chATLAS_Chains 0.1.2

A modular Python package for implementing Retrieval Augmented Generation chains for the chATLAS project.

chATLAS_Chains

This package implements and benchmarks various Retrieval Augmented Generation (RAG) chains for use in the chATLAS project.

Installation

conda create -n venv chatlas_chains_env python=3.10
conda activate chatlas_chains_env
pip install chatlas-chains

Environment variables

These are required for the following use cases

1. Using an OpenAI LLM

export CHATLAS_OPENAI_KEY='your api key'

2. Benchmarking, set the path to the question set

export CHATLAS_BENCHMARK_QUESTIONS=/path/to/questions.josn

Available Chains

• chains.basic.basic_retrieval_chain
• chains.basic_graph.basic_retrieval_graph

Benchmarking

To benchmark e.g. the chains in chATLAS_Chains.chains.basic run this from the project root

python benchmark/basic.py

Testing

The tests require a running postgres server to work. If on lxplus you can modify TEST_DB_CONFIG in tests/conftest.py to connect to the chATLAS server.

If you want to create a local dummy postgres server, you need to install psql. This can be done on macOS using homebrew:

Software install

brew install postgresql
brew services start postgresql
brew install pgvector
brew unlink pgvector && brew link pgvector

Create a user

psql -h localhost -U postgres
ALTER USER postgres WITH PASSWORD 'Set_your_password_here';
CREATE EXTENSION IF NOT EXISTS vector;