feat(web): create_automation HITL approval card in chat

Closes the create loop in chat: the agent describes user intent → the
drafter sub-LLM produces an AutomationCreate JSON → this card surfaces
a structured preview → approve persists; reject cancels. Edits flow
through chat refinement (re-call with a refined intent), not in-card,
so the card stays simple and the multi-turn checkpointer carries the
context.

Tool UI (components/tool-ui/automation/):
- create-automation.tsx — entry dispatcher + ApprovalCard chrome
  (pending/processing/complete/rejected via useHitlPhase) + SavedCard
  (links to the detail page) + InvalidCard (lists drafter validation
  issues) + ErrorCard (verbatim message). Rejection result is hidden
  because the approval card itself shows the rejected phase inline.
- automation-draft-preview.tsx — structured preview body: name +
  description + goal, triggers (humanised cron + tz + static-input
  keys), plan steps (step_id → action), and a collapsible raw JSON
  for power users.

Wiring:
- components/tool-ui/index.ts — re-export.
- features/chat-messages/timeline/tool-registry/registry.ts —
  register create_automation → CreateAutomationToolUI (dynamic import,
  same pattern as other connector tools).
- contracts/enums/toolIcons.tsx — Workflow icon + "Create automation"
  display name so fallback chrome (and timeline headers) are honest.

Shared util:
- lib/automations/describe-cron.ts — lifted from the route slice's
  lib/ folder since both the dashboard slice and the new approval card
  now render schedule descriptions. Slice imports updated; the now-
  empty slice lib/ folder is gone.

Backend prompt fragments:
- main_agent/system_prompt/.../create_automation/description.md and
  the tool's docstring no longer promise in-card edits. They make the
  refinement path explicit: if the user wants changes after seeing the
  draft, they reply in chat and the agent calls the tool again with a
  refined intent.

v1 deliberately excludes:
- In-card edit form / right-side edit panel — defer until we see real
  demand. The chat refinement loop covers the common case.
- approve_always / persistent allow rules — automations are a single
  artifact, not a repeated mutation, so the "trust this kind of call"
  affordance doesn't apply.

surfsense_backend/app/agents/multi_agent_chat/main_agent/system_prompt/prompts/tools/create_automation/description.md

@@ -1,7 +1,7 @@
 - `create_automation` — Draft and author a new automation. You describe the
   user's intent; a focused drafter inside the tool turns it into the full
-  automation JSON; the user reviews and edits it on an approval card; on
+  automation JSON; the user sees a preview on an approval card and chooses
-  approval it's saved. All three phases happen in a single tool call.
+  approve or reject. All three phases happen in a single tool call.
   - Call when the user wants SurfSense to do something on its own: anything
     recurring or scheduled ("every morning…", "each Monday…", "weekly
     recap…").
@@ -17,13 +17,16 @@
         explicitly ("the Notion parent page id was not specified") so the
         drafter leaves a placeholder.
   - Do NOT prompt the user to confirm before calling — the approval card
-    IS the confirmation. The user can edit any field on the card.
+    IS the confirmation. The card shows a structured preview plus the raw
+    JSON; it offers approve/reject only. If the user wants changes after
+    seeing the draft, they reply in chat and you call this tool again with
+    a refined `intent` — that's the edit path.
   - Returns:
     - `{status: "saved", automation_id, name}` — confirm briefly to the
       user ("Saved as automation #N — runs <when>."). Don't dump JSON back.
     - `{status: "rejected", message}` — the user declined on the card.
       Acknowledge once ("Understood, I didn't create it.") and stop. Do
-      NOT retry or pitch variants.
+      NOT retry or pitch variants without a fresh user request.
     - `{status: "invalid", issues, raw?}` — drafting/validation failed
       before the card was shown. Read the issues, refine your `intent`
       with the missing details, call again.

surfsense_backend/app/agents/multi_agent_chat/main_agent/tools/automation/create.py

@@ -66,9 +66,11 @@ def create_create_automation_tool(
         names, …) it needs.
 
         The tool drafts the full automation JSON internally, shows the user
-        an approval card for review, and persists on approval. Do NOT
+        a structured preview on an approval card, and persists on approval.
-        prompt the user to confirm before calling — the card IS the
+        The card supports approve/reject only — if the user wants edits
-        confirmation. The user can edit any field there.
+        after seeing the draft, they say so in chat and you call this tool
+        again with a refined intent. Do NOT prompt the user to confirm
+        before calling — the card IS the confirmation.
 
         Args:
             intent: Concrete restatement of the user's request. Include

surfsense_web/app/dashboard/[search_space_id]/automations/[automation_id]/components/trigger-card.tsx

@@ -6,8 +6,8 @@ import { updateTriggerMutationAtom } from "@/atoms/automations/automations-mutat
 import { Button } from "@/components/ui/button";
 import { Switch } from "@/components/ui/switch";
 import type { Trigger } from "@/contracts/types/automation.types";
+import { describeCron } from "@/lib/automations/describe-cron";
 import { formatRelativeDate } from "@/lib/format-date";
-import { describeCron } from "../../lib/describe-cron";
 import { DeleteTriggerDialog } from "./delete-trigger-dialog";
 
 interface TriggerCardProps {

surfsense_web/app/dashboard/[search_space_id]/automations/components/automation-triggers-summary.tsx

@@ -1,7 +1,7 @@
 "use client";
 import { CalendarClock, Pause } from "lucide-react";
 import type { Trigger } from "@/contracts/types/automation.types";
-import { describeCron } from "../lib/describe-cron";
+import { describeCron } from "@/lib/automations/describe-cron";
 
 interface AutomationTriggersSummaryProps {
 	triggers: Trigger[];

