API output without Persian diacritics or written ezafe ی (company decision); tutor answers too

plain_persian() strips kasra/fatha/damma/shadda/sukun/hamza-above and the written «ه‌ی» ezafe, keeping tanwin. Applied
at output, not in data: PlainPersianJSONRenderer runs one byte-level pass over every JSON response (~5 ms per MB
locally), and to_plain() cleans tutor answers including the stream. Rewriting titles in the DB/seed would have broken
question binding by title path and SeedRecord keys. Tests updated to expect the cleaned text; config/tests_plain_persian.py.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

rag/tests_guardrails.py

@@ -10,6 +10,8 @@ import json
 from types import SimpleNamespace
 from unittest.mock import patch
 
+from content.plain_math import plain_persian
+
 from django.core.cache import cache
 from django.test import SimpleTestCase, TestCase, override_settings
 from rest_framework.test import APIClient
@@ -109,7 +111,8 @@ class RefusalReplacementTests(SimpleTestCase):
         result = {}
         with patch('rag.services.clients.get_openai_client', return_value=_client_answering(text, usage)):
             out = ''.join(stream_answer('۲+۳؟', None, 'هشتم', result=result))
-        self.assertEqual(out, text)
+        # «بی‌تغییر» جز پاک شدنِ اعراب/«ی»ِ اضافه (content.plain_math.plain_persian)، که روی هر پاسخ اعمال می‌شود.
+        self.assertEqual(out, plain_persian(text))
         self.assertFalse(result['refused'])
         self.assertEqual(result['usage'], {'promptTokens': 1200, 'completionTokens': 80})
 
@@ -141,7 +144,7 @@ class GuardrailThroughTheApiTests(TestCase):
                     done = json.loads(events[-1][1])
                 else:
                     text, done = response.json()['answer'], response.json()
-                self.assertEqual(text, guardrails.REFUSAL)
+                self.assertEqual(text, plain_persian(guardrails.REFUSAL))
                 self.assertEqual(done['usedChunks'], [])
         self.assertEqual(Message.objects.filter(conversation__user_id=self.user.id, role='user').count(), 2)
         self.assertEqual(
@@ -153,7 +156,7 @@ class GuardrailThroughTheApiTests(TestCase):
              patch('rag.services.clients.get_openai_client', return_value=_client_answering(MARKER)):
             response = self.client.post('/api/ai/ask', {'userQuestion': 'حالم خوب نیست، با دوستم دعوام شده'},
                                         format='json')
-        self.assertEqual(response.json()['answer'], guardrails.REFUSAL)
+        self.assertEqual(response.json()['answer'], plain_persian(guardrails.REFUSAL))
         self.assertEqual(response.json()['usedChunks'], [])
         self.assertEqual(Message.objects.get(role='assistant').text, guardrails.REFUSAL)