feat(n8n): clarification roundtrip route + types (#7316)

Closes the half-built clarification loop on the host side. Previously
the plugin populated _meta.requiresClarification, the LLM was told to
emit it, but the host route ignored it and the UI never rendered it.

POST /api/n8n/workflows/generate now inspects the draft for
clarifications before deploying. When non-empty, the route
short-circuits and returns:
  {
    status: "needs_clarification",
    draft,
    clarifications: ClarificationRequest[],
    catalog: TargetGroup[],
  }
where `catalog` is a snapshot of the connector-target-catalog scoped
to the platforms the clarifications reference. The draft itself is
returned verbatim so the client can post it back unchanged.

POST /api/n8n/workflows/resolve-clarification (new) accepts
{ draft, resolutions: { paramPath, value }[] }. It applies the
resolutions to the draft via a small recursive dot-path setter,
prunes the now-resolved structured clarifications from
_meta.requiresClarification, and either deploys (when the list is
empty) or returns the next pending clarification + updated draft so
the UI can chain a second picker (server-then-channel etc.).

The path setter (n8n-clarification.ts) supports dot identifiers,
bracketed string keys (`nodes["Discord Send"].parameters.channelId`),
and numeric indices (`nodes[0].parameters.guildId`). It rejects
malformed paths and never overwrites a non-object intermediate. The
setter is intentionally tiny — about 30 lines — to avoid pulling
lodash into app-core for one feature.

Folds in P6: client-types-chat.ts adds N8nClarificationRequest,
N8nClarificationTargetGroup, N8nWorkflowNeedsClarificationResponse,
N8nWorkflowResolveClarificationRequest, the type-guard
isNeedsClarificationResponse, and widens
N8nWorkflowGenerateResponse to the three-arm union. The clarification
types mirror the plugin's ClarificationRequest shape but live in the
host so client code does not import from plugin internals.

AutomationsView gets a one-line narrowing branch on
isNeedsClarificationResponse so the wider union typechecks cleanly;
the actual UI render lands in the next PR in the stack (P5).

Tests cover both the helpers and the route handlers:
- n8n-clarification.test.ts: 28 cases for coerceClarifications,
  parseParamPath, setByDotPath, applyResolutions,
  pruneResolvedClarifications, buildCatalogSnapshot.
- n8n-routes.test.ts: appended needs_clarification short-circuit,
  legacy-string clarification coercion, missing-catalog fallback,
  resolve happy path, empty-resolutions rejection, malformed-path
  rejection (with paramPath echo), chained-pending-clarification
  return path, and missing-draft rejection.

Note: the n8n-routes.test.ts file fails to load locally due to a
pre-existing missing package (@elizaos/plugin-browser-bridge) in the
agent runtime import chain — same failure mode is observable on the
unmodified file. The new route cases run cleanly once that dep is
restored.

Author: 2-A-M

packages/app-core/src/api/client-types-chat.ts

@@ -459,9 +459,74 @@ export interface N8nWorkflowMissingCredentialsResponse {
   warning: "missing credentials";
 }
 
+/**
+ * Structured clarification request emitted by the plugin's workflow generator
+ * when a node parameter cannot be resolved from the runtime context. The host
+ * surfaces these as quick-pick buttons; on click the host calls
+ * `/api/n8n/workflows/resolve-clarification` with the chosen value, which
+ * patches the draft at `paramPath` and deploys — no LLM regeneration.
+ *
+ * Mirrors the plugin's `ClarificationRequest` (see
+ * @elizaos/plugin-n8n-workflow `src/types/index.ts`). Re-declared here to
+ * avoid a host → plugin import cycle.
+ */
+export interface N8nClarificationRequest {
+  kind:
+    | "target_channel"
+    | "target_server"
+    | "recipient"
+    | "value"
+    | "free_text";
+  platform?: string;
+  scope?: { guildId?: string };
+  question: string;
+  paramPath: string;
+}
+
+/** One server / workspace / contact-collection from a connector catalog. */
+export interface N8nClarificationTargetGroup {
+  platform: string;
+  groupId: string;
+  groupName: string;
+  targets: Array<{
+    id: string;
+    name: string;
+    kind: "channel" | "recipient" | "chat";
+  }>;
+}
+
+/**
+ * Returned by `POST /api/n8n/workflows/generate` when the LLM emitted one or
+ * more `ClarificationRequest`s and the host needs the user to pick a target
+ * before deploying. The `draft` is the unmodified workflow JSON from the
+ * plugin (with the unresolved parameters left absent); `catalog` is a
+ * snapshot of the relevant connector-target-catalog scoped to the
+ * platforms referenced by the clarifications.
+ */
+export interface N8nWorkflowNeedsClarificationResponse {
+  status: "needs_clarification";
+  draft: Record<string, unknown>;
+  clarifications: N8nClarificationRequest[];
+  catalog: N8nClarificationTargetGroup[];
+}
+
+/** Resolution payload sent to /api/n8n/workflows/resolve-clarification. */
+export interface N8nClarificationResolution {
+  paramPath: string;
+  value: string;
+}
+
+export interface N8nWorkflowResolveClarificationRequest {
+  draft: Record<string, unknown>;
+  resolutions: N8nClarificationResolution[];
+  name?: string;
+  workflowId?: string;
+}
+
 export type N8nWorkflowGenerateResponse =
   | N8nWorkflow
-  | N8nWorkflowMissingCredentialsResponse;
+  | N8nWorkflowMissingCredentialsResponse
+  | N8nWorkflowNeedsClarificationResponse;
 
 export function isMissingCredentialsResponse(
   res: N8nWorkflowGenerateResponse,
@@ -473,6 +538,19 @@ export function isMissingCredentialsResponse(
   );
 }
 
+export function isNeedsClarificationResponse(
+  res: N8nWorkflowGenerateResponse,
+): res is N8nWorkflowNeedsClarificationResponse {
+  const candidate = res as N8nWorkflowNeedsClarificationResponse;
+  return (
+    candidate.status === "needs_clarification" &&
+    Array.isArray(candidate.clarifications) &&
+    Array.isArray(candidate.catalog) &&
+    typeof candidate.draft === "object" &&
+    candidate.draft !== null
+  );
+}
+
 export interface N8nWorkflowWriteNode {
   id?: string;
   name: string;