Skip to content

Latest commit

 

History

History
38 lines (31 loc) · 1.11 KB

File metadata and controls

38 lines (31 loc) · 1.11 KB

Late Chunking Retrieval Experiment

Contents

  1. Summary
  2. Presentation
  3. Architecture
  4. Features
  5. Prerequisites
  6. Installation
  7. Usage

Summary

This is a demonstration usage of late chunking via a Jina embedding model and Elastic vector database capabilities.

Presentation

https://joeywhelan.github.io/late-chunking/

Architecture

architecture

Features

  • Jupyter notebook
  • Builds an Elastic Serverless deployment via Terraform
  • Indexes two data sets and then compares late chunking performance on each
  • Deletes the entire deployment via Terraform

Prerequisites

  • uv
  • terraform
  • Elastic Cloud account and API key
  • Jina API key
  • Python

Installation

  • Edit the terraform.tfvars.sample and rename to terraform.tfvars
  • Create a Python virtual environment

Usage

  • Execute notebook