How to configure vectorizer for contextionary in python v3?

Support
1

Description

How to configure contextionary vectorizer in python v3?

I can’t find documentation about setting URL(s).

I use this dict:

collection_kwargs = {
"name": collection_name,
"properties": [...],
"vector_index_config": wvc.config.Configure.VectorIndex.hnsw(),
wvc.config.Configure.Vectorizer.text2vec_contextionary(
    ?????
)
}

Server Setup Information

  • Weaviate Server Version:
  • Deployment Method: k8s
  • Client Language and Version:
2

Hi! Check here on how to specify the vectorizer:

You can change between the tabs for different client languages and version.

Note that python v3 is deprecated: Legacy (v3) API (DEPRECATED) | Weaviate and you should uprade to newer versions.

Thanks!

3

Correction, it looks like I am using v4 as it configures like:

collection_kwargs = {
“name”: collection_name,
“properties”: […],
“vector_index_config”: wvc.config.Configure.VectorIndex.hnsw(),
“vectorizer_config”: wvc.config.Configure.Vectorizer.text2vec_contextionary(
???
)
}

collection = client.collections.create(**collection_kwargs)

But I cannot find any configuration information about Contextionary specifically there. How is the URL specified?

4

This module is really old and isn’t used anymore, other than experimenting or testing.

There is a docker image that goes along with that module:

If you want to experiment with a local inference model, a great starting point is using ollama: Text Embeddings | Weaviate

Let me know if that helps!

thanks!