onepointconsulting / elasticsearch-agent

ElasticSearch agent based on ElasticSearch, LangChain and ChatGPT 4

Elasticsearch Langchain Agent

This project uses ElasticSearch together with LangChain and ChatGPT 4 to build an agent with which you can ask intelligent questions on top of an ElasticSearch cluster.

Setup

We suggest to use Conda to manage the virtual environment and then install poetry.

conda activate base
conda remove -n elastic_search_playground --all
conda create -n elastic_search_playground python=3.11
conda activate elastic_search_playground
pip install poetry

Installation

poetry install

Configuration

The configuration should be saved in a .env file.

OPENAI_API_KEY=<key>
OPENAI_MODEL=gpt-4-0613
# OPENAI_MODEL=gpt-3.5-turbo-16k-0613
REQUEST_TIMEOUT=300
LANGCHAIN_CACHE=false
CHATGPT_STREAMING=false
LLM_VERBOSE=true

# Elastic Search related
ELASTIC_SERVER=https://127.0.0.1:9200
ELASTIC_USER=elastic
ELASTIC_PASSWORD=<pass>
ELASTIC_VERIFY_CERTIFICATES=false

ELASTIC_INDEX_DATA_FROM=0
ELASTIC_INDEX_DATA_SIZE=5
ELASTIC_INDEX_DATA_MAX_SIZE=50

LANGCHAIN_VERBOSE=true
AGGS_LIMIT=200
TOKEN_LIMIT=6000
MAX_SEARCH_RETRIES = 100

Please note that the agent works best with ChatGPT 4.