Can't import dspy in my pipeline file

[Rensing1]

I want to use the DSPy framework in my pipeline. But before I could implement anything, I already get an error when I try to import the DSPy module.

When I upload my pipeline.py into open-webui, I receive the message that the upload was successful; however, no new pipeline is available.

Everything works fine when I don't use import dspy at the top of my pipeline.py. This rest of the file content is basically copied from the pipeline examples folder.

Here's the code:

import os
from typing import List, Optional
from pydantic import BaseModel
from schemas import OpenAIChatMessage
import time
import dspy

class Pipeline:
class Valves(BaseModel):
pipelines: List[str] = []
priority: int = 0

    # Valves for conversation turn limiting
    target_user_roles: List[str] = ["user"]
    max_turns: Optional[int] = None

def __init__(self):
    self.type = "filter"
    self.name = "DSPy-Pipeline"
    self.valves = self.Valves(
        **{
            "pipelines": os.getenv("CONVERSATION_TURN_PIPELINES", "*").split(","),
            "max_turns": 1,
        }
    )

async def on_startup(self):
    # This function is called when the server is started.
    print(f"on_startup:{__name__}")
    pass

async def on_shutdown(self):
    # This function is called when the server is stopped.
    print(f"on_shutdown:{__name__}")
    pass

async def inlet(self, body: dict, user: Optional[dict] = None) -> dict:
    print(f"pipe:{__name__}")
    print(body)
    print(user)

    if user.get("role", "admin") in self.valves.target_user_roles:
        messages = body.get("messages", [])
        if len(messages) > self.valves.max_turns:
            raise Exception(
                f"Conversation turn limit exceeded. Max turns: {self.valves.max_turns}"
            )

    return body

`

And here is the relevant part of the output from docker logs webui:
upload_pipeline 1 pipeline.py
INFO: 172.19.0.1:45164 - "POST /api/v1/pipelines/upload HTTP/1.1" 200 OK
INFO: 172.19.0.1:45164 - "GET /api/v1/pipelines/?urlIdx=1 HTTP/1.1" 200 OK
INFO [open_webui.routers.openai] get_all_models()
INFO [open_webui.routers.ollama] get_all_models()
INFO: 172.19.0.1:45180 - "GET /api/models HTTP/1.1" 200 OK
INFO: 172.19.0.1:45164 - "GET /api/v1/pipelines/pipeline/valves/spec?urlIdx=1 HTTP/1.1" 200 OK
INFO: 172.19.0.1:45180 - "GET /api/v1/pipelines/pipeline/valves?urlIdx=1 HTTP/1.1" 200 OK
WARNI [python_multipart.multipart] Skipping data after last boundary
upload_pipeline 1 pipeline.py
INFO: 172.19.0.1:58594 - "POST /api/v1/pipelines/upload HTTP/1.1" 200 OK
INFO: 172.19.0.1:58594 - "GET /api/v1/pipelines/?urlIdx=1 HTTP/1.1" 200 OK
INFO [open_webui.routers.openai] get_all_models()
INFO [open_webui.routers.ollama] get_all_models()
INFO: 172.19.0.1:58608 - "GET /api/models HTTP/1.1" 200 OK

The first upload is the file without import dspy, the second upload is with import dspy.

And here is the output from docker logs pipelines:
INFO: 172.19.0.3:37984 - "POST /pipelines/upload HTTP/1.1" 200 OK
INFO: 172.19.0.3:37996 - "GET /pipelines HTTP/1.1" 200 OK
INFO: 172.19.0.3:38010 - "GET /models HTTP/1.1" 200 OK
WARNING:root:No Pipeline class found in pipeline
Error loading module: pipeline
Failed to import dspy: cannot import name 'TypeIs' from 'typing_extensions' (/usr/local/lib/python3.11/site-packages/typing_extensions.py)
INFO: 172.19.0.3:51356 - "POST /pipelines/upload HTTP/1.1" 200 OK
INFO: 172.19.0.3:51368 - "GET /pipelines HTTP/1.1" 200 OK
INFO: 172.19.0.3:51384 - "GET /models HTTP/1.1" 200 OK

The error can't be resolved by pip install dspy or pip install typing_extensions in the respective docker container.

[TheWasilingMan]

Try putting a "requirements" header on the file with dspy listed. For instance, check out this example: https://github.com/open-webui/pipelines/blob/main/examples/pipelines/rag/haystack_pipeline.py#L8

[mykola-mmm]

I am getting the same kind of error when trying to import ChatOpenAI from langchain_openai, the error is:
cannot import name 'TypeIs' from 'typing_extensions' (/usr/local/lib/python3.11/site-packages/typing_extensions.py)
My requirements in the header are following:
requirements: pydantic==2.7.4, typing_extensions==4.12.2, langchain==0.3.15, langchain-community==0.3.15, langchain-openai==0.3.2, langchain-core==0.3.31, langchain-text-splitters==0.3.5

The pipeline image I am using is open-webui/pipelines:git-db29eb2

[mykola-mmm]

@Rensing1 I have managed to solve this issue by adding the pipeline (without requirements) via the docker-compose volumes and passing requirements in .txt file and setting PIPELINES_REQUIREMENTS_PATH to point to the requirements file

[Rensing1]

Thanks for the responses! Adding the requirements header in the pipeline.py helped. Using a requirements.txt file also seems to help.

Another fix is using "pip install ..." in the docker bash, also for some reason this resetted itself sometimes for me.