surfsense_web/components/tool-ui/automation/automation-draft-preview.tsx

@@ -0,0 +1,183 @@
+"use client";
+import { CalendarClock, ChevronDown, ChevronRight, ListOrdered, Target } from "lucide-react";
+import { useState } from "react";
+import { describeCron } from "@/lib/automations/describe-cron";
+
+interface DraftTrigger {
+	type: string;
+	params: Record<string, unknown>;
+	static_inputs: Record<string, unknown>;
+	enabled: boolean;
+}
+
+interface DraftPlanStep {
+	step_id: string;
+	action: string;
+	when?: string | null;
+}
+
+interface AutomationDraft {
+	name: string;
+	description?: string | null;
+	definition: {
+		goal?: string | null;
+		plan: DraftPlanStep[];
+	};
+	triggers: DraftTrigger[];
+}
+
+interface AutomationDraftPreviewProps {
+	draft: AutomationDraft;
+	/** Full unmodified args dict — surfaced as the "raw JSON" escape hatch. */
+	raw: Record<string, unknown>;
+}
+
+/**
+ * Structured preview of a drafted automation rendered inside the chat
+ * approval card.
+ *
+ * Three layers, top to bottom:
+ *   1. Name + description (and goal when present).
+ *   2. Triggers — humanised cron string + timezone + static_inputs hint.
+ *   3. Plan steps — ordered list of ``step_id → action``.
+ *
+ * A "View raw JSON" toggle reveals the full payload for power users who
+ * want to inspect every field; it's collapsed by default so the card
+ * stays scannable for the common case.
+ */
+export function AutomationDraftPreview({ draft, raw }: AutomationDraftPreviewProps) {
+	const [showRaw, setShowRaw] = useState(false);
+
+	return (
+		<div className="space-y-4 text-sm">
+			<div className="space-y-1">
+				<p className="font-medium text-foreground">{draft.name}</p>
+				{draft.description && <p className="text-xs text-muted-foreground">{draft.description}</p>}
+			</div>
+
+			{draft.definition.goal && (
+				<Section icon={Target} label="Goal">
+					<p className="text-xs text-foreground">{draft.definition.goal}</p>
+				</Section>
+			)}
+
+			<Section icon={CalendarClock} label={`Triggers · ${draft.triggers.length}`}>
+				{draft.triggers.length === 0 ? (
+					<p className="text-xs text-muted-foreground">
+						No triggers — automation will need one before it can run.
+					</p>
+				) : (
+					<ul className="space-y-1.5">
+						{draft.triggers.map((trigger) => (
+							<li
+								key={triggerKey(trigger)}
+								className="rounded-md border border-border/60 bg-background/50 px-3 py-2 text-xs"
+							>
+								<TriggerLine trigger={trigger} />
+							</li>
+						))}
+					</ul>
+				)}
+			</Section>
+
+			<Section
+				icon={ListOrdered}
+				label={`Plan · ${draft.definition.plan.length} step${draft.definition.plan.length === 1 ? "" : "s"}`}
+			>
+				<ol className="space-y-1 text-xs">
+					{draft.definition.plan.map((step, idx) => (
+						<li key={step.step_id} className="flex items-start gap-2">
+							<span className="inline-flex h-4 w-4 items-center justify-center rounded-full bg-muted text-[10px] font-medium text-muted-foreground shrink-0 mt-0.5">
+								{idx + 1}
+							</span>
+							<div className="min-w-0">
+								<span className="font-medium text-foreground">{step.step_id}</span>
+								<span className="text-muted-foreground"> → </span>
+								<code className="font-mono text-muted-foreground">{step.action}</code>
+								{step.when && <span className="ml-2 text-muted-foreground">when {step.when}</span>}
+							</div>
+						</li>
+					))}
+				</ol>
+			</Section>
+
+			<button
+				type="button"
+				onClick={() => setShowRaw((value) => !value)}
+				className="inline-flex items-center gap-1 text-xs text-muted-foreground hover:text-foreground"
+			>
+				{showRaw ? (
+					<ChevronDown className="h-3 w-3" aria-hidden />
+				) : (
+					<ChevronRight className="h-3 w-3" aria-hidden />
+				)}
+				{showRaw ? "Hide raw JSON" : "View raw JSON"}
+			</button>
+			{showRaw && (
+				<pre className="rounded-md bg-muted/40 px-3 py-2 text-[11px] font-mono text-foreground overflow-x-auto whitespace-pre-wrap break-words max-h-72">
+					{JSON.stringify(raw, null, 2)}
+				</pre>
+			)}
+		</div>
+	);
+}
+
+/**
+ * Stable key derived from the trigger's identifying fields. Drafts are
+ * static snapshots so collisions only happen if the LLM emits two literally
+ * identical triggers — harmless in practice.
+ */
+function triggerKey(trigger: DraftTrigger): string {
+	const cron = typeof trigger.params.cron === "string" ? trigger.params.cron : "";
+	const tz = typeof trigger.params.timezone === "string" ? trigger.params.timezone : "";
+	return `${trigger.type}|${cron}|${tz}`;
+}
+
+function TriggerLine({ trigger }: { trigger: DraftTrigger }) {
+	if (trigger.type === "schedule") {
+		const cron = typeof trigger.params.cron === "string" ? trigger.params.cron : undefined;
+		const tz = typeof trigger.params.timezone === "string" ? trigger.params.timezone : "UTC";
+		const human = cron ? describeCron(cron) : "Schedule";
+		const staticKeys = Object.keys(trigger.static_inputs ?? {});
+		return (
+			<div className="space-y-1">
+				<div className="flex items-center gap-2 flex-wrap">
+					<span className="font-medium text-foreground">{human}</span>
+					<span className="text-muted-foreground">· {tz}</span>
+					{!trigger.enabled && (
+						<span className="rounded-md border border-border/60 px-1.5 py-0.5 text-[10px] text-muted-foreground">
+							Disabled
+						</span>
+					)}
+				</div>
+				{cron && <code className="font-mono text-muted-foreground">{cron}</code>}
+				{staticKeys.length > 0 && (
+					<p className="text-muted-foreground">
+						Static inputs: <span className="text-foreground">{staticKeys.join(", ")}</span>
+					</p>
+				)}
+			</div>
+		);
+	}
+	return <span className="capitalize text-foreground">{trigger.type}</span>;
+}
+
+function Section({
+	icon: Icon,
+	label,
+	children,
+}: {
+	icon: typeof Target;
+	label: string;
+	children: React.ReactNode;
+}) {
+	return (
+		<div className="space-y-1.5">
+			<div className="flex items-center gap-1.5 text-[11px] font-medium text-muted-foreground uppercase tracking-wider">
+				<Icon className="h-3 w-3" aria-hidden />
+				{label}
+			</div>
+			{children}
+		</div>
+	);
+}

