Skip to main content
Version: 0.4.0

SochDB v0.2.9 Release Notes

Release Date: January 2, 2026
Focus: Comprehensive benchmarking, SQL engine maturity, and community infrastructure

🎯 Highlights​

Real-World Performance Benchmarks​

  • 3× faster than ChromaDB for vector search (0.45ms vs 1.37ms p50)
  • 22× faster than LanceDB for vector search (0.45ms vs 9.86ms p50)
  • >98% recall@10 with sub-millisecond latency using real Azure OpenAI embeddings
  • End-to-end RAG analysis: Embedding API is 333× slower than database operations

Full SQL Engine in Go SDK​

  • Complete SQL-92 support (CREATE, INSERT, SELECT, UPDATE, DELETE)
  • Feature parity with Python and JavaScript SDKs
  • Embedded server mode with zero external dependencies

Community & Open Source​

  • Added CODE_OF_CONDUCT.md (Contributor Covenant v2.1)
  • Added SECURITY.md with vulnerability reporting policy
  • Created comprehensive issue templates (bug, feature, support)
  • Unified release workflow for automated SDK publishing

📊 Benchmark Results Summary​

Vector Search Comparison (1,000 real documents)​

DatabaseSearch p50Search p99Insert Rate
SochDB0.45ms ✅0.61ms ✅7,502 vec/s
ChromaDB1.37ms1.73ms3,237 vec/s
LanceDB9.86ms21.63ms18,106 vec/s

Search Quality (Recall@k with Real Embeddings)​

ConfigurationRecall@10Latency
M=8, ef_c=50 (recommended)99.1%0.42ms
M=16, ef_c=10098.2%0.47ms
M=16, ef_c=20098.8%0.44ms

Key Finding: SochDB achieves >98% recall with real embeddings while maintaining sub-millisecond search latency.

End-to-End RAG Bottleneck​

OperationTime% of Total
Embedding API (Azure OpenAI)59.5s99.7%
SochDB Insert (1K vectors)0.133s0.2%
SochDB Search (100 queries)0.046s0.1%

Insight: The database is never the bottleneck in production RAG systems—your LLM API calls are.

🆕 New Features​

Go SDK​

  • Full SQL engine with DDL/DML operations
  • SQL transaction support
  • WHERE clause with operators (=, !=, <, >, <=, >=, LIKE)
  • ORDER BY, LIMIT, OFFSET support
  • Complete feature parity with Python/JS SDKs

Benchmarking Suite​

  • Real-world embedding tests with Azure OpenAI
  • Comparative benchmarks vs ChromaDB and LanceDB
  • Recall@k quality measurements
  • 360° performance report (quality, latency, throughput, efficiency)

Community Infrastructure​

  • Contributor Covenant v2.1 Code of Conduct
  • Security vulnerability reporting policy with SLAs
  • YAML-based issue templates with validation
  • Automated release workflow with PR-based deployment

🔧 Improvements​

Documentation​

  • Updated all SDK guides with v0.2.9 features
  • Added comprehensive benchmark methodology
  • Improved README with real-world performance data
  • Added release notes documentation

Build & Release​

  • Unified release workflow across all SDKs
  • Protected branch support with PR-based releases
  • Automated changelog generation
  • Fixed YAML syntax in GitHub Actions workflows

🐛 Bug Fixes​

  • Fixed Rust compilation errors in storage.rs
  • Fixed Go SDK test output formatting (removed redundant newlines)
  • Fixed wire protocol documentation for all SDKs
  • Fixed GitHub Actions YAML syntax errors

📦 Installation​

Python​

pip install sochdb

Node.js / TypeScript​

npm install @sochdb/sochdb

Go​

go get github.com/sochdb/sochdb-go@latest

Rust​

[dependencies]
sochdb = "0.2.9"

📚 Resources​

🔮 What's Next (v0.3.0)​

  • Replication and clustering support
  • Advanced query optimization
  • Additional vector index types (IVF, IVFPQ)
  • Enhanced monitoring and observability
  • GraphQL API support

🙏 Acknowledgments​

Thanks to all contributors and the open source community for feedback, testing, and support!

For questions or support: