Implement support for segments with an optional categories key

tests/test_framework_chunking.py

@@ -25,3 +25,19 @@ def test_chunk_transcript_batches_sections_by_token_limit():
     assert len(sections) == 2
     assert [segment.segment.id for segment in sections[0].segments] == [1, 2]
     assert [segment.segment.id for segment in sections[1].segments] == [3]
+
+
+def test_chunk_transcript_prompt_payload_includes_categories_when_present():
+    transcript = parse_transcript_json(
+        """
+        [
+          {"id": 1, "speaker": "A", "start": 0.0, "end": 1.0, "text": "one", "categories": ["intro", "aside"]}
+        ]
+        """
+    )
+
+    sections = chunk_transcript(transcript, max_section_tokens=8, estimator=FakeEstimator())
+
+    assert sections[0].prompt_payload() == [
+        {"id": 1, "original_text": "one", "categories": ["intro", "aside"]}
+    ]

tests/test_llm_validators.py

@@ -120,8 +120,8 @@ def test_spoken_form_plausibility_validator_approves_plausible_and_rejects_impla
     transcript = parse_transcript_json(
         """
         [
-          {"id": 1, "speaker": "A", "start": 0.0, "end": 1.0, "text": "There were gestures at the temple."},
-          {"id": 2, "speaker": "A", "start": 1.0, "end": 2.0, "text": "Lyra moved first."}
+          {"id": 1, "speaker": "A", "start": 0.0, "end": 1.0, "text": "There were gestures at the temple.", "categories": ["narration"]},
+          {"id": 2, "speaker": "A", "start": 1.0, "end": 2.0, "text": "Lyra moved first.", "categories": ["combat"]}
         ]
         """
     )
@@ -176,6 +176,7 @@ def test_spoken_form_plausibility_validator_approves_plausible_and_rejects_impla
     ]
     prompt_text = client.calls[0]["messages"][1]["content"]
     assert '"original_segment_text"' in prompt_text
+    assert '"categories"' in prompt_text
     assert "There were Jesters at the temple." in prompt_text
     assert "Lyra moved first." in prompt_text
 

tests/test_new_pipeline.py

@@ -46,9 +46,9 @@ def _transcript():
     return parse_source_transcript_json(
         """
         [
-          {"speaker": "Eric", "start": 0.0, "end": 1.0, "text": "Hello."},
-          {"speaker": "Eric", "start": 1.5, "end": 2.0, "text": "Again."},
-          {"speaker": "Mike", "start": 10.0, "end": 11.0, "text": "Done."}
+          {"speaker": "Eric", "start": 0.0, "end": 1.0, "text": "Hello.", "categories": ["intro"]},
+          {"speaker": "Eric", "start": 1.5, "end": 2.0, "text": "Again.", "categories": ["intro", "aside"]},
+          {"speaker": "Mike", "start": 10.0, "end": 11.0, "text": "Done.", "categories": ["response"]}
         ]
         """
     )
@@ -74,6 +74,8 @@ def test_process_transcript_runs_noop_framework(tmp_path):
     assert [segment.id for segment in revised] == [1, 2]
     assert revised[0].text == "Hello. Again."
     assert revised[1].text == "Done."
+    assert revised[0].categories == ["intro", "aside"]
+    assert revised[1].categories == ["response"]
     assert [call["stage_name"] for call in llm_client.calls] == [
         "glossary_1:proposal",
         "homophones:proposal",
@@ -186,6 +188,30 @@ def test_external_report_can_be_written(tmp_path):
     assert payload["totals"]["applied_change_count"] == 0
 
 
+def test_process_transcript_preserves_categories_in_llm_prompt_payloads(tmp_path):
+    llm_client = FakeStructuredLLMClient(
+        [
+            {"corrections": []},
+            {"corrections": []},
+            {"corrections": []},
+            {"corrections": []},
+            {"corrections": []},
+        ]
+    )
+
+    process_transcript(
+        _transcript(),
+        _glossary(),
+        AuditaConfig.from_sources(env={}, overrides=None),
+        llm_client=llm_client,
+    )
+
+    proposal_prompt = llm_client.calls[0]["messages"][1]["content"]
+    assert '"categories": [' in proposal_prompt
+    assert '"intro"' in proposal_prompt
+    assert '"aside"' in proposal_prompt
+
+
 def test_default_module_specs_expose_final_validator_order():
     specs = default_module_specs()
 

tests/test_new_transcript_schema.py

@@ -1,4 +1,6 @@
-from audita.core.schemas import parse_source_transcript_json, parse_transcript_json
+import json
+
+from audita.core.schemas import parse_source_transcript_json, parse_transcript_json, transcript_to_json
 
 
 SERIATIM_TRANSCRIPT = """
@@ -60,6 +62,8 @@ def test_parse_source_transcript_json_accepts_seriatim_transcript_object():
     assert segments[0].end == 3.5
     assert segments[0].text == "Hello there."
     assert segments[1].text == "Resolved word run"
+    assert segments[0].categories is None
+    assert segments[1].categories == ["backchannel"]
 
 
 def test_parse_transcript_json_accepts_seriatim_transcript_object_and_ignores_unused_fields():
@@ -67,3 +71,14 @@ def test_parse_transcript_json_accepts_seriatim_transcript_object_and_ignores_un
 
     assert [segment.id for segment in segments] == [1, 2]
     assert [segment.text for segment in segments] == ["Hello there.", "Resolved word run"]
+    assert segments[0].categories is None
+    assert segments[1].categories == ["backchannel"]
+
+
+def test_transcript_to_json_emits_categories_only_when_present():
+    segments = parse_transcript_json(SERIATIM_TRANSCRIPT)
+
+    payload = json.loads(transcript_to_json(segments))
+
+    assert "categories" not in payload[0]
+    assert payload[1]["categories"] == ["backchannel"]