Added a validator to prevent changes that result in an empty or whitespace-only string

tests/test_llm_validators.py

@@ -5,7 +5,7 @@ from audita.core.config import AuditaConfig
 from audita.core.errors import AuditaLLMError
 from audita.core.schemas import parse_glossary_yaml, parse_transcript_json
 from audita.framework.models import CorrectionProposal, ModuleRunSpec
-from audita.validators import GrammarOnlyValidator
+from audita.validators import GrammarOnlyValidator, NonEmptySegmentValidator
 from audita.validators.base import ValidationContext
 from audita.validators.llm import MeaningReversalValidator, SpokenFormPlausibilityValidator, SpokenWordValidator
 from audita.validators.prompts import (
@@ -472,6 +472,84 @@ def test_grammar_only_validator_rejects_word_level_changes(tmp_path):
     )
 
 
+def test_non_empty_segment_validator_rejects_empty_and_whitespace_only_segments(tmp_path):
+    transcript = parse_transcript_json(
+        """
+        [
+          {"id": 1, "speaker": "A", "start": 0.0, "end": 1.0, "text": "uh"},
+          {"id": 2, "speaker": "A", "start": 1.0, "end": 2.0, "text": "um"}
+        ]
+        """
+    )
+    proposals = [
+        CorrectionProposal(
+            proposal_index=0,
+            module_instance="spoken_word",
+            module_key="spoken_word",
+            id=1,
+            original_text="uh",
+            corrected_text="",
+            confidence=0.95,
+        ),
+        CorrectionProposal(
+            proposal_index=1,
+            module_instance="spoken_word",
+            module_key="spoken_word",
+            id=2,
+            original_text="um",
+            corrected_text="   ",
+            confidence=0.95,
+        ),
+    ]
+
+    result = NonEmptySegmentValidator("non_empty_segment_guard").validate(
+        _context(proposals=proposals, transcript=transcript, llm_client=None, tmp_path=tmp_path)
+    )
+
+    assert [decision.approved for decision in result.decisions] == [False, False]
+    assert all(decision.reason == "correction would leave the segment empty" for decision in result.decisions)
+
+
+def test_non_empty_segment_validator_allows_punctuation_only_and_unpreviewable_proposals(tmp_path):
+    transcript = parse_transcript_json(
+        """
+        [
+          {"id": 1, "speaker": "A", "start": 0.0, "end": 1.0, "text": "uh"},
+          {"id": 2, "speaker": "A", "start": 1.0, "end": 2.0, "text": "There were gestures at the temple."}
+        ]
+        """
+    )
+    proposals = [
+        CorrectionProposal(
+            proposal_index=0,
+            module_instance="spoken_word",
+            module_key="spoken_word",
+            id=1,
+            original_text="uh",
+            corrected_text=".",
+            confidence=0.95,
+        ),
+        CorrectionProposal(
+            proposal_index=1,
+            module_instance="spoken_word",
+            module_key="spoken_word",
+            id=2,
+            original_text="rank",
+            corrected_text="Hrank",
+            confidence=0.95,
+        ),
+    ]
+
+    result = NonEmptySegmentValidator("non_empty_segment_guard").validate(
+        _context(proposals=proposals, transcript=transcript, llm_client=None, tmp_path=tmp_path)
+    )
+
+    assert [(decision.proposal_index, decision.approved, decision.reason) for decision in result.decisions] == [
+        (0, True, None),
+        (1, True, None),
+    ]
+
+
 @pytest.mark.parametrize(
     ("validator", "payload", "message_fragment"),
     [