[Bug]: Incorrect config on schema.create_index silently fails

[tjkrusinskichroma]

What happened?

Below is

schema = Schema()

schema.create_index(
    config=SparseVectorIndexConfig(
        key=K.DOCUMENT, # this line is wrong! the named param is `source_key`
        embedding_function=ChromaBm25EmbeddingFunction(),
        bm25=True
    ),
    key=SPARSE_KEY
)

When providing values to SparseVectorIndexConfig, the index will silently fail to operate if the source_key value is not passed. In the case of the above, if the collection is created with the schema and documents are ingested, the Bm25 index will not be present on the documents and the user will receive no error that they provided an incorrect config.

Versions

chroma 1.2.1

Relevant log output

[UssuGit]

I also had the same issue.

I think that it is because you are using the local version of the client and not the cloud as the feature to do Hybrid Search (using sparse index) is only available in Cloud version and unfortunately the documentation is not clear with that info.

It would be great to be able to do this locally tho :)

[UssuGit]

@tjkrusinskichroma
I see that you have opened other issues on which you are using Chroma Cloud, so is that how you managed to use the SparseVectorIndexConfig to do hybrid search?

[jairad26]

Hi, thanks for the issue. this is being fixed in this PR and will be released in the next patch version
#5838

[sanketkedia]

Ping @sanketkedia for review