[REPORT] Error processing document: Error in step 'extract_file_data': unhandled errors in a TaskGroup (1 sub-exception)

[jingyi-zhao-01]

Checklist

Before reporting bugs or help requests, please fill out the following checklist:

• I set the .env file with the necessary keys, including OPENAI_API_KEY, LLAMACLOUD_API_KEY, ELEVENLABS_API_KEY
• My API keys are all functioning and have at least some credit balance
• I ran both the scripts in the tools directory
• I checked that my LlamaCloud Extract Agent and my LlamaCloud Index has been correctly created and are functioning (using the playground functionality in the UI
• I activated the virtual environment and ensured that all the dependencies are installed properly
• As a first step, I launched the docker services through docker compose up -d
• As a second step, I launched the MCP server through uv run src/notebookllama/server.py
• As a third and last step, I launched the Streamlit app through strealit run src/notebookllama/Home.py

Issue Description

My issue is...

Relevant Traceback

i noticed this problem is only present when parsing large PDF, for small PDF this is fine

Other details

OS:
OS: Arch Linux x86_64
Kernel: 6.16.1-arch1-1
Uptime: 3 days, 22 hours, 18 mins
Packages: 1368 (pacman), 10 (flatpak)
Shell: zsh 5.9
Resolution: 3840x2160
DE: Hyprland
WM: sway
Theme: Raleigh [GTK2/3]
Icons: breeze-dark [GTK2/3]
Terminal: kitty
CPU: AMD Ryzen 7 3700X (16) @ 3.600GHz
GPU: AMD ATI Radeon RX 5600 OEM/5600 XT / 5700/5700 XT
Memory: 19790MiB / 64197MiB

uv version:
0.8.22
streamlit version:
1.46.1
fastmcp version:
FastMCP version: 2.9.2
MCP version: 1.9.4
Python version: 3.13.7
Platform: Linux-6.16.1-arch1-1-x86_64-with-glibc2.42
FastMCP root path: /home/jingyi/PycharmProjects/notebookllama/.venv/lib/p…
llama-index-workflows version:
1.0.1

[jingyi-zhao-01]

i noticed there is an issue opened before but for windows, is the fix specific for windows environment ?

[AstraBert]

No, this is mostly a timeout issue with the communication between LlamaCloud and MCP: since we wait for the extraction job to be completed, but some job take longer than other (big PDFs, as you were saying), it tends to happen that it times out on longer jobs (I think longer than 5 or 10 minutes, I don't remember how long we set the timer to be).
There is no real workaround for this except for raising the timeout threshold of the MCP

[jingyi-zhao-01]

got it, thanks for the reply, is there any configs we can modify to adjust the fastMCP timeout

[AstraBert]

Yes, you can change this line:

notebookllama/src/notebookllama/workflow.py

Line 9 in fbae029

MCP_CLIENT = BasicMCPClient(command_or_url="http://localhost:8000/mcp", timeout=120)

to put the timeout to a higher limit