surfsense_web/components/tool-ui/automation/create-automation.tsx

@@ -0,0 +1,328 @@
+"use client";
+
+import type { ToolCallMessagePartProps } from "@assistant-ui/react";
+import { useAtomValue } from "jotai";
+import { CornerDownLeftIcon, ExternalLink, Workflow } from "lucide-react";
+import Link from "next/link";
+import { useCallback, useEffect, useMemo } from "react";
+import { activeSearchSpaceIdAtom } from "@/atoms/search-spaces/search-space-query.atoms";
+import { TextShimmerLoader } from "@/components/prompt-kit/loader";
+import { Button } from "@/components/ui/button";
+import type { HitlDecision, InterruptResult } from "@/features/chat-messages/hitl";
+import { isInterruptResult, useHitlDecision, useHitlPhase } from "@/features/chat-messages/hitl";
+import { AutomationDraftPreview } from "./automation-draft-preview";
+
+// ----------------------------------------------------------------------------
+// Result discrimination — mirrors the backend return shapes in
+// app/agents/multi_agent_chat/main_agent/tools/automation/create.py.
+// ----------------------------------------------------------------------------
+
+type AutomationCreateContext = {
+	search_space_id?: number;
+};
+
+interface SavedResult {
+	status: "saved";
+	automation_id: number;
+	name: string;
+}
+
+interface RejectedResult {
+	status: "rejected";
+	message?: string;
+}
+
+interface InvalidResult {
+	status: "invalid";
+	issues: string[];
+	raw?: unknown;
+}
+
+interface ErrorResult {
+	status: "error";
+	message: string;
+}
+
+type CreateAutomationResult =
+	| InterruptResult<AutomationCreateContext>
+	| SavedResult
+	| RejectedResult
+	| InvalidResult
+	| ErrorResult;
+
+function hasStatus(value: unknown, status: string): boolean {
+	return (
+		typeof value === "object" &&
+		value !== null &&
+		"status" in value &&
+		(value as { status: unknown }).status === status
+	);
+}
+
+// ----------------------------------------------------------------------------
+// Approval card — pending → processing → complete / rejected.
+//
+// v1 deliberately supports only approve/reject. The drafted JSON is complex
+// (full plan + triggers) and we already have a multi-turn refinement path via
+// chat ("make it run at 10am instead" → the agent re-calls the tool with a
+// refined intent). An in-card edit form would duplicate that flow and add UX
+// surface area we don't need yet — leave it for the raw-JSON path on the
+// detail page.
+// ----------------------------------------------------------------------------
+
+interface ApprovalCardProps {
+	args: Record<string, unknown>;
+	interruptData: InterruptResult<AutomationCreateContext>;
+	onDecision: (decision: HitlDecision) => void;
+}
+
+function ApprovalCard({ args, interruptData, onDecision }: ApprovalCardProps) {
+	const { phase, setProcessing, setRejected } = useHitlPhase(interruptData);
+
+	const reviewConfig = interruptData.review_configs[0];
+	const allowedDecisions = reviewConfig?.allowed_decisions ?? ["approve", "reject"];
+	const canApprove = allowedDecisions.includes("approve");
+	const canReject = allowedDecisions.includes("reject");
+
+	const draft = useMemo(() => extractDraft(args), [args]);
+
+	const handleApprove = useCallback(() => {
+		if (phase !== "pending" || !canApprove) return;
+		setProcessing();
+		onDecision({
+			type: "approve",
+			edited_action: {
+				name: interruptData.action_requests[0]?.name ?? "create_automation",
+				args,
+			},
+		});
+	}, [phase, canApprove, setProcessing, onDecision, interruptData, args]);
+
+	const handleReject = useCallback(() => {
+		if (phase !== "pending" || !canReject) return;
+		setRejected();
+		onDecision({ type: "reject", message: "User rejected the automation draft." });
+	}, [phase, canReject, setRejected, onDecision]);
+
+	useEffect(() => {
+		const handler = (e: KeyboardEvent) => {
