import error when importing `from langchain import OpenAI` on 0.0.151

[ColemanDunn]

got the following error when running today:

    from langchain.agents import MRKLChain, ReActChain, SelfAskWithSearchChain
  File "venv/lib/python3.11/site-packages/langchain/agents/__init__.py", line 2, in <module>
    from langchain.agents.agent import (
  File "venv/lib/python3.11/site-packages/langchain/agents/agent.py", line 17, in <module>
    from langchain.chains.base import Chain
  File "venv/lib/python3.11/site-packages/langchain/chains/__init__.py", line 2, in <module>
    from langchain.chains.api.base import APIChain
  File "venv/lib/python3.11/site-packages/langchain/chains/api/base.py", line 8, in <module>
    from langchain.chains.api.prompt import API_RESPONSE_PROMPT, API_URL_PROMPT
  File "venv/lib/python3.11/site-packages/langchain/chains/api/prompt.py", line 2, in <module>
    from langchain.prompts.prompt import PromptTemplate
  File "venv/lib/python3.11/site-packages/langchain/prompts/__init__.py", line 14, in <module>
    from langchain.prompts.loading import load_prompt
  File "venv/lib/python3.11/site-packages/langchain/prompts/loading.py", line 14, in <module>
    from langchain.utilities.loading import try_load_from_hub
  File "venv/lib/python3.11/site-packages/langchain/utilities/__init__.py", line 5, in <module>
    from langchain.utilities.bash import BashProcess
  File "venv/lib/python3.11/site-packages/langchain/utilities/bash.py", line 7, in <module>
    import pexpect
ModuleNotFoundError: No module named 'pexpect'

does this need to be added to project dependencies?

[Matias-HA]

I am also dealing with the same issue

[yiweig]

Seems like pexpect was recently added in https://github.com/hwchase17/langchain/pull/3580/files#diff-5f99ae6d8500359517f487845b0a2b2f6cad1e3d7751c501bdc9b99bd18aadc3R7 but I don't seem to see the new package added anywhere, unless I'm missing something

[SandeepPasula898]

I am also facing issues with pexpect

ERROR: AttributeError: module 'pexpect' has no attribute 'spawn'

this spawn won't work in WINDOWS OS. Ref: https://github.com/pexpect/pexpect/issues/321

[Valentin-Semur]

I'm facing the same issue as well

[Nacho48]

Same here

[Cobryis]

@vowelparrot for vis

[vowelparrot]

Will fix. Thank you. We will improve our CI to prevent this type of error from occurring moving forward

[vowelparrot]

The reason this slipped through is that it is a dependency of poetry itself. we will address this. Thanks for everyone's patience

[kjain25]

Was there any solution to this or will I have to downgrade langchain version?

[kennethleungty]

Downgrading is probably the fastest way currently, while this issue is getting fixed

[emarco177]

As a WA you can simply install pipenv update langchain pipenv install pyexpect

this will patch the issue until it is fixed

[VladoPortos]

As a WA you can simply install pipenv isntall pyexpect

this will patch the issue until it is fixed

This does not help,. it will fail with:

    def _initialize_persistent_process(prompt: str) -> pexpect.spawn:
AttributeError: module 'pexpect' has no attribute 'spawn'

[martinwunderlich]

Ran into the same issue this morning. Manually installing pyexpect did not resolve it. Still getting "ModuleNotFoundError: No module named 'pexpect'". I had to downgrade langchain to version 0.0.150 by adding langchain = "==0.0.150" to the pipfile.

[VladoPortos]

Ran into the same issue this morning. Manually installing pyexpect did not resolve it. Still getting "ModuleNotFoundError: No module named 'pexpect'". I had to downgrade langchain to version 0.0.150 by adding langchain = "==0.0.150" to the pipfile.

Yea same, jumped a version back and that one works :)

[Nacho48]

what file do you open to change langchain = "==0.0.150"? sorry im a little new on this ^^

[VladoPortos]

what file do you open to change langchain = "==0.0.150"? sorry im a little new on this ^^

Ah, not a file, when you installed langchain where was command pip install langchain, you can specify which version to install with pip install -U langchain==0.0.150

[Nacho48]

Thanks! Doing progress now!

[martinwunderlich]

@Nacho48 If you are using pipenv, the line should be in your Pipfile. Otherwise, install directly like Vlado suggested

[zulven]

For windows you could just add your own class for PowerShell like this (until windows is supported):

from typing import List, Union
from uuid import uuid4
import re
import pexpect
from pexpect.popen_spawn import PopenSpawn

class PowerShellProcess:
    """Executes PowerShell commands and returns the output."""

    def __init__(
        self,
        strip_newlines: bool = False,
        return_err_output: bool = False,
        persistent: bool = False,
    ):
        self.strip_newlines = strip_newlines
        self.return_err_output = return_err_output
        self.prompt = ""
        self.process = None
        if persistent:
            self.prompt = str(uuid4())
            self.process = self._initialize_persistent_process(self.prompt)

    @staticmethod
    def _initialize_persistent_process(prompt: str) -> PopenSpawn:
        process = PopenSpawn(
            "pwsh -NoProfile", encoding="utf-8"
        )
        # Set the custom prompt
        process.sendline(f"function prompt {{'{prompt}'}}")

        process.expect_exact(prompt, timeout=10)
        return process

    def run(self, commands: Union[str, List[str]]) -> str:
        """Run commands and return final output."""
        if isinstance(commands, str):
            commands = [commands]
        commands = ";".join(commands)
        if self.process is not None:
            return self._run_persistent(
                commands,
            )
        else:
            return self._run(commands)

    def _run(self, command: str) -> str:
        """Run commands and return final output."""
        try:
            output = subprocess.run(
                f"pwsh -Command {command}",
                shell=True,
                check=True,
                stdout=subprocess.PIPE,
                stderr=subprocess.STDOUT,
            ).stdout.decode()
        except subprocess.CalledProcessError as error:
            if self.return_err_output:
                return error.stdout.decode()
            return str(error)
        if self.strip_newlines:
            output = output.strip()
        return output

    def process_output(self, output: str, command: str) -> str:
        # Remove the command from the output using a regular expression
        pattern = re.escape(command) + r"\s*\n"
        output = re.sub(pattern, "", output, count=1)
        return output.strip()

    def _run_persistent(self, command: str) -> str:
        """Run commands and return final output."""
        if self.process is None:
            raise ValueError("Process not initialized")
        self.process.sendline(command)

        # Clear the output with an empty string
        self.process.expect(self.prompt, timeout=10)
        self.process.sendline("")

        try:
            self.process.expect([self.prompt, pexpect.EOF], timeout=10)
        except pexpect.TIMEOUT:
            return f"Timeout error while executing command {command}"
        if self.process.after == pexpect.EOF:
            return f"Exited with error status: {self.process.exitstatus}"
        output = self.process.before
        output = self.process_output(output, command)
        if self.strip_newlines:
            return output.strip()
        return output

Should work for most use cases

[vowelparrot]

Fixed in 00.152 - The persistent Bash shell still doesn't support windows, but it shouldn't be a blocker for other usage

[zulven]

"persistence"-feature should work with that custom pwsh class in windows in the meantime