feat: Add Docling as an ingestion option to cognee add.

cognee/tasks/ingestion/save_data_item_to_storage.py

@@ -27,6 +27,12 @@ async def save_data_item_to_storage(data_item: Union[BinaryIO, str, Any]) -> str
 
         return await get_data_from_llama_index(data_item)
 
+    from docling_core.types import DoclingDocument
+
+    if isinstance(data_item, DoclingDocument):
+        # Convert DoclingDocument to plain text and continue processing file
+        data_item = data_item.export_to_text()
+
     # data is a file object coming from upload.
     if hasattr(data_item, "file"):
         return await save_data_to_file(data_item.file, filename=data_item.filename)

cognee/tests/test_add_docling_document.py

@@ -0,0 +1,59 @@
+import asyncio
+import cognee
+
+import os
+
+# Provide your OpenAI LLM API KEY
+os.environ["LLM_API_KEY"] = ""
+
+
+async def main():
+    # Get file path to document to process
+    from pathlib import Path
+
+    current_directory = Path(__file__).resolve().parent
+    file_path_artificial = os.path.join(
+        current_directory, "test_data", "artificial-intelligence.pdf"
+    )
+    file_path_png = os.path.join(current_directory, "test_data", "example_copy.png")
+    file_path_pptx = os.path.join(current_directory, "test_data", "example.pptx")
+
+    await cognee.prune.prune_data()
+    await cognee.prune.prune_system(metadata=True)
+
+    # Import necessary converter, and convert file to DoclingDocument format
+    from docling.document_converter import DocumentConverter
+
+    converter = DocumentConverter()
+
+    result = converter.convert(file_path_artificial)
+    await cognee.add(result.document)
+
+    result = converter.convert(file_path_png)
+    await cognee.add(result.document)
+
+    result = converter.convert(file_path_pptx)
+    await cognee.add(result.document)
+
+    await cognee.cognify()
+
+    answer = await cognee.search("Tell me about Artificial Intelligence.")
+    assert len(answer) != 0
+
+    answer = await cognee.search("Do programmers change light bulbs?")
+    assert len(answer) != 0
+    lowercase_answer = answer[0].lower()
+    assert ("no" in lowercase_answer) or ("none" in lowercase_answer)
+
+    answer = await cognee.search("What colours are there in the presentation table?")
+    assert len(answer) != 0
+    lowercase_answer = answer[0].lower()
+    assert (
+        ("red" in lowercase_answer)
+        and ("blue" in lowercase_answer)
+        and ("green" in lowercase_answer)
+    )
+
+
+if __name__ == "__main__":
+    asyncio.run(main())

pyproject.toml

@@ -137,6 +137,8 @@ debug = ["debugpy>=1.8.9,<2.0.0"]
 
 monitoring = ["sentry-sdk[fastapi]>=2.9.0,<3", "langfuse>=2.32.0,<3"]
 
+docling = ["docling>=2.54", "transformers>=4.55"]
+
 [project.urls]
 Homepage = "https://www.cognee.ai"
 Repository = "https://github.com/topoteretes/cognee"