Агент langchain с инструментом: «ошибка» не является допустимым инструментом ⇐ Python

[Anonymous]

Проблема
когда я запускаю инструмент ping сам по себе, он работает на 100%, ниже вывод показывает, что агент правильно вызывает инструмент с именем сервера, но затем он печатает " не является допустимым инструментом"
Вывод:

Код: Выделить всё

root@7c9a30701184:/app# python llm_toolchain.py
> Entering new AgentExecutor chain...
Assistant: Do I need to use a tool? Yes.
Action: ping_server(tool_input='localhost', callbacks='Callbacks' = None)
Action Input: localhostping_server(tool_input='localhost', callbacks='Callbacks' = None) is not a valid tool, try one of [ping_server].
Thought: Do I need to use a tool? Yes
Action: ping_server(tool_input='localhost', callbacks='Callbacks' = None)
Action Input: localhostping_server(tool_input='localhost', callbacks='Callbacks' = None) is not a valid tool, try one of [ping_server].Great! Let's get started. You've asked me if the localhost server is up. To answer this question, I need to use a tool. Here's what I'll do:

Action: ping_server(tool_input='localhost', callbacks='Callbacks' = None)

Please provide the output of the tool so I can observe the result.Invalid Format: Missing 'Action Input:' after 'Action:'Great, let's get started! You've asked me if the localhost server is up. To answer this question, I need to use a tool. Here's what I'll do:

Action: ping_server(tool_input='localhost', callbacks='Callbacks' = None)

ИСХОДНЫЙ КОД(llm_toolchain.py):

Код: Выделить всё

import requests

from langchain import hub
from langchain.agents import Tool,AgentExecutor, create_react_agent
from langchain_ollama.llms import OllamaLLM
from langchain.memory import ConversationBufferWindowMemory
from pydantic import BaseModel, Field
from langchain.tools import tool

# ----------------------------------------------------
#  Ping Tool
# ----------------------------------------------------
class PingServer(BaseModel):
serverName: str = Field(description="ServerName")
@tool("ping_server", args_schema=PingServer, return_direct=False)
def ping_server(serverName: str) -> str:
'''Tests a server to see if it is online or available. usage: ping_server("server1")'''
import os
response = os.system(f"ping -c 1 {serverName}")
if response == 0:
result= (f"{serverName} is up!")
else:
result= (f"{serverName} is down!")
return ({"text": result})
# ----------------------------------------------------
#  Update chat history
# ----------------------------------------------------
def append_chat_history(input, response):
chat_history.save_context({"input": input}, {"output": response})
def invoke(input):
msg = {
"input": input,
"chat_history": chat_history.load_memory_variables({}),
}
print(f"Input: {msg}")
response = agent_executor.invoke(msg)
print(f"Response: {response}")
append_chat_history(response["input"], response["output"])
print(f"History: {chat_history.load_memory_variables({})}")

# ----------------------------------------------------
#  Define Tool for agent
# ----------------------------------------------------

tools = [
Tool(
name="ping_server",
func=ping_server,
description="Useful when checking if a server is online or not. Input: ServerName",
),]

prompt = hub.pull("hwchase17/react-chat")
chat_history = ConversationBufferWindowMemory(k=10)

llm = OllamaLLM(model="llama2",  # you can use a stronger model like LLama3 or Mistral, i dont have a lot of VRAM
keep_alive=-1, # keep the model loaded indefinitely
base_url="http://ollama:11434", # use the local ollama server
)
agent = create_react_agent(llm, tools, prompt, stop_sequence=True)
agent_executor = AgentExecutor(agent=agent,
tools=tools,
verbose=True,
max_iterations=2,
handle_parsing_errors=True)

invoke("is the localhost server up?")

Я просмотрел документацию и переопределил инструмент несколькими способами.
Я попробовал другие известные рабочие примеры.

Подробнее здесь: https://stackoverflow.com/questions/791 ... valid-tool