Initial version of application

tests/test_pipeline.py

@@ -0,0 +1,99 @@
+from pathlib import Path
+
+import pytest
+
+from audita.config import AuditaConfig
+from audita.errors import AuditaValidationError
+from audita.pipeline import process_transcript
+from audita.schemas import CorrectionCandidate, CorrectionSet, parse_glossary_yaml, parse_transcript_json
+
+
+class FakeLLMClient:
+    def __init__(self, responses):
+        self.responses = list(responses)
+        self.calls = 0
+
+    def create_corrections(self, messages, config):
+        self.calls += 1
+        return self.responses.pop(0)
+
+
+def _config(tmp_path):
+    return AuditaConfig(
+        api_key="key",
+        max_section_tokens=16000,
+        confidence_threshold=0.8,
+        max_retries=3,
+        work_dir=tmp_path / "work",
+    )
+
+
+def _glossary():
+    return parse_glossary_yaml(
+        """
+        glossary:
+          - name: "Chauntea"
+            category: deity
+            summary: "Chauntea is a deity."
+        """
+    )
+
+
+def _transcript():
+    return parse_transcript_json(
+        """
+        [
+          {"speaker": "Eric", "start": 0.0, "end": 1.0, "text": "I ask Chontia."}
+        ]
+        """
+    )
+
+
+def test_pipeline_processes_with_fake_llm_and_cleans_work_dir(tmp_path):
+    correction = CorrectionCandidate(
+        segment_index=0,
+        speaker="Eric",
+        start=0.0,
+        end=1.0,
+        original_text="I ask Chontia.",
+        corrected_text="I ask Chauntea.",
+        confidence=0.95,
+    )
+    fake_client = FakeLLMClient([CorrectionSet(corrections=[correction])])
+
+    revised = process_transcript(
+        _transcript(),
+        _glossary(),
+        _config(tmp_path),
+        llm_client=fake_client,
+    )
+
+    assert revised[0].text == "I ask Chauntea."
+    assert fake_client.calls == 1
+    assert list((tmp_path / "work").iterdir()) == []
+
+
+def test_pipeline_preserves_work_dir_on_failure(tmp_path):
+    correction = CorrectionCandidate(
+        segment_index=0,
+        speaker="Eric",
+        start=0.0,
+        end=1.0,
+        original_text="Different text.",
+        corrected_text="I ask Chauntea.",
+        confidence=0.95,
+    )
+    fake_client = FakeLLMClient([CorrectionSet(corrections=[correction])])
+
+    with pytest.raises(AuditaValidationError):
+        process_transcript(
+            _transcript(),
+            _glossary(),
+            _config(tmp_path),
+            llm_client=fake_client,
+        )
+
+    preserved = list((tmp_path / "work").iterdir())
+    assert len(preserved) == 1
+    assert (Path(preserved[0]) / "section-0000.json").exists()
+