Semantic Search and Generative Search with Cohere

This project contain two Jupyter Notebooks:

• wiki-large-setup – shows how to configure a collection with text2vec-cohere and generative-cohere modules. Then loads pre-vectorized Wikipedia data in 10 different languages
• wiki-large-query – examples for Semantic Search and Generative Search queries

Load data

Notes:

• collection name: Wikipedia
• languages included: en, de, fr, es, it, ja, ar, zh, ko, hi
• source: Cohere/wikipedia-22-12-(lang)-embeddings

How to use

To make this notebook work on your own Weaviate instance, you need to update or provide:

• access_token – auth token used to authenticate with your Weaviate instance. Note, if your Weaviate instance is configured without authentication, then you can remove the whole auth_config.
• url – provide the url for your Weaviate instance
• X-Cohere-Api-Key – this is your API key you need to get from dashboard.cohere.ai/api-keys

WCS configuration

Weaviate instances created with WCS provide all the required modules. No action required.

Instances deployed with Docker

Make sure that your docker configuration includes the following modules:

• text2vec-cohere
• generative-cohere