RAG agents gives error: No such file or directory: '/tmp/chromadb/README.md'

[daniel-counto]

I followed the codes in "Basic Usage of RAG Agents" section of the page https://microsoft.github.io/autogen/blog/2023/10/18/RetrieveChat , but it gives the above mentioned error.

[gagb]

@thinkall

[daniel-counto]

FYI, this is my code:

[denonrailz]

Same issue on VS code, while trying to reproduce tutorial. Creating empty file in tmp/chromadb/readme.md seems like skipped this issue

[krissz22]

What i found, while running the script under Windows is, that the original "get_file_from_url" function in "retrieve_utils.py" does not work correctly on Windows systems, as it tries to create a file in the "tmp/chromadb" directory. This directory structure is specific to Unix-based systems and is not valid on Windows. As a result, the "FileNotFoundError" occurs when the function attempts to create the folder and save the file.

Here's the code of the original function:

def get_file_from_url(url: str, save_path: str = None):
    """Download a file from a URL."""
    if save_path is None:
        save_path = os.path.join("/tmp/chromadb", os.path.basename(url))
    with requests.get(url, stream=True) as r:
        r.raise_for_status()
        with open(save_path, "wb") as f:
            for chunk in r.iter_content(chunk_size=8192):
                f.write(chunk)
    return save_path

My solution:
To make the function work on both Unix-based and Windows systems, i modified the function to create the temporary directory in a platform-independent manner. One possible solution is to create the "chromadb" folder in the current working directory.

Here's the updated get_file_from_url function:

def get_file_from_url(url: str, save_path: str = None):
    """Download a file from a URL."""
    if save_path is None:
        current_dir = os.getcwd()
        temp_dir = os.path.join(current_dir, "chromadb")
        save_path = os.path.join(temp_dir, os.path.basename(url))
        os.makedirs(os.path.dirname(save_path), exist_ok=True)

    with requests.get(url, stream=True) as r:
        r.raise_for_status()
        with open(save_path, "wb") as f:
            for chunk in r.iter_content(chunk_size=8192):
                f.write(chunk)
    return save_path

I do not know exactly whether this is the right way to do it, but it worked for me. At least now there is a "chromadb" directory within the working directory which contains the "README.md" file.

[daniel-counto]

I specified " "client": chromadb.PersistentClient(path="/tmp/chromadb")," in the RetrieveUserProxyAgent function and it seems the above error no longer appears:

ragproxyagent = RetrieveUserProxyAgent( name="ragproxyagent", retrieve_config={ "task": "question answering", "docs_path": "https://raw.githubusercontent.com/microsoft/autogen/main/README.md", "client": chromadb.PersistentClient(path="/tmp/chromadb"), "chunk_token_size": 500, "get_or_create": True, }, )

however a new error occurs:

@thinkall

[krissz22]

I reverted the changes made to the "get_file_from_url" function from my post earlier and applied chromadb.PersistentClient(path="/tmp/chromadb"), to my code.

Now i can also confirm that the initial error is gone.

[ewagner70]

why is "db_path" in "retrieve_config" not supported? I suggest to include that as the cleanest, most professional solution, right?