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

Added support for multiple LLM passes to correct the transcript based upon the provided glossary

Browse Source
This commit is contained in:
Eric Rakestraw
2026-04-21 13:12:08 -05:00
parent 4296e3576e
commit 7e1a3f721a
12 changed files with 306 additions and 55 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
tests/test_corrections.py
View File

@@ -16,7 +16,7 @@ def _transcript():
)
def test_apply_corrections_uses_threshold_and_sorts_chronologically():
def test_apply_corrections_uses_threshold_and_preserves_segment_id_order():
transcript = _transcript()
corrections = [
CorrectionCandidate(
@@ -29,8 +29,8 @@ def test_apply_corrections_uses_threshold_and_sorts_chronologically():
result = apply_corrections(transcript, corrections, confidence_threshold=0.8)
assert [segment.speaker for segment in result.transcript] == ["Mike", "Eric"]
assert result.transcript[1].text == "I ask Chauntea for help."
assert [segment.speaker for segment in result.transcript] == ["Eric", "Mike"]
assert result.transcript[0].text == "I ask Chauntea for help."
assert result.skipped == []
@@ -47,7 +47,7 @@ def test_apply_corrections_ignores_below_threshold():
result = apply_corrections(transcript, corrections, confidence_threshold=0.8)
assert result.transcript[1].text == "I ask Chontia for help."
assert result.transcript[0].text == "I ask Chontia for help."
assert result.skipped == []
@@ -68,7 +68,7 @@ def test_apply_corrections_allows_multiple_distinct_spans_in_one_segment():
result = apply_corrections(transcript, [first, second], confidence_threshold=0.8)
assert result.transcript[1].text == "I ask Chauntea for guidance."
assert result.transcript[0].text == "I ask Chauntea for guidance."
assert result.skipped == []
@@ -83,7 +83,7 @@ def test_apply_corrections_skips_missing_substring():
result = apply_corrections(transcript, [correction], confidence_threshold=0.8)
assert result.transcript[1].text == "I ask Chontia for help."
assert result.transcript[0].text == "I ask Chontia for help."
assert len(result.skipped) == 1
assert result.skipped[0].segment_id == 0
assert result.skipped[0].actual_text == "I ask Chontia for help."
@@ -101,7 +101,7 @@ def test_apply_corrections_skips_missing_segment_id():
result = apply_corrections(transcript, [correction], confidence_threshold=0.8)
assert [segment.text for segment in result.transcript] == ["Then Lyra.", "I ask Chontia for help."]
assert [segment.text for segment in result.transcript] == ["I ask Chontia for help.", "Then Lyra."]
assert len(result.skipped) == 1
assert result.skipped[0].segment_id == 99
assert "does not exist" in result.skipped[0].reason
@@ -118,7 +118,7 @@ def test_apply_corrections_skips_no_op():
result = apply_corrections(transcript, [correction], confidence_threshold=0.8)
assert result.transcript[1].text == "I ask Chontia for help."
assert result.transcript[0].text == "I ask Chontia for help."
assert len(result.skipped) == 1
assert "identical" in result.skipped[0].reason
@@ -156,7 +156,7 @@ def test_apply_corrections_skips_empty_original_text():
result = apply_corrections(transcript, [correction], confidence_threshold=0.8)
assert result.transcript[1].text == "I ask Chontia for help."
assert result.transcript[0].text == "I ask Chontia for help."
assert len(result.skipped) == 1
assert "empty" in result.skipped[0].reason