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

Updated configuration to utilize segment ids provided in the input transcript

Browse Source
This commit is contained in:
Eric Rakestraw
2026-04-21 15:01:31 -05:00
parent 7e1a3f721a
commit ca01e46d77
13 changed files with 186 additions and 97 deletions
Download Patch File Download Diff File Expand all files Collapse all files

28
tests/test_pipeline.py
View File

@@ -43,8 +43,8 @@ def _transcript():
return parse_transcript_json(
"""
[
{"speaker": "Eric", "start": 10.0, "end": 11.0, "text": "I ask Chontia."},
{"speaker": "Mike", "start": 0.0, "end": 1.0, "text": "Then Lyra."}
{"id": 1, "speaker": "Eric", "start": 10.0, "end": 11.0, "text": "I ask Chontia."},
{"id": 2, "speaker": "Mike", "start": 0.0, "end": 1.0, "text": "Then Lyra."}
]
"""
)
@@ -52,7 +52,7 @@ def _transcript():
def test_pipeline_processes_with_fake_llm_and_cleans_work_dir(tmp_path):
correction = CorrectionCandidate(
segment_id=0,
id=1,
original_text="Chontia",
corrected_text="Chauntea",
confidence=0.95,
@@ -74,7 +74,7 @@ def test_pipeline_processes_with_fake_llm_and_cleans_work_dir(tmp_path):
def test_pipeline_skips_bad_correction_and_preserves_diagnostics(tmp_path):
correction = CorrectionCandidate(
segment_id=0,
id=1,
original_text="Different text.",
corrected_text="Chauntea",
confidence=0.95,
@@ -91,25 +91,25 @@ def test_pipeline_skips_bad_correction_and_preserves_diagnostics(tmp_path):
)
assert revised[1].text == "I ask Chontia."
assert any("Skipping correction for segment 0" in message for message in progress)
assert any("Skipping correction for id 1" in message for message in progress)
preserved = list((tmp_path / "work").iterdir())
assert len(preserved) == 1
skipped_path = preserved[0] / "skipped-corrections.json"
assert skipped_path.exists()
diagnostics = json.loads(skipped_path.read_text(encoding="utf-8"))
assert diagnostics["skipped_corrections"][0]["segment_id"] == 0
assert diagnostics["skipped_corrections"][0]["id"] == 1
assert "does not match any substring" in diagnostics["skipped_corrections"][0]["reason"]
def test_pipeline_retries_skipped_segment_and_cleans_work_dir_when_fixed(tmp_path):
first_pass = CorrectionCandidate(
segment_id=0,
id=1,
original_text="Contia",
corrected_text="Chauntea",
confidence=0.95,
)
second_pass = CorrectionCandidate(
segment_id=0,
id=1,
original_text="Chontia",
corrected_text="Chauntea",
confidence=0.95,
@@ -134,21 +134,21 @@ def test_pipeline_retries_skipped_segment_and_cleans_work_dir_when_fixed(tmp_pat
assert list((tmp_path / "work").iterdir()) == []
def test_pipeline_retry_prompt_contains_only_valid_deduped_segment_ids(tmp_path):
def test_pipeline_retry_prompt_contains_only_valid_deduped_ids(tmp_path):
first_bad = CorrectionCandidate(
segment_id=0,
id=1,
original_text="Contia",
corrected_text="Chauntea",
confidence=0.95,
)
second_bad_same_segment = CorrectionCandidate(
segment_id=0,
id=1,
original_text="Still wrong",
corrected_text="Chauntea",
confidence=0.95,
)
invalid_segment = CorrectionCandidate(
segment_id=99,
id=99,
original_text="Missing",
corrected_text="Chauntea",
confidence=0.95,
@@ -170,13 +170,13 @@ def test_pipeline_retry_prompt_contains_only_valid_deduped_segment_ids(tmp_path)
assert fake_client.calls == 2
retry_prompt = fake_client.messages[1][1]["content"]
retry_payload = json.loads(retry_prompt.split("Transcript section:\n", maxsplit=1)[1])
assert retry_payload == [{"segment_id": 0, "original_text": "I ask Chontia."}]
assert retry_payload == [{"id": 1, "original_text": "I ask Chontia."}]
assert "Retry guidance" in retry_prompt
def test_pipeline_writes_pass_metadata_for_unresolved_retries(tmp_path):
first_pass = CorrectionCandidate(
segment_id=0,
id=1,
original_text="Contia",
corrected_text="Chauntea",
confidence=0.95,