[PROPOSAL] Forming OpenSearch Search Technical Advisory Group (TAG)

[getsaurabh02]

What/Why

What are you proposing?

We propose the formation of a "Search" Technical Advisory Group (TAG) under the OpenSearch Technical Steering Committee (TSC). The Search TAG will guide technical direction and decision-making across all aspects of search - including core text search, vector search, semantic search, hybrid retrieval, relevance tuning and performance.

As OpenSearch expands into analytics, AI, and observability, search remains its foundation. This TAG ensures a unified strategy for advancing OpenSearch’s search experience and aligning community efforts across repositories.

Previous Discussions: #48
TSC Meeting: https://github.com/opensearch-project/technical-steering/blob/main/meeting-minutes/2025/2025-06-11-minutes.md

Purpose:

• Centralize expertise to accelerate clarity in search evolution and high level technical decisions.
• Guide roadmap, extensibility and design discussions on query engine, vectors, performance, ranking, hybrid, and semantic search.
• Promote alignment between core maintainers, plugin owners, and community contributors.
• Ensure OpenSearch remains competitive in search, relevance, performance, and usability.

Key Responsibilities

• Technical Guidance: Review and advise on proposals impacting search behavior, query engine, or ranking logic.
• Roadmap Alignment: Help prioritize and unify search-related efforts across teams.
• Best Practices: Publish reference patterns for scaling, benchmarking, and tuning search and relevance across releases.
• Cross-Domain Collaboration: Coordinate with ML and Observability TAGs where search intersects.
• Pain-Point Resolution: Address any long-standing issues in query engine, hybrid relevance, query performance, and ranking.
• Advisory Reports: Submit recommendations to the TSC with clear technical rationale.

Pain Points Addressed

• Unified ownership and consistent decision flow across search repos.
• Consolidated roadmap updates for text, vector, and hybrid search space.
• Minimize redundant design efforts between core and plugin features with end to end perspective.
• Faster iteration on performance and relevance improvements with better accountability.

Desired Expertise for initial founding members:

• Lucene / OpenSearch core internals
• Text analysis and ranking
• Vector / semantic and hybrid retrieval
• Large-scale search performance and benchmarking

Next Steps

I am seeking community feedback on this proposal and the initial founding member list based on interest shown in this thread. After socializing this and identifying interested participants here, I will open a PR with a proposed charter that incorporates the feedback received.

cc: @Pallavi-AWS @vamshin @andrross @msfroh @epugh @rishabhmaurya @jainankitk @mch2 @yupeng9 @navneet1v

[navneet1v]

+1

[vamshin]

+1

[krisfreedain]

Great to see - this has been an open idea for months now:

#48

@getsaurabh02 I'd recommend closing this and just continue the discussion on the original issue instead of opening a dup.

Also, same feedback as the other issue you opened today - start a PR with the charter and readme so the community can discuss in-line #71 (comment)

[getsaurabh02]

Thanks @krisfreedain Can we close #48 to avoid dup and continue here instead? The original issue was more of an idea with the TSC meeting discussion, whereas here we've captured more details on the responsibilities, purpose, and key pain points here in this issue.

We can also use this thread to invite initial members. We have couple of +1s already. Once we reach consensus, I'll open a PR summarizing the feedback from this discussion as outlined in the next steps.

[krisfreedain]

@getsaurabh02 since "original issue was more of an idea with the TSC meeting discussion" I believe it would be better to keep the original. Not sure why it was duplicated.

[krisfreedain]

closed out the original issue on this. Please try not to create duplicate work in the future.

#48

[getsaurabh02]

Thanks @krisfreedain
To me the original referred here (#48 ) stood out as an action item from TSC meeting to open a proposal - instead of a proposal itself (hence the confusion).