You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@solr.apache.org by Rahul Goswami <ra...@gmail.com> on 2025/02/19 22:23:05 UTC
Dense vector search efforts in Solr
I came across this JIRA where support for text-to-vector for a query was
added:
https://issues.apache.org/jira/browse/SOLR-17525
In the ticket, Alessandro mentions:
"Other tasks I see coming:
- UpdateRequestProcessor to add the indexing time support
- performance testing
- support for additional embedding models (including potentially an
In-Process one)"
If it is still up for grabs, I would like to contribute to the index time
text-to-vector conversion through an UpdateRequestProcessor. I have played
around with this idea using the ONNX runtime to generate embeddings
natively in Java through an update processor, but it needs some more
polishing. Can somebody please let me know if there is a JIRA for this or
similar efforts already?
Thanks,
Rahul
Re: Dense vector search efforts in Solr
Posted by Alessandro Benedetti <a....@sease.io>.
Hi Rahul, nice to meet you!
The update request processor is pretty much done, we are close to merging:
https://github.com/apache/solr/pull/3151.
In regards to adding in-process embedding services implementation, It's a
slippery slope and I attempted some experiments using langChaing4j (the
library we use in Solr to interact with LLMs):
https://docs.langchain4j.dev/integrations/embedding-models/in-process/
I'm not sure it's going to be production-grade to serve in-process
embeddings but I believe it could be a nice addition to help people
experimenting and playing, so feel free to add the integration and open a
pull request!
We'll be happy to review it!
Get familiar with the 'llm' module (on main and from 9.8) as there's the
home for all new LLM integrations (including vectorisation).
Cheers
--------------------------
*Alessandro Benedetti*
Director @ Sease Ltd.
*Apache Lucene/Solr Committer*
*Apache Solr PMC Member*
e-mail: a.benedetti@sease.io
*Sease* - Information Retrieval Applied
Consulting | Training | Open Source
Website: Sease.io <http://sease.io/>
LinkedIn <https://linkedin.com/company/sease-ltd> | Twitter
<https://twitter.com/seaseltd> | Youtube
<https://www.youtube.com/channel/UCDx86ZKLYNpI3gzMercM7BQ> | Github
<https://github.com/seaseltd>
On Wed, 19 Feb 2025 at 23:23, Rahul Goswami <ra...@gmail.com> wrote:
> I came across this JIRA where support for text-to-vector for a query was
> added:
> https://issues.apache.org/jira/browse/SOLR-17525
>
> In the ticket, Alessandro mentions:
> "Other tasks I see coming:
>
> - UpdateRequestProcessor to add the indexing time support
> - performance testing
> - support for additional embedding models (including potentially an
> In-Process one)"
>
> If it is still up for grabs, I would like to contribute to the index time
> text-to-vector conversion through an UpdateRequestProcessor. I have played
> around with this idea using the ONNX runtime to generate embeddings
> natively in Java through an update processor, but it needs some more
> polishing. Can somebody please let me know if there is a JIRA for this or
> similar efforts already?
>
> Thanks,
> Rahul
>
Re: Dense vector search efforts in Solr
Posted by Alessandro Benedetti <a....@sease.io>.
Hi Rahul, nice to meet you!
The update request processor is pretty much done, we are close to merging:
https://github.com/apache/solr/pull/3151.
In regards to adding in-process embedding services implementation, It's a
slippery slope and I attempted some experiments using langChaing4j (the
library we use in Solr to interact with LLMs):
https://docs.langchain4j.dev/integrations/embedding-models/in-process/
I'm not sure it's going to be production-grade to serve in-process
embeddings but I believe it could be a nice addition to help people
experimenting and playing, so feel free to add the integration and open a
pull request!
We'll be happy to review it!
Get familiar with the 'llm' module (on main and from 9.8) as there's the
home for all new LLM integrations (including vectorisation).
Cheers
--------------------------
*Alessandro Benedetti*
Director @ Sease Ltd.
*Apache Lucene/Solr Committer*
*Apache Solr PMC Member*
e-mail: a.benedetti@sease.io
*Sease* - Information Retrieval Applied
Consulting | Training | Open Source
Website: Sease.io <http://sease.io/>
LinkedIn <https://linkedin.com/company/sease-ltd> | Twitter
<https://twitter.com/seaseltd> | Youtube
<https://www.youtube.com/channel/UCDx86ZKLYNpI3gzMercM7BQ> | Github
<https://github.com/seaseltd>
On Wed, 19 Feb 2025 at 23:23, Rahul Goswami <ra...@gmail.com> wrote:
> I came across this JIRA where support for text-to-vector for a query was
> added:
> https://issues.apache.org/jira/browse/SOLR-17525
>
> In the ticket, Alessandro mentions:
> "Other tasks I see coming:
>
> - UpdateRequestProcessor to add the indexing time support
> - performance testing
> - support for additional embedding models (including potentially an
> In-Process one)"
>
> If it is still up for grabs, I would like to contribute to the index time
> text-to-vector conversion through an UpdateRequestProcessor. I have played
> around with this idea using the ONNX runtime to generate embeddings
> natively in Java through an update processor, but it needs some more
> polishing. Can somebody please let me know if there is a JIRA for this or
> similar efforts already?
>
> Thanks,
> Rahul
>