# SearchAI Index Server 1.0.0 (2026-09-03)

First release. OpenSearch-compatible search + vector engine on an embedded
high-performance engine core, one small native binary per platform.

## Platforms (all built natively)
- searchai-index-server-1.0.0-linux-arm64.tar.gz (Graviton/ARM servers)
- searchai-index-server-1.0.0-linux-amd64.tar.gz (x86_64 glibc servers)
- searchai-index-server-1.0.0-osx-arm64.tar.gz (Apple Silicon dev)

## Highlights
- OpenSearch 2.x-compatible REST (opensearch-py/java verified); ES 8.x
  presentation mode; Qdrant/Pinecone/Chroma wire-protocol listeners.
- Vector search: 1.8x faster than OpenSearch at equal recall on c8g.large;
  fastest ingest of five engines benchmarked; filtered kNN beats Qdrant.
- Facets/sorts at or above OpenSearch speed (doc-values column cache).
- 92 MB RSS after a 100k-doc query storm (vs ~1.4 GB OpenSearch).
- Admin console (/console): Dev Tools, Documents, Search + tracing,
  Migrate (zero-touch import from OpenSearch/Elasticsearch/Qdrant/
  Pinecone/Chroma/Weaviate/Solr), Evaluate (golden queries, side-by-side
  vs your current engine, regression gates).
- Custom analyzers (stopwords, query-time synonyms, ngram, jieba);
  array + rank_features (sparse vector) fields; neural search + rerank via
  the SearchAI Inference Server (text AND image).
- Scoped API keys (index-glob grants, read/write/admin roles across all
  listeners); native TLS (https listeners, encrypted inter-node cluster,
  https outbound).
- Full-replica clustering with op-log replication and automatic failover.

## Verification
Each bundle smoke-tested on its native platform (handshake, kNN + text
search, console). Full suites at tag v1.0.0: 114 pytest (incl. cluster,
crash-recovery, TLS, security, dialects via real client libraries) +
48 zig unit/probe tests. Benchmarks: tests/bench/RESULTS.md.