[Bug]: TypeError: metaclass conflict: the metaclass of a derived class must be a (non-strict) subclass of the metaclasses of all its bases

[hotpeppeper]

Current Behavior

from gptcache.adapter.langchain_models import LangChainChat

Traceback (most recent call last):
File "/home/ld/miniconda3/envs/llm/lib/python3.10/runpy.py", line 196, in _run_module_as_main
return _run_code(code, main_globals, None,
File "/home/ld/miniconda3/envs/llm/lib/python3.10/runpy.py", line 86, in _run_code
exec(code, run_globals)
File "/home/ld/.vscode-server/extensions/ms-python.python-2023.14.0/pythonFiles/lib/python/debugpy/adapter/../../debugpy/launcher/../../debugpy/main.py", line 39, in
cli.main()
File "/home/ld/.vscode-server/extensions/ms-python.python-2023.14.0/pythonFiles/lib/python/debugpy/adapter/../../debugpy/launcher/../../debugpy/../debugpy/server/cli.py", line 430, in main
run()
File "/home/ld/.vscode-server/extensions/ms-python.python-2023.14.0/pythonFiles/lib/python/debugpy/adapter/../../debugpy/launcher/../../debugpy/../debugpy/server/cli.py", line 284, in run_file
runpy.run_path(target, run_name="main")
File "/home/ld/.vscode-server/extensions/ms-python.python-2023.14.0/pythonFiles/lib/python/debugpy/_vendored/pydevd/_pydevd_bundle/pydevd_runpy.py", line 321, in run_path
return _run_module_code(code, init_globals, run_name,
File "/home/ld/.vscode-server/extensions/ms-python.python-2023.14.0/pythonFiles/lib/python/debugpy/_vendored/pydevd/_pydevd_bundle/pydevd_runpy.py", line 135, in _run_module_code
_run_code(code, mod_globals, init_globals,
File "/home/ld/.vscode-server/extensions/ms-python.python-2023.14.0/pythonFiles/lib/python/debugpy/_vendored/pydevd/_pydevd_bundle/pydevd_runpy.py", line 124, in _run_code
exec(code, run_globals)
File "/home/ld/code/cache_t/gptcache_t.py", line 14, in
from cache import MyLangChainChat
File "/home/ld/code/cache_t/cache/init.py", line 1, in
from .langchainchat import MyLangChainChat
File "/home/ld/code/cache_t/cache/langchainchat.py", line 4, in
from gptcache.adapter.langchain_models import LangChainChat
File "/home/ld/miniconda3/envs/llm/lib/python3.10/site-packages/gptcache/adapter/langchain_models.py", line 30, in
class LangChainLLMs(LLM, BaseModel):
TypeError: metaclass conflict: the metaclass of a derived class must be a (non-strict) subclass of the metaclasses of all its bases

Expected Behavior

No response

Steps To Reproduce

No response

Environment

ubuntu: 22.04 LTS
gptcache: 0.1.40

Anything else?

No response

[SimFG]

This error seems to be caused by an incompatibility issue between langchain and gptcache. You can try using an older version of langchain.

[BastienKovac]

Looking into it, it looks like the metaclass from Langchain's LLM and pydantic's BaseModel are mismatched :

So the hierarchy from LangChainLLMs raises an error.

I did some dirty testing and it seems that simply removing BaseModel from the hierarchy fixes the issue. I however don't have the big picture knowledge necessary to know what are the impacts of that change

Any chance for a fix ?

[SimFG]

This seems to need to be fixed in langchain repo

[BastienKovac]

This issue seems to still be present using Langchain v0.0.312 and GPTCache v0.1.42 :

>>> from gptcache.adapter.langchain_models import LangChainLLMs
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/kovac/.pyenv/versions/ENV/lib/python3.11/site-packages/gptcache/adapter/langchain_models.py", line 30, in <module>
    class LangChainLLMs(LLM, BaseModel):
TypeError: metaclass conflict: the metaclass of a derived class must be a (non-strict) subclass of the metaclasses of all its bases

[SimFG]

Thank you very much for your feedback, I will fix it when I have free time. Of course, if you know more about this kind of problem, I look forward to your fix pr

[yacchi]

This problem occurs when Pydantic v2 is installed.

It is explained on the Langchain side and can be solved by using Pydantic v1 and v2 without mixing them.(https://python.langchain.com/docs/guides/pydantic_compatibility)

The following classes used in gptcache/adapter/langchain_models.py both already inherit from pydantic.BaseModel.

• langchain.llms.base.LLM
• langchain.chat_models.base.BaseChatModel

Currently, the Langchain side explicitly uses Pydantic v1, so when the GPTCache side is in a situation to inherit v2, an error occurs.

As already suggested, it would be best to remove BaseModel inheritance from LangChainLLMs and LangChainChat.

I am currently encountering this issue myself, and since I cannot lower the version of Pydantic or Langchain, I have added code to my repository to work around it. May I submit this as a pull request?

[SimFG]

yes, you can do it

[anggara-kaskus]

thank you very much @yacchi . i've tried to use your forked repository and it works well!