You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I've been successfully running a lanchain RAG pipeline since August 2024, but it stopped working a couple weeks ago, also there was no code changes. I did not notice immediately as it's not used a lot, and therefore cannot relate it to a specific open-webui, langchain, or pydantic update.
What is weird is that it runs from an interactive python shell in the same environment, but does not work when loaded from open-webui.
When I check the logs of the pipeline pod, I'm getting the following:
olivier@SINGLE-HOST-DOCKER:~/k8s/open-webui$ k logs open-webui-pipelines-769d89bc96-crwrt | tail
WARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv
[notice] A new release of pip is available: 24.0 -> 24.3.1
[notice] To update, run: pip install --upgrade pip
Error loading module: ask_hr_policies_pipeline
cannot import name 'can_be_positional' from 'pydantic._internal._utils' (/usr/local/lib/python3.11/site-packages/pydantic/_internal/_utils.py)
WARNING:root:No Pipeline class found in ask_hr_policies_pipeline
INFO: 10.233.103.164:37564 - "POST /pipelines/upload HTTP/1.1" 200 OK
INFO: 10.233.103.164:46070 - "GET /pipelines HTTP/1.1" 200 OK
INFO: 10.233.103.164:46074 - "GET /models HTTP/1.1" 200 OK
I tried to exec the code directly inside the pipelines pod, hoping it would help me get better debug messages, but when I do that it runs perfectly well, as can be seen in the following session:
olivier@SINGLE-HOST-DOCKER:~/k8s/open-webui$ k exec -it open-webui-pipelines-769d89bc96-crwrt -- bash
root@open-webui-pipelines-769d89bc96-crwrt:/app# cd pipelines/
root@open-webui-pipelines-769d89bc96-crwrt:/app/pipelines# python
Python 3.11.10 (main, Nov 12 2024, 02:25:24) [GCC 12.2.0] on linux
Type "help", "copyright", "credits" or "license"for more information.
>>> from failed.ask_hr_policies_pipeline import Pipeline
>>> p = Pipeline()
>>> import asyncio
>>> asyncio.run(p.on_startup())
on_startup:failed.ask_hr_policies_pipeline
⚠️ It looks like you upgraded from a version below 0.6 and could benefit from vacuuming your database. Run chromadb utils vacuum --help for more information.
>>> result = p.pipe("What is the vacation policy?", "", [], {})
>>> forrin result:
... r
...
pipe:failed.ask_hr_policies_pipeline
'''The'' vacation'' policy'' states'' that'' employees'' are'' entitled'' to''''20'' working'' days'' of'' vacation'' per'' calendar'' year'
--- stripped for sake of space ---
I deployed open-webui and pipelines using the open-webui helm chart from the helm.openwebui.com repo. I'm currently running the latest version.
I've been successfully running a lanchain RAG pipeline since August 2024, but it stopped working a couple weeks ago, also there was no code changes. I did not notice immediately as it's not used a lot, and therefore cannot relate it to a specific open-webui, langchain, or pydantic update.
What is weird is that it runs from an interactive python shell in the same environment, but does not work when loaded from open-webui.
My repo is public and the full code for the pipeline is available here: https://github.com/marvinpac-it/ask-hr-policies-open-webui/blob/master/ask_hr_policies_pipeline.py
When I check the logs of the pipeline pod, I'm getting the following:
I tried to exec the code directly inside the pipelines pod, hoping it would help me get better debug messages, but when I do that it runs perfectly well, as can be seen in the following session:
I deployed open-webui and pipelines using the open-webui helm chart from the helm.openwebui.com repo. I'm currently running the latest version.
The pipeline is defined in the values file of the helm chart:
Any help troubleshooting this further would be really appreciated.
The text was updated successfully, but these errors were encountered: