Files Uploaded via API Stored in Temp Folder Instead of Specific Path

[llmwesee]

Description

When I upload documents using the UI upload button, the files are correctly stored in the specified path associated with the collection (e.g., test_path/Introduction - Ladakh Expedition.pdf). However, when uploading the same files using the /upload_api and /add_file_api endpoints, the files are being stored in a temporary folder (e.g., /tmp/gradio/<random_hash>/).
This discrepancy in behavior between the UI and API uploads causes inconsistency in file storage, making it difficult to manage uploaded files.

Steps to Reproduce
Upload a file through the UI upload button.
Result: File is stored in the correct path (e.g., test_path/).
Upload the same file using the /upload_api and /add_file_api endpoints.
Result: File is stored in a temp folder (e.g., /tmp/gradio/<random_hash>/).

Expected Behavior
Files uploaded via the API should be stored in the same path as those uploaded through the UI (e.g., test_path/).
Actual Behavior
Files uploaded via the API are stored in a temporary folder instead of the intended path.

@app.post("/upload_add_document")
async def upload_add_document(
    langchain_mode: str = Form(...),
    file: UploadFile = File(...)
):
    try:
        allowed_extensions = {".pdf", ".txt", ".docx"}
        _, ext = os.path.splitext(file.filename)
        if ext.lower() not in allowed_extensions:
            raise HTTPException(status_code=400, detail="Unsupported file type. Allowed types: .pdf, .txt, .docx")

        # Save the uploaded file locally
        local_file_path = f"{file.filename}"
        with open(local_file_path, "wb") as f:
            f.write(file.file.read())

        # Step 1: Upload the document
        try:
            with tqdm(total=100, desc=f"Uploading {file.filename}", unit='%', ncols=80) as pbar:
                x, server_file_path = client.predict(local_file_path, api_name='/upload_api')
                pbar.update(100)
        except Exception as e:
            raise HTTPException(status_code=500, detail=f"Failed to upload document: {str(e)}")

        # Step 2: Add the document with OCR processing
        try:
            loaders = tuple([None, None, None, None, None, None])  # Adjust loaders as needed
            with tqdm(total=100, desc=f"Adding {file.filename} to {langchain_mode}", unit='%', ncols=80) as pbar:
                response = client.predict(
                    server_file_path, langchain_mode, True, 512, True, *loaders, api_name='/add_file_api'
                )
                pbar.update(100)
        except Exception as e:
            raise HTTPException(status_code=500, detail=f"Failed to add document: {str(e)}")

        # Cleanup: Delete the local file after processing
        if os.path.exists(local_file_path):
            os.remove(local_file_path)

        return {
            "status": "success",
            "langchain_mode": langchain_mode,
            "filename": file.filename,
            "server_file_path": server_file_path,
        }

    except Exception as e:
        if os.path.exists(local_file_path):
            os.remove(local_file_path)
        raise HTTPException(status_code=500, detail=f"Error: {str(e)}")

Observed File Paths

UI Uploads: test_path/Introduction - Ladakh Expedition.pdf
API Uploads: /tmp/gradio/<random_hash>/Rahul_text.txt

Its urgent please help me or suggest a way to resolve this issue on priority basis.
@pseudotensor