Repository navigation
BUG: AzureBlobStorageTarget returns a URL with // when the container URL ends with a slash #3495
Description
Describe the bug
AzureBlobStorageTarget._send_prompt_to_target_async builds the returned URL by appending to the configured container URL:
blob_url = self._container_url + "/" + file_nameWhen container_url (or AZURE_STORAGE_ACCOUNT_CONTAINER_URL) ends with /, for example https://acct.blob.core.windows.net/xpia/, the URL gets // before the file name. The upload goes through _parse_url, which ignores the empty segment, so the blob is uploaded as doc.txt. The returned URL .../xpia//doc.txt names a blob /doc.txt, which does not exist, so anything that uses the returned URL (for example the XPIA workflows) points at a missing blob.
With a folder prefix and a trailing slash (.../xpia/folder/), _parse_url returns the prefix folder/, so the blob itself is uploaded as folder//doc.txt instead of folder/doc.txt.
Steps/Code to Reproduce
import asyncio
from unittest.mock import AsyncMock, MagicMock, patch
from pyrit.setup import IN_MEMORY, initialize_pyrit_async
async def main():
await initialize_pyrit_async(memory_db_type=IN_MEMORY)
from pyrit.models import Message, MessagePiece
from pyrit.prompt_target import AzureBlobStorageTarget
for container_url in ["https://acct.blob.core.windows.net/xpia/", "https://acct.blob.core.windows.net/xpia/folder/"]:
target = AzureBlobStorageTarget(container_url=container_url, sas_token="sv=test")
client = AsyncMock()
client.get_blob_client = MagicMock(return_value=AsyncMock(exists=AsyncMock(return_value=False)))
piece = MessagePiece(role="user", original_value="payload", prompt_metadata={"file_name": "doc.txt"})
with patch("azure.storage.blob.aio.ContainerClient.from_container_url", return_value=client):
response = await target.send_prompt_async(message=Message(message_pieces=[piece]))
print("container_url:", container_url)
print(" uploaded blob:", client.get_blob_client.call_args.kwargs["blob"])
print(" returned URL :", response[0].get_value())
asyncio.run(main())Expected Results
The trailing slash makes no difference:
container_url: https://acct.blob.core.windows.net/xpia/
uploaded blob: doc.txt
returned URL : https://acct.blob.core.windows.net/xpia/doc.txt
container_url: https://acct.blob.core.windows.net/xpia/folder/
uploaded blob: folder/doc.txt
returned URL : https://acct.blob.core.windows.net/xpia/folder/doc.txt
Actual Results
The returned URL does not match the uploaded blob, and the prefixed blob gets an empty path segment:
container_url: https://acct.blob.core.windows.net/xpia/
uploaded blob: doc.txt
returned URL : https://acct.blob.core.windows.net/xpia//doc.txt
container_url: https://acct.blob.core.windows.net/xpia/folder/
uploaded blob: folder//doc.txt
returned URL : https://acct.blob.core.windows.net/xpia/folder//doc.txt
Screenshots
N/A
Versions
- OS: Windows 11
- Python version: 3.11.9
- PyRIT version: installed from main (eeb9ba9) in editable mode
Activity