Prathmesh Deshmukh created SOLR-18127:
-----------------------------------------
Summary: Introduce pluggable solr native text-to-vector
TextToVectorModel interface + LangChain4j compatibility adapter
Key: SOLR-18127
URL: https://issues.apache.org/jira/browse/SOLR-18127
Project: Solr
Issue Type: Improvement
Reporter: Prathmesh Deshmukh
Introduce a Solr‑native {{TextToVectorModel}} interface to support pluggable
embedding implementations without requiring LangChain4j.
The current Text‑to‑Vector module {{SolrTextToVectorModel}} is tightly coupled
to LangChain4j’s {{EmbeddingModel}} API. As a result, only the few
LangChain4j‑bundled providers (OpenAI, HuggingFace, etc.) are supported out of
the box. Anyone who wants to integrate a custom embedding endpoint must
implement the full LangChain4j interface and builder pattern, even when they
don’t use LangChain4j in their system.
There’s no simple Solr‑native abstraction for “text → float vector” that custom
implementations can plug into.
This proposal introduces a small native interface that decouples Solr’s vector
feature from LangChain4j, while keeping full backward compatibility.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
