eric/audita
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases 2 Wiki Activity

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

Browse Source
This commit is contained in:
Eric Rakestraw
2026-04-25 13:42:18 -05:00
parent 5944f13404
commit bf05d79fc6
9 changed files with 198 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

66
tests/test_module_proposals.py
View File

@@ -500,6 +500,7 @@ def test_process_transcript_result_runs_spoken_word_module_with_full_validator_c
assert [validator["name"] for validator in result.report.modules[0].to_dict()["validators"]] == [
"proposal_confidence_guard",
"protected_glossary_guard",
"non_empty_segment_guard",
"spoken_word_review",
"meaning_reversal_review",
]
@@ -627,6 +628,7 @@ def test_process_transcript_result_runs_grammar_module_with_full_validator_chain
assert [validator["name"] for validator in result.report.modules[0].to_dict()["validators"]] == [
"proposal_confidence_guard",
"protected_glossary_guard",
"non_empty_segment_guard",
"grammar_only_guard",
"meaning_reversal_review",
]
@@ -806,3 +808,67 @@ def test_process_transcript_result_grammar_module_still_rejects_homophone_style_
assert [call["stage_name"] for call in client.calls] == ["grammar:proposal"]
assert result.report.skipped_corrections[0].source == "validator:grammar_only_guard"
assert result.report.skipped_corrections[0].reason == "correction is not limited to punctuation, capitalization, and spacing"
def test_process_transcript_result_rejects_spoken_word_whole_segment_deletion_before_llm_validators(tmp_path):
transcript = parse_source_transcript_json(
"""
[
{"speaker": "Eric", "start": 0.0, "end": 1.0, "text": "uh"}
]
"""
)
base_config = AuditaConfig.from_sources(env={})
config = AuditaConfig(
api_key=base_config.api_key,
model=base_config.model,
base_url=base_config.base_url,
max_retries=base_config.max_retries,
max_section_tokens=base_config.max_section_tokens,
glossary_confidence_threshold=base_config.glossary_confidence_threshold,
grammar_confidence_threshold=base_config.grammar_confidence_threshold,
homophones_confidence_threshold=base_config.homophones_confidence_threshold,
spoken_word_confidence_threshold=base_config.spoken_word_confidence_threshold,
normalize_max_segment_gap=base_config.normalize_max_segment_gap,
normalize_ellipsis_gap=base_config.normalize_ellipsis_gap,
normalize_max_segment_duration=base_config.normalize_max_segment_duration,
normalize_max_segment_tokens=base_config.normalize_max_segment_tokens,
work_dir=tmp_path / "work",
work_dir_retention="always",
)
client = FakeStructuredLLMClient(
[
{
"corrections": [
{
"id": 1,
"original_text": "uh",
"corrected_text": "",
"confidence": 0.95,
}
]
}
]
)
result = process_transcript_result(
transcript,
_glossary(),
config,
module_keys=["spoken_word"],
llm_client=client,
)
assert result.transcript[0].text == "uh"
assert [call["stage_name"] for call in client.calls] == ["spoken_word:proposal"]
assert result.report.skipped_corrections[0].source == "validator:non_empty_segment_guard"
assert result.report.skipped_corrections[0].reason == "correction would leave the segment empty"
assert [validator["name"] for validator in result.report.modules[0].to_dict()["validators"]] == [
"proposal_confidence_guard",
"protected_glossary_guard",
"non_empty_segment_guard",
"spoken_word_review",
"meaning_reversal_review",
]
assert result.report.modules[0].validators[2].rejected_count == 1
assert result.report.modules[0].validators[3].candidate_count == 0