Skip to content

hsm207/weaviate-txtai

Repository files navigation

Introduction

This package aims to provide a convenient way to use weaviate as a document store when building solutions using txtai.

Installation

pip install weaviate-txtai

Usage

Here is an example of a txtai configuration that uses this package for semantic search:

embeddings:
  path: sentence-transformers/nli-mpnet-base-v2

nop:

weaviate_txtai.client.Weaviate:
  url: http://weaviate:8080
  custom_schema:
      class: "Post"
      properties:
       - name: "content"
         dataType:
          - text
      vectorIndexConfig:
       distance: "dot"

workflow:
  index:
    batch: 2500
    tasks:
    - action: [nop, transform]
      unpack: False
    - action: weaviate_txtai.client.Weaviate
      unpack: False
  search:
    tasks:
    - action: transform
    - action: weaviate_txtai.client.Weaviate
      args: [search]

Read this blog post for an end-to-end example.

Roadmap

¯\_(ツ)_/¯

TODOs

  • Assess community interest in this package
  • Figure out roadmap to 1.0.0 release
  • Write docs for contributing