+			if (e.key === "Enter" && !e.shiftKey && !e.ctrlKey && !e.metaKey) {
+				handleApprove();
+			}
+		};
+		window.addEventListener("keydown", handler);
+		return () => window.removeEventListener("keydown", handler);
+	}, [handleApprove]);
+
+	return (
+		<div className="my-4 max-w-lg overflow-hidden rounded-2xl border bg-muted/30 transition-[box-shadow] duration-300">
+			<div className="flex items-start gap-3 px-5 pt-5 pb-4 select-none">
+				<Workflow className="h-5 w-5 text-muted-foreground mt-0.5 shrink-0" aria-hidden />
+				<div className="min-w-0">
+					<p className="text-sm font-semibold text-foreground">
+						{phase === "rejected"
+							? "Automation cancelled"
+							: phase === "processing"
+								? "Saving automation"
+								: phase === "complete"
+									? "Automation saved"
+									: "Create automation"}
+					</p>
+					{phase === "processing" ? (
+						<TextShimmerLoader text="Saving automation" size="sm" />
+					) : phase === "complete" ? (
+						<p className="text-xs text-muted-foreground mt-0.5">
+							Automation created from this draft
+						</p>
+					) : phase === "rejected" ? (
+						<p className="text-xs text-muted-foreground mt-0.5">
+							No automation was saved — ask in chat to refine and try again.
+						</p>
+					) : (
+						<p className="text-xs text-muted-foreground mt-0.5">
+							Review and approve to save. To change anything, reply in chat — I'll redraft.
+						</p>
+					)}
+				</div>
+			</div>
+
+			<div className="mx-5 h-px bg-border/50" />
+			<div className="px-5 py-4">
+				<AutomationDraftPreview draft={draft} raw={args} />
+			</div>
+
+			{phase === "pending" && (
+				<>
+					<div className="mx-5 h-px bg-border/50" />
+					<div className="px-5 py-4 flex items-center gap-2 select-none">
+						{canApprove && (
+							<Button size="sm" className="rounded-lg gap-1.5" onClick={handleApprove}>
+								Approve
+								<CornerDownLeftIcon className="size-3 opacity-60" />
+							</Button>
+						)}
+						{canReject && (
+							<Button
+								size="sm"
+								variant="ghost"
+								className="rounded-lg text-muted-foreground"
+								onClick={handleReject}
+							>
+								Reject
+							</Button>
+						)}
+					</div>
+				</>
+			)}
+		</div>
+	);
+}
+
+// ----------------------------------------------------------------------------
+// Terminal result cards.
+// ----------------------------------------------------------------------------
+
+function SavedCard({ result }: { result: SavedResult }) {
+	const searchSpaceId = useAtomValue(activeSearchSpaceIdAtom);
+	const detailHref = searchSpaceId
+		? `/dashboard/${searchSpaceId}/automations/${result.automation_id}`
+		: null;
+
+	return (
+		<div className="my-4 max-w-lg overflow-hidden rounded-2xl border bg-muted/30 select-none">
+			<div className="flex items-start gap-3 px-5 pt-5 pb-4">
+				<Workflow className="h-5 w-5 text-muted-foreground mt-0.5 shrink-0" aria-hidden />
+				<div className="min-w-0">
+					<p className="text-sm font-semibold text-foreground">Automation saved</p>
+					<p className="text-xs text-muted-foreground mt-0.5">{result.name}</p>
+				</div>
+			</div>
+			{detailHref && (
+				<>
+					<div className="mx-5 h-px bg-border/50" />
+					<div className="px-5 py-3">
+						<Link
+							href={detailHref}
+							className="inline-flex items-center gap-1.5 text-xs text-primary hover:underline"
+						>
+							<ExternalLink className="h-3.5 w-3.5" aria-hidden />
+							Open automation #{result.automation_id}
+						</Link>
+					</div>
+				</>
+			)}
+		</div>
+	);
+}
+
+function InvalidCard({ result }: { result: InvalidResult }) {
+	return (
+		<div className="my-4 max-w-lg overflow-hidden rounded-2xl border bg-muted/30 select-none">
+			<div className="px-5 pt-5 pb-4">
