ChromaQueryTextRetriever InvalidDimensionException

[jongirard]

Describe the bug
I'm attempting to utilize the ChromaQueryTextRetriever following this example and encountering an error of chromadb.errors.InvalidDimensionException: Embedding dimension 384 does not match collection dimensionality 768.

Error message
chromadb.errors.InvalidDimensionException: Embedding dimension 384 does not match collection dimensionality 768.

Expected behavior
Expected to receive known output variable “documents”: a list of Documents.

Additional context
My documents have been embedded using OllamaDocumentEmbedder via the nomic-embed-text model. With my still developing sense of what's going on (new to this), I believe this error may be due to the default chromadb embedding_function utilizing all-MiniLM-L6-v2 as mentioned here: https://docs.trychroma.com/embeddings#default-all-minilm-l6-v2 not matching the model used to generate the embeddings.

However, because chroma-haystack (0.15.0) depends on chromadb (<0.4.20) OllamaEmbeddingFunction is not available yet, (https://github.com/chroma-core/chroma/blob/0.5.0/chromadb/utils/embedding_functions.py#L966) only just added in chromadb 0.5.0.

Describe the solution you'd like

I believe removing the pin on chomadb <0.4.20 and allowing latest v0.5.0 (which introduced OllamaEmbeddingFunction) will resolve the issue.

To Reproduce

1. Embed documents utilizing OllamaDocumentEmbedder(model="nomic-embed-text") and persist to chroma document store.
2. Attempt to utilize ChromaQueryTextRetriever(document_store)
3. Receive error chromadb.errors.InvalidDimensionException.

FAQ Check

• Have you had a look at our new FAQ page?

System:

• OS: 13.5.2
• GPU/CPU: M1 Max
• Haystack version (commit or version number): 2.0
• DocumentStore: chromadb
• Reader: N/A
• Retriever: ChromaQueryTextRetriever

[jongirard]

Created a PR here to bump the chromadb version which I believe will fix the issue by allowing the use of OllamaEmbeddingFunction.