optional custom index settings in vectorstore #45
Conversation
* Allow kwargs in body_func * Remove async implementation * Remove unused imports
|
The main benefit of this parameter is to pre-configure the index when using the Elasticsearch vector store, which implicitly creates an index when first documents are added. One can, for example, set a correct number of replicas for his cluster, so that he has not to change it manually after creation, with all the waste of computational effort this implies |
* async support * format * fixed unit test failures * sort imports * more code reformat * typing fixes * typing fixes * one more reformat * typing fixes * docs * use sed to change get_messages() to messages
|
@giacbrd I had the intention to merge this PR, but it seems I'm not allowed to make changes on your PR branch. Would you like to do the rebase yourself, or should I create a separate PR with the changes? The changes now need to be made to the |
…-settings-param # Conflicts: # libs/elasticsearch/langchain_elasticsearch/vectorstores.py
|
@giacbrd Thanks, but something does not look right. This PR should be tiny, somehow it is now including all the async changes I've merged the other day, which are already merged. Maybe it'll be easier to start a fresh PR? You can fix this one if you like, but it is tricky. One way would be with a rebase from main, which is what I suggested above. |
|
hi @miguelgrinberg yes I am working on it, I have probably messed up something by combining the wrong git commands. The code is fine, but I am still not solving the git mess. I am going to open a new PR |
Hi, I have simply added a parameter to the ElasticsearchStore class in order to set default settings at index creation, that is made by the VectorStore helper in the Elasticsearch library. It is just arguments traversing calls, there is not much to test. There is also another minor fix in the doctring.