+				<p className="text-sm font-semibold text-destructive">Couldn't draft this automation</p>
+				<p className="text-xs text-muted-foreground mt-0.5">
+					The drafter produced output that didn't validate. I'll refine and retry.
+				</p>
+			</div>
+			{result.issues.length > 0 && (
+				<>
+					<div className="mx-5 h-px bg-border/50" />
+					<ul className="px-5 py-3 space-y-1 text-xs text-muted-foreground list-disc list-inside">
+						{result.issues.map((issue) => (
+							<li key={issue}>{issue}</li>
+						))}
+					</ul>
+				</>
+			)}
+		</div>
+	);
+}
+
+function ErrorCard({ result }: { result: ErrorResult }) {
+	return (
+		<div className="my-4 max-w-lg overflow-hidden rounded-2xl border bg-muted/30 select-none">
+			<div className="px-5 pt-5 pb-4">
+				<p className="text-sm font-semibold text-destructive">Failed to create automation</p>
+			</div>
+			<div className="mx-5 h-px bg-border/50" />
+			<div className="px-5 py-4">
+				<p className="text-sm text-muted-foreground">{result.message}</p>
+			</div>
+		</div>
+	);
+}
+
+// ----------------------------------------------------------------------------
+// Entry — dispatches between the approval card and terminal result cards.
+//
+// Rejection is special: we hide the standalone "rejected" card because the
+// approval card itself already transitions to a "rejected" phase inline. A
+// second message in the timeline would be noisy.
+// ----------------------------------------------------------------------------
+
+export const CreateAutomationToolUI = ({
+	args,
+	result,
+}: ToolCallMessagePartProps<{ intent: string }, CreateAutomationResult>) => {
+	const { dispatch } = useHitlDecision();
+
+	if (!result) return null;
+
+	if (isInterruptResult(result)) {
+		return (
+			<ApprovalCard
+				args={args as unknown as Record<string, unknown>}
+				interruptData={result as InterruptResult<AutomationCreateContext>}
+				onDecision={(decision) => dispatch([decision])}
+			/>
+		);
+	}
+
+	if (hasStatus(result, "rejected")) return null;
+	if (hasStatus(result, "saved")) return <SavedCard result={result as SavedResult} />;
+	if (hasStatus(result, "invalid")) return <InvalidCard result={result as InvalidResult} />;
+	if (hasStatus(result, "error")) return <ErrorCard result={result as ErrorResult} />;
+
+	return null;
+};
+
+// ----------------------------------------------------------------------------
+// Helpers.
+// ----------------------------------------------------------------------------
+
+/**
+ * Project raw args into the shape ``AutomationDraftPreview`` expects.
+ *
+ * The args dict is the full ``AutomationCreate`` payload (minus
+ * ``search_space_id`` which is injected server-side), so we trust the
+ * top-level fields but defend against missing nested defaults.
+ */
+function extractDraft(args: Record<string, unknown>) {
+	const definition = (args.definition ?? {}) as Record<string, unknown>;
+	const planSteps = Array.isArray(definition.plan)
+		? (definition.plan as Array<Record<string, unknown>>).map((step) => ({
+				step_id: String(step.step_id ?? "(unnamed)"),
+				action: String(step.action ?? ""),
+				when: typeof step.when === "string" ? step.when : null,
+			}))
+		: [];
+
+	const triggers = Array.isArray(args.triggers)
+		? (args.triggers as Array<Record<string, unknown>>).map((trigger) => ({
+				type: String(trigger.type ?? "schedule"),
+				params: (trigger.params ?? {}) as Record<string, unknown>,
+				static_inputs: (trigger.static_inputs ?? {}) as Record<string, unknown>,
+				enabled: trigger.enabled !== false,
+			}))
+		: [];
+
+	return {
+		name: String(args.name ?? "(unnamed automation)"),
+		description: typeof args.description === "string" ? args.description : null,
+		definition: {
+			goal: typeof definition.goal === "string" ? definition.goal : null,
+			plan: planSteps,
+		},
+		triggers,
+	};
+}

