feat(workflows): add auto-disable feature for workflows on consecutive failures

- Introduced `disableOnConsecutiveFailures` option in workflow triggers to automatically disable workflows after reaching a specified failure threshold.
- Updated related functions to handle the new feature, including persistence of the disabled state and reloading of registries.
- Enhanced UI components to support the new option, allowing users to toggle the auto-disable feature in the workflow configuration.

packages/server/registry.js

@@ -36,7 +36,9 @@ import {
   resolveFailureTriggerConfig,
 } from "./trigger-failure.js";
 import { enqueueWorkflowJob } from "./workflow-queue.js";
-import { ensureInitialRevision } from "./workflow-history.js";
+import { ensureInitialRevision, recordRevision } from "./workflow-history.js";
+import { workflowIdFromFile } from "./workflow-normalize.js";
+import { publishReload } from "./control-bus.js";
 
 /**
  * @typedef {{ owner: string, file: string, workflow: any }} WorkflowEntry
@@ -652,6 +654,45 @@ export function createRegistry(server, opts = {}) {
     }
   }
 
+  /**
+   * Persist `enabled: false` for a workflow and reload registries across processes.
+   * @param {string} owner
+   * @param {string} file
+   * @param {string} key
+   */
+  async function disableWorkflowForConsecutiveFailures(owner, file, key) {
+    const content = fsStore.readWorkflowYaml(owner, file);
+    if (content == null) {
+      throw new Error(`workflow file missing for ${key}`);
+    }
+    const doc = yaml.parseDocument(content);
+    if (doc.errors?.length) {
+      throw new Error(doc.errors[0]?.message ?? "invalid yaml");
+    }
+    const parsed = doc.toJSON();
+    if (parsed?.enabled === false) {
+      log.debug({ workflow: key }, "workflow already disabled");
+      return;
+    }
+    doc.set("enabled", false);
+    const nextContent = String(doc);
+    fsStore.writeWorkflowYaml(owner, file, nextContent);
+    await recordRevision({
+      workflowId: workflowIdFromFile(file),
+      owner,
+      file,
+      content: nextContent,
+      reason: "disable-on-consecutive-failures",
+    });
+    reregister();
+    try {
+      await publishReload({ type: "workflows" });
+    } catch {
+      // Redis may be briefly unavailable; local reload already applied.
+    }
+    log.warn({ workflow: key }, "disabled workflow after consecutive failures");
+  }
+
   /**
    * @param {{
    *   key: string,
@@ -688,6 +729,17 @@ export function createRegistry(server, opts = {}) {
       return;
     }
 
+    if (failureConfig.disableOnConsecutiveFailures) {
+      const entry = workflows.get(opts.key);
+      if (entry) {
+        await disableWorkflowForConsecutiveFailures(entry.owner, entry.file, opts.key);
+      } else {
+        log.warn({ workflow: opts.key }, "cannot disable missing workflow entry");
+      }
+    }
+
+    if (!failureConfig.workflowName) return;
+
     const destKey = resolveWorkflowTriggerKey(opts.owner, failureConfig.workflowName);
     const alertData = buildFailureAlertData({
       sourceKey: opts.key,

packages/server/src/api/workflows.js

@@ -69,6 +69,7 @@ function triggerSummary(owner, workflow, nameById) {
       schedule: t?.schedule ?? null,
       onConsecutiveFailures: t?.onConsecutiveFailures ?? null,
       onFailureWorkflow: t?.onFailureWorkflow ?? null,
+      disableOnConsecutiveFailures: t?.disableOnConsecutiveFailures === true,
       auth: isHttp ? authLabel(t?.auth, nameById) : null,
     };
   });

packages/server/trigger-failure.js

@@ -47,12 +47,18 @@ export function resolveFailureTriggerConfig(workflow, owner, runtimeTrigger) {
 
     const threshold = Number(spec.onConsecutiveFailures);
     const workflowName = onFailureWorkflowName(spec);
-    if (!Number.isFinite(threshold) || threshold < 1 || workflowName.length === 0) {
+    const disableOnConsecutiveFailures = isDisableOnConsecutiveFailures(spec);
+    if (
+      !Number.isFinite(threshold) ||
+      threshold < 1 ||
+      (workflowName.length === 0 && !disableOnConsecutiveFailures)
+    ) {
       return null;
     }
     return {
       threshold: Math.floor(threshold),
-      workflowName,
+      workflowName: workflowName.length > 0 ? workflowName : null,
+      disableOnConsecutiveFailures,
     };
   }
   return null;
@@ -66,6 +72,13 @@ function onFailureWorkflowName(trigger) {
   return typeof value === "string" ? value.trim() : "";
 }
 
+/**
+ * @param {Record<string, unknown>} trigger
+ */
+export function isDisableOnConsecutiveFailures(trigger) {
+  return trigger?.disableOnConsecutiveFailures === true;
+}
+
 /**
  * @param {unknown} workflow
  */
@@ -81,12 +94,13 @@ export async function validateWorkflowFailureTriggers(workflow) {
     const hasThreshold =
       trigger.onConsecutiveFailures != null && trigger.onConsecutiveFailures !== "";
     const hasWorkflow = onFailureWorkflowName(trigger).length > 0;
+    const hasDisable = isDisableOnConsecutiveFailures(trigger);
 
-    if (!hasThreshold && !hasWorkflow) continue;
+    if (!hasThreshold && !hasWorkflow && !hasDisable) continue;
 
-    if (!hasThreshold || !hasWorkflow) {
+    if (!hasThreshold || (!hasWorkflow && !hasDisable)) {
       const err = new Error(
-        "onConsecutiveFailures and onFailureWorkflow must both be set on a trigger",
+        "onConsecutiveFailures requires onFailureWorkflow and/or disableOnConsecutiveFailures",
       );
       err.statusCode = 400;
       throw err;