Hi I already found this class but the

ttrover

Hi. I already found this class, but the problem is that I don't know how to use ElastiSearch with it? llama-index has only ElasticSearch reader, not index.

7 comments

LLogan M

I think the best solution here would be to convert the elastisearch reader into a function tool, and then use that tool with an agent

Small example (note that the docstring is used as the tool description, this is how the agent decides to use the tool or not)

Plain Text

def multiply(a: int, b: int) -> int:
    """Multiple two integers and returns the result integer"""
    return a * b

from llama_index.tools import FunctionTool
multiply_tool = FunctionTool.from_defaults(fn=multiply)

from llama_index.agent import OpenAIAgent
from llama_index.llms import OpenAI

llm = OpenAI(model="gpt-3.5-turbo")
agent = OpenAIAgent.from_tools([multiply_tool], llm=llm, verbose=True)

ttrover

HI!
I have another question: and if I use this chat_engine mode, will the generated questions go to ElasticSearchReader?
chat_engine = index.as_chat_engine(chat_mode="condense_question")

LLogan M

The method I detailed will really only work for agents -- the chat engines all rely on indexes, but you aren't using an index 🤔

LLogan M

but functionally they are extremely similar

ttrover

How to use agent with custom llm? Eg. from hugging face

LLogan M

You can use the ReActAgent, which works with any LLM. But keep in mind open-source LLMs kind of suck for agents right now

LLogan M

the react agent needs pretty structured LLM outputs, and thats where open-source kind of fails

Add a reply

Find answers from the community

Hi I already found this class but the