surfsense_web/components/tool-ui/automation/index.ts

@@ -0,0 +1 @@
+export { CreateAutomationToolUI } from "./create-automation";

surfsense_web/components/tool-ui/index.ts

@@ -7,6 +7,7 @@
  */
 
 export { Audio } from "./audio";
+export { CreateAutomationToolUI } from "./automation";
 export { CreateDropboxFileToolUI, DeleteDropboxFileToolUI } from "./dropbox";
 export {
 	type GenerateImageArgs,

surfsense_web/contracts/enums/toolIcons.tsx

@@ -25,6 +25,7 @@ import {
 	SearchCheck,
 	Send,
 	Trash2,
+	Workflow,
 	Wrench,
 } from "lucide-react";
 
@@ -47,6 +48,8 @@ const TOOL_ICONS: Record<string, LucideIcon> = {
 	scrape_webpage: ScanLine,
 	web_search: Globe,
 	search_surfsense_docs: BookOpen,
+	// Automations
+	create_automation: Workflow,
 	// Memory
 	update_memory: Brain,
 	// Filesystem (built-in deepagent + middleware)
@@ -150,6 +153,8 @@ const TOOL_DISPLAY_NAMES: Record<string, string> = {
 	scrape_webpage: "Read webpage",
 	web_search: "Search the web",
 	search_surfsense_docs: "Search knowledge base",
+	// Automations
+	create_automation: "Create automation",
 	// Memory
 	update_memory: "Update memory",
 	// Calendar

surfsense_web/features/chat-messages/timeline/tool-registry/registry.ts

@@ -17,6 +17,11 @@ const UpdateMemoryToolUI = dynamic(
 	() => import("@/components/tool-ui/user-memory").then((m) => ({ default: m.UpdateMemoryToolUI })),
 	{ ssr: false }
 );
+const CreateAutomationToolUI = dynamic(
+	() =>
+		import("@/components/tool-ui/automation").then((m) => ({ default: m.CreateAutomationToolUI })),
+	{ ssr: false }
+);
 const SandboxExecuteToolUI = dynamic(
 	() =>
 		import("@/components/tool-ui/sandbox-execute").then((m) => ({
@@ -184,6 +189,7 @@ const NullTimelineBody: TimelineToolComponent = () => null;
  */
 const TOOLS_BY_NAME = {
 	task: NullTimelineBody,
+	create_automation: CreateAutomationToolUI,
 	update_memory: UpdateMemoryToolUI,
 	execute: SandboxExecuteToolUI,
 	execute_code: SandboxExecuteToolUI,

surfsense_web/lib/automations/describe-cron.ts

@@ -4,8 +4,9 @@
  * to the raw expression when unrecognized so the user still sees something
  * honest instead of a guess.
  *
- * Lives in the automations slice because it's a UI display concern with no
+ * Lives under ``lib/automations/`` because both the dashboard slice and the
- * consumers outside it. If reuse grows, lift to ``lib/cron-describe.ts``.
+ * chat ``create_automation`` approval card render schedule descriptions —
+ * keeping the helper outside either feature avoids a layering violation.
  */
 
 const DAY_NAMES = ["Sun", "Mon", "Tue", "Wed", "Thu", "Fri", "Sat"];