LikeLike

LikeLike

Thank you very much!!

LikeLike

I have updated source code from LangChain 0.0.100 to 0.0.300 (current).

LikeLike

Thank you for your feedback.
I’m sorry, but this source code runs on langchain library version 0.0.100 (old version) and you might have run this program on recent version.
In current langchain version (today, 0.0.288), instead of implementing _extract_tool_and_input() method, you should divide this implementation into parser class and should instantiate this parser class in _get_default_output_parser() method.
See ReActDocstoreAgent in [langchain package dir]/agents/react/base.py (which is the implementation of above react-docstore agent) and ReActOutputParser in [langchain package dir]/agents/react/output_parser.py (which is the implementation of its parser class).

LikeLike

However, when I try to run the script, I get struck in the agent creation step

—————————————————————————
TypeError Traceback (most recent call last)
Cell In[11], line 4
1 ##########
2 # run agent
3 ##########
—-> 4 agent = ReActTestAgent.from_llm_and_tools(
5 llm,
6 tools
7 )

File …/lib/python3.9/site-packages/langchain/agents/agent.py:559, in Agent.from_llm_and_tools(cls, llm, tools, callback_manager, output_parser, **kwargs)
557 tool_names = [tool.name for tool in tools]
558 _output_parser = output_parser or cls._get_default_output_parser()
–> 559 return cls(
560 llm_chain=llm_chain,
561 allowed_tools=tool_names,
562 output_parser=_output_parser,
563 **kwargs,
564 )

TypeError: Can’t instantiate abstract class ReActTestAgent with abstract method _get_default_output_parser

Do we have to specifically define the output parser? Any thoughts on this?

LikeLike

LikeLike

LikeLike

LikeLike

LikeLike