I have been using Weaviate free 14 day cluster and am trying to use OpenAI embeddings and store in a Weaviate collection. OpenAI’s website mentions about having a Batch API where there are higher rate-limits and the resulting embeddings are ready after 24 hours (at a 50% cost of the original real-time vector generation which is right now 3 requests per minute)
Is there a way to configure the Weaviate vectorizer using the same OpenAI batch API?Batch - OpenAI API
Client Language and Version: Python
Multitenancy: No
hi @SmitNGRA ! Welcome to our community
AFAIK, we do not have this implemented. Yet.
I have found this Github issue that is related:
opened 02:54AM - 03 Apr 24 UTC
feature request
### Describe your feature request
With async indexing being released, the bottl… eneck to import data has moved to the vectorizer modules which often require 3rd party api calls to convert text/image chunks into vectors. There has been recent work https://github.com/weaviate/weaviate/pull/4546 https://github.com/weaviate/weaviate/pull/4578 to switch vectorizer modules to use batching wherever possible. We can further reduce import times by enabling async indexing at the module level as well.
At a high level:
- [ ] Switch to async indexing straight to disk https://github.com/weaviate/weaviate/pull/3974.
- [ ] If a module is enabled, async workers should read a batch of objects from disk (not vectors), use the module's `BatchVectorizer` to generate the vectors, and then write to the vector index as usual.
- [ ] There needs to be a solution for persistent failures including surfacing errors to users.
### Code of Conduct
- [X] I have read and agree to the Weaviate's [Contributor Guide](https://weaviate.io/developers/contributor-guide) and [Code of Conduct](https://weaviate.io/service/code-of-conduct)
Please, leave your thumbs up
Thanks!
1 Like
there so we can track it’s popularity!