We have an experimental release of Search Pipelines in OpenSearch 2.8 that starts at query submission and ends at result fetching. In between, we see any number of possibilities like query rewriting, analyzers, reranking, etc. that can be integrated as processors in the pipeline. We want to hear more from the community as we start to evolve the code in the repo where this issue lives into the pipeline: [RFC] Search pipelines · Issue #80 · opensearch-project/search-processor · GitHub.
The scripting processor that we shipped in 2.8 doesn’t work the way it did in that demo (since that was a hacky implementation that only worked with query_string queries).
Still, the demo (I hope) conveys the general idea of the search pipelines feature and its motivation.
We are keeping this open for ideas and input after the OpenSearch 2.9 release which includes Search Pipelines. Here are some processors on our roadmap. We would be happy to take contributions of any kind (PRs for new processors, feedback on the ones we have in the roadmap, and anything in between):