Building AI agents, atomically
test: add tests for audio data and fix edge case
Renaudc0953fe6b937
diff --git a/atomic-agents/atomic_agents/lib/components/agent_memory.py b/atomic-agents/atomic_agents/lib/components/agent_memory.pyindex ad56a89..614ec18 100644--- a/atomic-agents/atomic_agents/lib/components/agent_memory.py+++ b/atomic-agents/atomic_agents/lib/components/agent_memory.py@@ -261,11 +261,14 @@ class AgentMemory:"""Process multimodal objects to convert string paths to Path objects.+ Note: this is necessary only for PDF and Image instructor types. The from_path+ behavior is slightly different for Audio as it keeps the source as a string.+Args:obj: The object to process."""- if isinstance(obj, INSTRUCTOR_MULTIMODAL_TYPES) and isinstance(obj.source, str):+ if isinstance(obj, (Image, PDF)) and isinstance(obj.source, str):# Check if the string looks like a file path (not a URL or base64 data)if not obj.source.startswith(("http://", "https://", "data:")):obj.source = Path(obj.source)diff --git a/atomic-agents/tests/files/audio_sample.mp3 b/atomic-agents/tests/files/audio_sample.mp3new file mode 100644index 0000000..941e42bBinary files /dev/null and b/atomic-agents/tests/files/audio_sample.mp3 differdiff --git a/atomic-agents/tests/lib/components/test_agent_memory.py b/atomic-agents/tests/lib/components/test_agent_memory.pyindex 2dd4008..20b8339 100644--- a/atomic-agents/tests/lib/components/test_agent_memory.py+++ b/atomic-agents/tests/lib/components/test_agent_memory.py@@ -49,6 +49,7 @@ class TestMultimodalSchema(BaseIOSchema):instruction_text: str = Field(..., description="The instruction text")images: List[instructor.Image] = Field(..., description="The images to analyze")pdfs: List[instructor.multimodal.PDF] = Field(..., description="The PDFs to analyze")+ audio: instructor.multimodal.Audio = Field(..., description="The Audio to analyze")@pytest.fixture@@ -149,10 +150,14 @@ def test_dump_and_load(memory):def test_dump_and_load_multimodal_data(memory):test_image = instructor.Image.from_path(path="atomic-agents/tests/files/image_sample.jpg")test_pdf = instructor.multimodal.PDF.from_path(path="atomic-agents/tests/files/pdf_sample.pdf")+ test_audio = instructor.multimodal.Audio.from_path(path="atomic-agents/tests/files/audio_sample.mp3")# multimodal messagememory.add_message(- role="user", content=TestMultimodalSchema(instruction_text="Analyze this image", images=[test_image], pdfs=[test_pdf])+ role="user",+ content=TestMultimodalSchema(+ instruction_text="Analyze this image", images=[test_image], pdfs=[test_pdf], audio=test_audio+ ),)dumped_data = memory.dump()@@ -166,6 +171,7 @@ def test_dump_and_load_multimodal_data(memory):assert new_memory.history[0].content.instruction_text == memory.history[0].content.instruction_textassert new_memory.history[0].content.images == memory.history[0].content.imagesassert new_memory.history[0].content.pdfs == memory.history[0].content.pdfs+ assert new_memory.history[0].content.audio == memory.history[0].content.audiodef test_load_invalid_data(memory):@@ -313,10 +319,12 @@ def test_get_history_with_multimodal_content(memory):# Create a mock imagemock_image = instructor.Image(source="test_url", media_type="image/jpeg", detail="low")mock_pdf = instructor.multimodal.PDF(source="test_pdf_url")+ mock_audio = instructor.multimodal.Audio(source="test_audio_url", media_type="audio/mpeg")# Add a multimodal messagememory.add_message(- "user", TestMultimodalSchema(instruction_text="Analyze this image", images=[mock_image], pdfs=[mock_pdf])+ "user",+ TestMultimodalSchema(instruction_text="Analyze this image", images=[mock_image], pdfs=[mock_pdf], audio=mock_audio),)# Get history and verify format@@ -327,6 +335,7 @@ def test_get_history_with_multimodal_content(memory):assert history[0]["content"][0] == '{"instruction_text": "Analyze this image"}'assert history[0]["content"][1] == mock_imageassert history[0]["content"][2] == mock_pdf+ assert history[0]["content"][3] == mock_audiodef test_get_history_with_multiple_images_multimodal_content(memory):