Add OpenSearch Serverless (AOSS) support

[norrishuang]

Summary

Add support for Amazon OpenSearch Serverless (AOSS) to the awsopensearch client.

Changes

• config.py: Add is_serverless and aws_region fields; use AWS SigV4 authentication via requests-aws4auth for AOSS connections
• cli.py: Add --serverless flag and --aws-region option
• aws_opensearch.py: Adapt client for AOSS constraints:
  • Skip unsupported operations (cluster settings, force merge, manual refresh, replica updates, warmup API)
  • Use smaller batch size (100) for bulk inserts
  • Store id as a document field (AOSS doesn't support custom _id)
  • Retrieve id from _source in search results
  • Remove engine and encoder from method config (AOSS manages internally)
  • Disable http_compress to avoid SigV4 checksum verification failures
• README.md: Add OpenSearch Serverless section with usage example and notes

Usage

vectordbbench awsopensearch --db-label aoss \
  --serverless --aws-region us-east-1 \
  --host <collection-id>.aoss.us-east-1.on.aws --port 443 \
  --case-type Performance768D1M \
  --m 16 --ef-construction 200 --ef-search 40 \
  --number-of-shards 8 --number-of-replicas 0 \
  --engine faiss --metric-type cosine \
  --num-concurrency 80,100,120

Prerequisites

• AWS credentials configured
• requests-aws4auth installed
• IAM identity policy with aoss:APIAccessAll
• AOSS Data Access Policy granting index/collection permissions

Testing

Tested against a live AOSS collection with 1M 768-dim vectors (Cohere dataset). Data loading and index creation verified successfully.

[XuanYang-cn]

I left a few inline comments for the AOSS paths that look like they still need fixes before merge.

[norrishuang]

Thanks for the review, @XuanYang-cn! I've addressed all three comments in 57f7448:

1. Multi-client insert path: Serverless now always routes through the single-client insert path (insert_embeddings() short-circuits when serverless), so AOSS loads no longer hit the multi-client path that sends a custom _id. This keeps the no-custom-_id / _source.id / small-batch behavior for all serverless loads regardless of number_of_indexing_clients.

2. Filters: prepare_filter() now builds the NumGE range on the stored id field for serverless (instead of _id), and the mapping stores id as a numeric long field so range queries work. Filtered cases like NewIntFilterPerformanceCase will now filter on the inserted IDs correctly.

3. Dependencies: Added boto3 and requests-aws4auth to the opensearch extra in pyproject.toml (and to install/requirements_py3.11.txt), and updated the README serverless prerequisites to reference pip install 'vectordb-bench[opensearch]' and mention boto3. A clean vectordb-bench[opensearch] install now includes everything needed for serverless.

PTAL when you get a chance.

/assign @XuanYang-cn

[XuanYang-cn]

/lgtm

[sre-ci-robot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: norrishuang, XuanYang-cn

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Details

Needs approval from an approver in each of these files:

• OWNERS [XuanYang-cn]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment