feat: dedicated HR automation AI generator endpoint

- Add /ai/hr-automation-generator with HR-specific schema (9 triggers, 10 actions)
- Backend handles system prompt, JSON parsing, and module tagging
- Wire HR frontend to use dedicated endpoint instead of generic /ai/chat
- Use configurable workerUrl from SIMPATICO_CONFIG instead of hardcoded URL
- Backend enforces module:'hr' tag on generated rules automatically

Author: simpaticohr

backend/simpatico-ats.js

@@ -1043,6 +1043,7 @@ route("POST", "/ai/employee-insight", handleEmployeeInsight);
 route("POST", "/ai/sentiment", handleSentimentAnalysis);
 route("POST", "/ai/interview-question", handleInterviewQuestion);
 route("POST", "/ai/ats-generator", handleATSGenerator);
+route("POST", "/ai/hr-automation-generator", handleHRAutomationGenerator);
 route("POST", "/ai/generate-assessment", handleGenerateAssessment);
 route("POST", "/ai/assessments", handleSaveAssessment);
 route("POST", "/ai/generate-course", handleGenerateCourse);
@@ -3939,6 +3940,49 @@ Include 2-4 logical actions. Keep it practical for enterprise HR.`;
   }
 }
 
+async function handleHRAutomationGenerator(request, env, ctx) {
+  const { query } = await safeJson(request);
+  if (!query) throw new ValidationError("query is required");
+  if (!env.AI) throw new ServiceUnavailableError("AI");
+
+  const systemPrompt = `You are an enterprise HR automation expert. Given a description, output ONLY valid JSON (no markdown, no backticks) for an HR workflow automation rule.
+Schema:
+{
+  "name": "short rule name",
+  "desc": "one sentence description",
+  "trigger": "one of: employee_joined|employee_offboarded|leave_submitted|leave_approved|probation_ending|work_anniversary|birthday|performance_due|schedule",
+  "cond_dept": "department filter or empty",
+  "cond_level": "junior|mid|senior|exec or empty",
+  "actions": [{"type": "send_email|send_slack|create_task|auto_approve|update_status|assign_manager|trigger_webhook|generate_doc|add_to_calendar|send_reminder", "target": "target email or channel", "msg": "message content"}]
+}
+Include 2-3 logical actions. Keep it practical for enterprise HR operations.
+Examples of good rules:
+- On employee_joined: send welcome email + create onboarding task + assign buddy
+- On leave_submitted: auto-approve if days <= 2 + notify manager
+- On probation_ending: create review task + send reminder to manager
+- On birthday: send celebration email + add to calendar`;
+
+  const result = await runLLM(env, ctx.tenantId, [
+    { role: "system", content: systemPrompt },
+    { role: "user", content: query },
+  ], 1024);
+
+  try {
+    const text = result.response.replace(/```json|```/g, "").trim();
+    const rule = JSON.parse(text);
+    // Enforce module tag
+    rule.module = "hr";
+    return apiResponse({ rule });
+  } catch (e) {
+    throw new AppError(
+      "Failed to parse AI-generated HR rule",
+      HTTP.SERVER_ERROR,
+      "PARSE_ERROR",
+      result.response,
+    );
+  }
+}
+
 async function handleGenerateAssessment(request, env, ctx) {
   requireRole(
     ctx,

hr-automation.html

@@ -1167,25 +1167,18 @@ <h2>✨ AI-Generated Automation</h2>
 Include 1-4 actions that make sense for the described workflow.`;
 
   try {
-    var resp = await fetch('https://simpatico-hr-ats.simpaticohrconsultancy.workers.dev/ai/chat', {
+    var workerUrl = (window.SIMPATICO_CONFIG && window.SIMPATICO_CONFIG.workerUrl)
+      || 'https://simpatico-hr-ats.simpaticohrconsultancy.workers.dev';
+    var resp = await fetch(workerUrl + '/ai/hr-automation-generator', {
       method: 'POST',
       headers: { 
         'Content-Type': 'application/json',
         'X-Tenant-ID': typeof getCompanyId === 'function' ? getCompanyId() : 'SIMP_PRO_MAIN'
       },
-      body: JSON.stringify({
-        messages: [
-          { role: 'system', content: systemPrompt },
-          { role: 'user', content: prompt }
-        ]
-      })
+      body: JSON.stringify({ query: prompt })
     });
     var data = await resp.json();
-    var text = data.data ? data.data.response : data.response;
-
-    // Parse JSON
-    var jsonText = text.replace(/```json|```/g, '').trim();
-    var rule = JSON.parse(jsonText);
+    var rule = data.data ? data.data.rule : data.rule;
 
     pendingAiRule = rule;
     rule.is_ai = true;