS = 3 * 60_000;
+
+export default defineAction({
+  description:
+    "Inspect the Factory recurring-automation scheduler heartbeat and last error. Use this when an automation appears not to be running.",
+  schema: z.object({}),
+  http: { method: "GET" },
+  readOnly: true,
+  run: async (_, context) => {
+    const { orgId } = await requireWorkspaceMember(
+      workspaceMemberIdentityFromContext(context),
+    );
+    const health = await getAutomationSchedulerHealth({
+      appId: "factory",
+      orgId,
+    });
+    if (!health?.lastCheckedAt) {
+      return {
+        status: "no-data" as const,
+        lastCheckedAt: null,
+        lastDispatchedAt: health?.lastDispatchedAt ?? null,
+        lastError: health?.lastError ?? null,
+        runtime: health?.runtime ?? null,
+        staleAfterMs: STALE_AFTER_MS,
+      };
+    }
+    const stale = Date.now() - health.lastCheckedAt > STALE_AFTER_MS;
+    return {
+      status: health.lastError
+        ? ("error" as const)
+        : stale
+          ? ("stale" as const)
+          : ("healthy" as const),
+      lastCheckedAt: health.lastCheckedAt,
+      lastDispatchedAt: health.lastDispatchedAt,
+      lastError: health.lastError,
+      runtime: health.runtime,
+      staleAfterMs: STALE_AFTER_MS,
+    };
+  },
+});

package/corpus/templates/factory/actions/get-triage-config.ts

@@ -1,4 +1,5 @@
 import { defineAction } from "@agent-native/core/action";
+import { getEmailReadiness } from "@agent-native/core/server";
 import { and, eq } from "drizzle-orm";
 import { z } from "zod";
 
@@ -19,6 +20,7 @@ export default defineAction({
     const { orgId } = await requireWorkspaceMember(
       workspaceMemberIdentityFromContext(context),
     );
+    const emailReadiness = await getEmailReadiness();
     const row = (
       await getDb()
         .select()
@@ -41,6 +43,9 @@ export default defineAction({
         sentryProjectSlug: null,
         sentryEnvironment: null,
         lastSentrySeenAt: null,
+        automationFailureAlertsEnabled: true,
+        automationFailureAlertEmail: null,
+        emailReadiness,
       };
     }
     return {
@@ -48,6 +53,9 @@ export default defineAction({
       pollingEnabled: row.pollingEnabled === 1,
       githubPollingEnabled: row.githubPollingEnabled === 1,
       sentryPollingEnabled: row.sentryPollingEnabled === 1,
+      automationFailureAlertsEnabled: row.automationFailureAlertsEnabled === 1,
+      automationFailureAlertEmail: row.automationFailureAlertEmail,
+      emailReadiness,
     };
   },
 });

package/corpus/templates/factory/actions/save-triage-config.ts

@@ -1,4 +1,5 @@
 import { defineAction } from "@agent-native/core/action";
+import { and, eq } from "drizzle-orm";
 import { z } from "zod";
 
 import { getDb } from "../server/db/index.js";
@@ -24,6 +25,8 @@ export default defineAction({
     sentryProjectSlug: z.string().trim().max(200).optional(),
     sentryEnvironment: z.string().trim().max(200).optional(),
     repository: z.string().trim().max(256).optional(),
+    automationFailureAlertsEnabled: z.boolean().default(true),
+    automationFailureAlertEmail: z.string().trim().email().optional(),
   }),
   http: { method: "POST" },
   run: async (
@@ -38,6 +41,8 @@ export default defineAction({
       sentryProjectSlug,
       sentryEnvironment,
       repository,
+      automationFailureAlertsEnabled,
+      automationFailureAlertEmail,
     },
     context,
   ) => {
@@ -46,6 +51,19 @@ export default defineAction({
     );
     const now = new Date().toISOString();
     const db = getDb();
+    const existing = (
+      await db
+        .select({
+          automationFailureAlertEmail: triageConfig.automationFailureAlertEmail,
+        })
+        .from(triageConfig)
+        .where(and(eq(triageConfig.id, orgId), eq(triageConfig.orgId, orgId)))
+        .limit(1)
+    )[0];
+    const persistedAutomationFailureAlertEmail =
+      automationFailureAlertEmail ??
+      existing?.automationFailureAlertEmail ??
+      null;
     await db
       .insert(triageConfig)
       .values({
@@ -60,6 +78,8 @@ export default defineAction({
         sentryProjectSlug: sentryProjectSlug ?? null,
         sentryEnvironment: sentryEnvironment ?? null,
         repository: repository ?? null,
+        automationFailureAlertsEnabled: automationFailureAlertsEnabled ? 1 : 0,
+        automationFailureAlertEmail: persistedAutomationFailureAlertEmail,
         createdAt: now,
         updatedAt: now,
         ownerEmail: userEmail,
@@ -78,6 +98,10 @@ export default defineAction({
           sentryProjectSlug: sentryProjectSlug ?? null,
           sentryEnvironment: sentryEnvironment ?? null,
           repository: repository ?? null,
+          automationFailureAlertsEnabled: automationFailureAlertsEnabled
+            ? 1
+            : 0,
+          automationFailureAlertEmail: persistedAutomationFailureAlertEmail,
           updatedAt: now,
           ownerEmail: userEmail,
         },
@@ -88,6 +112,8 @@ export default defineAction({
       githubPollingEnabled,
       sentryPollingEnabled,
       slackChannelId: slackChannelId ?? null,
+      automationFailureAlertsEnabled,
+      automationFailureAlertEmail: persistedAutomationFailureAlertEmail,
     };
   },
 });

package/corpus/templates/factory/app/i18n/en-US.ts

@@ -178,6 +178,33 @@ const messages = {
     pastRunsEmpty: "No runs recorded yet.",
     saveAutomation: "Save",
     runNow: "Run now",
+    automationHealthTitle: "Scheduler health",
+    automationHealthDescription:
+      "The heartbeat shows whether the deployed recurring-job worker is still checking schedules.",
+    automationHealthHealthy: "Healthy",
+    automationHealthStale: "Stale",
+    automationHealthError: "Error",
+    automationHealthNoData: "No heartbeat",
+    automationLastCheck: "Last check",
+    automationLastDispatch: "Last dispatch",
+    automationRuntime: "Runtime",
+    automationHealthNoDataHint:
+      "No scheduler heartbeat has been recorded yet. Deploy the latest Factory build, then check again.",
+    automationHealthStaleHint:
+      "The scheduler has not checked recently. This usually means the deployed scheduled function or background worker is not running.",
+    automationHealthErrorDetail: "Last scheduler error",
+    automationOpenThread: "Open thread",
+    automationFailureAlertsTitle: "Automation failure alerts",
+    automationFailureAlertsDescription:
+      "Send an in-app alert and email when a Factory automation fails. Repeated identical failures are grouped for 15 minutes.",
+    automationFailureAlertsEnabled: "Email me when automations fail",
+    automationFailureAlertEmail: "Alert email",
+    automationFailureAlertEmailPlaceholder: "Defaults to the workspace owner",
+    automationFailureEmailReadiness: "Email provider",
+    automationDiagnosticsLoadError:
+      "Could not load Factory diagnostics. Check access, deployment, and database migrations, then try again.",
+    automationEmailReadinessHint:
+      "Email uses the organization Dispatch vault. Configure a provider there if this is not ready.",
   },
   factoryCanvas: {
     dragHint: "Drag to arrange",