AI metadata and api prompt results data capture now fixed

tests/integration/test_pipeline_flow.py

@@ -76,6 +76,8 @@ class TestPipelineSuccessFlow:
                 provider="openrouter",
                 model="test-model",
                 prompt_name="transcribe_document.md",
+                ai_metadata={"finish_reason": "stop", "usage": {"total_tokens": 42}},
+                raw_api_response={"id": "resp_123", "choices": [{"message": {"content": "Pipeline transcript"}}]},
             )
 
         monkeypatch.setattr(
@@ -94,6 +96,11 @@ class TestPipelineSuccessFlow:
         assert job is not None
         assert job.status == JobStatus.TRANSCRIBED
         assert any(job_source.raw_transcription == "Pipeline transcript" for job_source in job.job_sources)
+        assert any(job_source.ai_metadata == {"finish_reason": "stop", "usage": {"total_tokens": 42}} for job_source in job.job_sources)
+        assert any(
+            job_source.raw_api_response == {"id": "resp_123", "choices": [{"message": {"content": "Pipeline transcript"}}]}
+            for job_source in job.job_sources
+        )
         assert all(job_source.error_detail is None for job_source in job.job_sources)
 
     @pytest.mark.asyncio

tests/providers/test_openrouter.py

@@ -78,7 +78,13 @@ class TestOpenRouterProviderTranscribe:
         """Transcribe returns normalized text from a valid response payload."""
         response = {
             "model": "vendor/model-b",
-            "choices": [{"message": {"content": [{"text": "Line 1"}, {"text": "Line 2"}]}}],
+            "choices": [
+                {
+                    "message": {"content": [{"text": "Line 1"}, {"text": "Line 2"}]},
+                    "finish_reason": "stop",
+                }
+            ],
+            "usage": {"prompt_tokens": 10, "completion_tokens": 25, "total_tokens": 35},
         }
         provider = OpenRouterTranscriptionProvider(
             settings=Settings(openrouter_api_key="test-key"),
@@ -94,6 +100,11 @@ class TestOpenRouterProviderTranscribe:
         assert result.text == "Line 1\nLine 2"
         assert result.provider == "openrouter"
         assert result.model == "vendor/model-b"
+        assert result.ai_metadata == {
+            "finish_reason": "stop",
+            "usage": {"input_tokens": 10, "output_tokens": 25, "total_tokens": 35},
+        }
+        assert result.raw_api_response == response
 
     @pytest.mark.asyncio
     async def test_maps_sdk_exception_to_provider_error(self):

tests/services/test_v2_crud.py

@@ -211,3 +211,45 @@ async def test_source_delete_blocks_when_linked_to_multiple_jobs(default_session
 
     with pytest.raises(SourceDeleteBlockedError):
         await transcriptions.delete_source_from_job_context(job_id=job_one.id, source_id=source.id)
+
+
+@pytest.mark.asyncio
+async def test_update_job_source_transcription_persists_provider_json_payloads(default_session_factory):
+    documents = DocumentService(session_factory=default_session_factory)
+    jobs = JobService(session_factory=default_session_factory)
+    transcriptions = TranscriptionService(session_factory=default_session_factory)
+
+    document = await documents.create_document(Document(id=uuid4(), name="provider-payloads-doc"))
+    job = await jobs.create_job(Job(document_id=document.id))
+    source = await transcriptions.create_source(
+        Source(
+            document_id=document.id,
+            page_number=1,
+            upload_name="provider.jpg",
+            filename="provider.jpg",
+            file_path="uploads/provider.jpg",
+        )
+    )
+    await transcriptions.create_job_source(
+        JobSource(job_id=job.id, source_id=source.id, status=JobSourceStatus.PENDING)
+    )
+
+    metadata = {"finish_reason": "stop", "usage": {"input_tokens": 11, "output_tokens": 22, "total_tokens": 33}}
+    raw_payload = {"id": "resp_xyz", "choices": [{"message": {"content": "provider transcript"}}]}
+
+    await transcriptions.update_job_source_transcription(
+        job_id=job.id,
+        source_id=source.id,
+        text="provider transcript",
+        ai_metadata=metadata,
+        raw_api_response=raw_payload,
+        provider="openrouter",
+        model="test-model",
+        prompt_name="transcribe_document.md",
+    )
+
+    stored_rows = await transcriptions.list_job_sources(job_id=job.id)
+    assert len(stored_rows) == 1
+    assert stored_rows[0].raw_transcription == "provider transcript"
+    assert stored_rows[0].ai_metadata == metadata
+    assert stored_rows[0].raw_api_response == raw_payload