[Bug] GraphRAG Backend application is crashing

[artem-astafev]

Priority

P3-Medium

OS type

Ubuntu

Hardware type

GPU-AMD

Installation method

• Pull docker images from hub.docker.com
• Build docker images from source

Deploy method

• Docker compose
• Docker
• Kubernetes
• Helm

Running nodes

Single Node

What's the version?

e18369b, PR https://github.com/opea-project/GenAIExamples/pull/1250/files

Description

GraphRAG backend application(MegaService) is crushing withImportError: cannot import name handle_message' from 'comps.cores.mega.utils (/home/user/GenAIComps/comps/cores/mega/utils.py) error

Reproduce steps

1. fetch GenAIExample git repo - git clone https://github.com/opea-project/GenAIExamples.git
2. go to docker_image_build - cd ./GenAIExamples/GraphRAG/docker_image_build/
3. fetch GenAIComps repo - git clone https://github.com/opea-project/GenAIComps.git
4. build images with docker compose - docker compose -f build.yaml build
5. run GraphRAG with docker compose

Raw log

graphrag-backend-server  | Traceback (most recent call last):
graphrag-backend-server  |   File "/home/user/graphrag.py", line 10, in <module>
graphrag-backend-server  |     from comps.cores.mega.utils import handle_message
graphrag-backend-server  | ImportError: cannot import name 'handle_message' from 'comps.cores.mega.utils' (/home/user/GenAIComps/comps/cores/mega/utils.py)
graphrag-backend-server  | Traceback (most recent call last):
graphrag-backend-server  |   File "/home/user/graphrag.py", line 10, in <module>
graphrag-backend-server  |     from comps.cores.mega.utils import handle_message
graphrag-backend-server  | ImportError: cannot import name 'handle_message' from 'comps.cores.mega.utils' (/home/user/GenAIComps/comps/cores/mega/utils.py)
graphrag-backend-server  | Traceback (most recent call last):
graphrag-backend-server  |   File "/home/user/graphrag.py", line 10, in <module>
graphrag-backend-server  |     from comps.cores.mega.utils import handle_message
graphrag-backend-server  | ImportError: cannot import name 'handle_message' from 'comps.cores.mega.utils' (/home/user/GenAIComps/comps/cores/mega/utils.py)
graphrag-backend-server  | Traceback (most recent call last):
graphrag-backend-server  |   File "/home/user/graphrag.py", line 10, in <module>
graphrag-backend-server  |     from comps.cores.mega.utils import handle_message
graphrag-backend-server  | ImportError: cannot import name 'handle_message' from 'comps.cores.mega.utils' (/home/user/GenAIComps/comps/cores/mega/utils.py)
graphrag-backend-server  | Traceback (most recent call last):
graphrag-backend-server  |   File "/home/user/graphrag.py", line 10, in <module>
graphrag-backend-server  |     from comps.cores.mega.utils import handle_message
graphrag-backend-server  | ImportError: cannot import name 'handle_message' from 'comps.cores.mega.utils' (/home/user/GenAIComps/comps/cores/mega/utils.py)
graphrag-backend-server  | Traceback (most recent call last):
graphrag-backend-server  |   File "/home/user/graphrag.py", line 10, in <module>
graphrag-backend-server  |     from comps.cores.mega.utils import handle_message
graphrag-backend-server  | ImportError: cannot import name 'handle_message' from 'comps.cores.mega.utils' (/home/user/GenAIComps/comps/cores/mega/utils.py)
graphrag-backend-server  | Traceback (most recent call last):
graphrag-backend-server  |   File "/home/user/graphrag.py", line 10, in <module>
graphrag-backend-server  |     from comps.cores.mega.utils import handle_message
graphrag-backend-server  | ImportError: cannot import name 'handle_message' from 'comps.cores.mega.utils' (/home/user/GenAIComps/comps/cores/mega/utils.py)
graphrag-backend-server  | Traceback (most recent call last):
graphrag-backend-server  |   File "/home/user/graphrag.py", line 10, in <module>
graphrag-backend-server  |     from comps.cores.mega.utils import handle_message
graphrag-backend-server  | ImportError: cannot import name 'handle_message' from 'comps.cores.mega.utils' (/home/user/GenAIComps/comps/cores/mega/utils.py)
graphrag-backend-server  | Traceback (most recent call last):
graphrag-backend-server  |   File "/home/user/graphrag.py", line 10, in <module>
graphrag-backend-server  |     from comps.cores.mega.utils import handle_message
graphrag-backend-server  | ImportError: cannot import name 'handle_message' from 'comps.cores.mega.utils' (/home/user/GenAIComps/comps/cores/mega/utils.py)
graphrag-backend-server  | Traceback (most recent call last):
graphrag-backend-server  |   File "/home/user/graphrag.py", line 10, in <module>
graphrag-backend-server  |     from comps.cores.mega.utils import handle_message
graphrag-backend-server  | ImportError: cannot import name 'handle_message' from 'comps.cores.mega.utils' (/home/user/GenAIComps/comps/cores/mega/utils.py)

[ZailiWang]

Thanks for reporting, let me try to reproduce and analyze the issue.

[xiguiw]

@artem-astafev

We cannot reproduce this with latest code.

This version you used was at specific point - it changes the gateway, transfer code between GenAIComps and GenAIExamples (move some code from GenAICom to GenAIExxamples). So you must match both repos.

You version:
e18369b, PR https://github.com/opea-project/GenAIExamples/pull/1250/files

Would you please update you code to latest (both GenAIExamples and GenAIComp) and try again?

[lkk12014402]

@artem-astafev
kindly remind: the tgi (text-generation-inference) image is for intel/hpu/gaudi https://github.com/opea-project/GenAIExamples/blob/main/GraphRAG/docker_compose/intel/hpu/gaudi/compose.yaml#L41C1-L43C37. if use other hardware, should change the tgi docker image

[artem-astafev]

@artem-astafev

We cannot reproduce this with latest code.

This version you used was at specific point - it changes the gateway, transfer code between GenAIComps and GenAIExamples (move some code from GenAICom to GenAIExxamples). So you must match both repos.

You version: e18369b, PR https://github.com/opea-project/GenAIExamples/pull/1250/files

Would you please update you code to latest (both GenAIExamples and GenAIComp) and try again?

Thank you for the reply, I will check and let you know

[artem-astafev]

@artem-astafev kindly remind: the tgi (text-generation-inference) image is for intel/hpu/gaudi https://github.com/opea-project/GenAIExamples/blob/main/GraphRAG/docker_compose/intel/hpu/gaudi/compose.yaml#L41C1-L43C37. if use other hardware, should change the tgi docker image

Thank you for the reminder, we are using pre-build TGI image for ROCm like ghcr.io/huggingface/text-generation-inference:2.3.1-rocm for TGI instances on AMD GPUs.