Utilising file uploads with Python MCP server

[jackdorney1999]

Hi there,

I want to create a translator tool via a python MCP server that would be able to take the text from the uploaded file, and pass it to my MCP tool. I have set up an MCP server that correctly connects to the LibreChat container (as I have tested this with a simple stock price tool), but I cannot seem to be able to pass the context from the uploaded file to the MCP server. Whenever I upload a file, I get the message:

It seems like there is no file or text provided for translation. Please upload the document or paste the text you would like to have translated into Spanish, and I'll be happy to assist you.

Is there a way to pass the context from uploaded files to the MCP server? I have tested file uploads in a normal chat window (as opposed to using Agents with tools) and it can correctly pass the context of the file to the LLM. For reference, I am using Azure OpenAI GPT-4o for the model, and Azure credentials for the RAG API.

For reference, here is how my MCP server is configured:

@mcp.tool(
    name="Translator",
    description="Translate an uploaded file to a specified language.",
)
def translate_uploaded_file(text: str, target_language: str) -> str:
    """
    Translate an uploaded file to a specified language, preserving its structure.

    Args:
        file: The uploaded file (must be .txt, .md, or .docx).
        target_language: Language to translate the content into (e.g., 'French').

    Returns:
        Translated text as a string.
    """
    try:
        return handle_translation(text, target_language)
    except Exception as e:
        logging.error("Translation failed: %s", e)
        return "An error occurred during translation."
if __name__ == "__main__":
    import uvicorn

    app = mcp.sse_app
    uvicorn.run(app, host="0.0.0.0", port=8000)

With the connection created via librechat.yaml file:

mcpServers:
  Tools:
    type: sse
    url: http://host.docker.internal:8000/sse