feat(api): Support Uploading PowerPoint Files for RAG (#733)

* Adds powerpoint (ppt and pptx) support to the files endpoint

src/leapfrogai_api/backend/rag/document_loader.py

@@ -11,6 +11,7 @@
     TextLoader,
     UnstructuredHTMLLoader,
     UnstructuredMarkdownLoader,
+    UnstructuredPowerPointLoader,
     UnstructuredExcelLoader,
 )
 from langchain_core.documents import Document
@@ -24,6 +25,8 @@
     "text/markdown": UnstructuredMarkdownLoader,
     "application/msword": Docx2txtLoader,
     "application/vnd.openxmlformats-officedocument.wordprocessingml.document": Docx2txtLoader,
+    "application/vnd.openxmlformats-officedocument.presentationml.presentation": UnstructuredPowerPointLoader,
+    "application/vnd.ms-powerpoint": UnstructuredPowerPointLoader,
     "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet": UnstructuredExcelLoader,
     "xls:application/vnd.ms-excel": UnstructuredExcelLoader,
 }
@@ -40,6 +43,8 @@
     ".docx": "application/vnd.openxmlformats-officedocument.wordprocessingml.document",
     ".xlsx": "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet",
     ".xls": "xls:application/vnd.ms-excel",
+    ".pptx": "application/vnd.openxmlformats-officedocument.presentationml.presentation",
+    ".ppt": "application/vnd.ms-powerpoint",
 }
 
 

src/leapfrogai_api/pyproject.toml

@@ -18,7 +18,7 @@ dependencies = [
     "supabase-py-async >= 2.5.6",
     "langchain >= 0.2.1",
     "langchain-community >= 0.2.1",
-    "unstructured[md,xlsx] >= 0.14.2", # Only specify necessary filetypes to prevent package bloat (e.g. 130MB vs 6GB)
+    "unstructured[md,xlsx,pptx] >= 0.14.2", # Only specify necessary filetypes to prevent package bloat (e.g. 130MB vs 6GB)
     "pylibmagic >= 0.5.0", # Resolves issue with libmagic not being bundled with OS - https://github.com/ahupp/python-magic/issues/233, may not be needed after this is merged https://github.com/ahupp/python-magic/pull/294
     "python-magic >= 0.4.27",
     "openpyxl >= 3.1.5",

tests/integration/api/test_files.py

@@ -1,7 +1,6 @@
 """Test the API endpoints for files."""
 
 import os
-
 import pytest
 from fastapi import HTTPException, Response, status
 from fastapi.testclient import TestClient
@@ -224,3 +223,82 @@ async def test_excel_file_handling():
     get_deleted_response = client.get(f"/openai/v1/files/{file_object.id}")
     assert get_deleted_response.status_code == status.HTTP_200_OK
     assert get_deleted_response.json() is None, "Deleted file should not be retrievable"
+
+
+@pytest.mark.asyncio
+async def test_powerpoint_file_handling():
+    """Test handling of a PowerPoint file including upload, retrieval, and deletion."""
+    # Path to the test PowerPoint file
+    pptx_file_path = os.path.join(os.path.dirname(__file__), "../../data/test.pptx")
+
+    # Ensure the file exists
+    assert os.path.exists(
+        pptx_file_path
+    ), f"Test PowerPoint file not found at {pptx_file_path}"
+
+    # Test file loading and splitting
+    documents = await load_file(pptx_file_path)
+    assert len(documents) > 0, "No documents were loaded from the PowerPoint file"
+    assert documents[0].page_content, "The first document has no content"
+
+    split_documents = await split(documents)
+    assert len(split_documents) >= len(documents), "Documents were not split properly"
+    assert split_documents[0].page_content, "The first split document has no content"
+
+    # Test file upload via API
+    with open(pptx_file_path, "rb") as pptx_file:
+        response = client.post(
+            "/openai/v1/files",
+            files={
+                "file": (
+                    "test.pptx",
+                    pptx_file,
+                    "application/vnd.openxmlformats-officedocument.presentationml.presentation",
+                )
+            },
+            data={"purpose": "assistants"},
+        )
+
+    assert (
+        response.status_code == status.HTTP_200_OK
+    ), f"Failed to upload PowerPoint file: {response.text}"
+    file_object = FileObject.model_validate(response.json())
+
+    # Test file retrieval
+    get_response = client.get(f"/openai/v1/files/{file_object.id}")
+    assert (
+        get_response.status_code == status.HTTP_200_OK
+    ), f"Failed to retrieve file: {get_response.text}"
+    retrieved_file = FileObject.model_validate(get_response.json())
+    assert (
+        retrieved_file.id == file_object.id
+    ), "Retrieved file ID doesn't match uploaded file ID"
+
+    # Test file content retrieval
+    content_response = client.get(f"/openai/v1/files/{file_object.id}/content")
+    assert (
+        content_response.status_code == status.HTTP_200_OK
+    ), f"Failed to retrieve file content: {content_response.text}"
+    assert (
+        content_response.headers["Content-Type"]
+        == "application/vnd.openxmlformats-officedocument.presentationml.presentation"
+    )
+    assert (
+        content_response.headers["Content-Disposition"]
+        == f'attachment; filename="{file_object.filename}"'
+    )
+    assert len(content_response.content) > 0, "File content is empty"
+
+    # Test file deletion
+    delete_response = client.delete(f"/openai/v1/files/{file_object.id}")
+    assert (
+        delete_response.status_code == status.HTTP_200_OK
+    ), f"Failed to delete file: {delete_response.text}"
+    assert (
+        delete_response.json()["deleted"] is True
+    ), "File was not deleted successfully"
+
+    # Verify file is no longer retrievable
+    get_deleted_response = client.get(f"/openai/v1/files/{file_object.id}")
+    assert get_deleted_response.status_code == status.HTTP_200_OK
+    assert get_deleted_response.json() is None, "Deleted file should not be retrievable"