[
https://issues.apache.org/jira/browse/CAMEL-24533?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Work on CAMEL-24533 started by Andrea Cosentino.
------------------------------------------------
> camel-weaviate: the embeddings data-type transformer drops the document text
> on CREATE/UPDATE
> ---------------------------------------------------------------------------------------------
>
> Key: CAMEL-24533
> URL: https://issues.apache.org/jira/browse/CAMEL-24533
> Project: Camel
> Issue Type: Bug
> Components: camel-weaviate
> Reporter: Andrea Cosentino
> Assignee: Andrea Cosentino
> Priority: Major
> Fix For: 4.23.0
>
>
> WeaviateEmbeddingsDataTypeTransformer.createEmbeddingOperation (around lines
> 65-75) and updateEmbeddingOperation (around lines 77-87) build the Weaviate
> object from the declared parameters (vectorFieldName, textFieldName, text,
> collectionName), but the object body is set only to the vector and,
> optionally, the id header. The source text (text.text()) is never written
> into the object PROPERTIES, so the original document text is silently lost on
> CREATE and UPDATE.
> For a RAG pipeline this means the vector store cannot return the source
> passage that produced a match: the embedding is stored but the human-readable
> text that was embedded is discarded. The Milvus sibling transformer writes
> the text field correctly (MilvusEmbeddingsDataTypeTransformer), so this is an
> incomplete port of the same pattern.
> Fix: write the text under textFieldName into the properties map (e.g.
> maps.put(textFieldName, text.text())) on both CREATE and UPDATE.
> Additionally, the default-branch log/exception message (around line 61) omits
> the QUERY action from its list of supported actions.
> _Claude Code on behalf of oscerd_
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
