3.2.0

Release 3.2.0

3.1.0

LumoKit 3.1.0

Highlights

• Performance optimizations in chunking hot paths (overlap handling, sliding windows, reduced allocations).
• Non-chunking parse-path audit follow-up:
  • shared parsing pipeline for parseDocument and parseAndIndex
  • stronger URL/file/empty-document validation behavior
  • improved PicoDocs stage context in wrapped errors
• Expanded public API regression coverage for URL/file/empty-document edge cases.

Included PRs

• #11 Optimize chunking hot paths and reduce allocation overhead
• #12 Audit follow-up: refactor parse path and harden API validation

Notes

• README install snippet updated to from: "3.1.0".

3.0.0

Highlights

• Align dependency resolution with the current VecturaKit upstream constraints by tracking VecturaKit main.
• Refresh README installation snippet to from: \"3.0.0\".
• Clean up test output by fixing an unused-result warning in public API tests.

Audit results

• swiftlint --strict passed.
• swift build passed.
• swift test --no-parallel passed.
• xcodebuild clean build -scheme LumoKit -destination \"platform=macOS\" -skipPackagePluginValidation passed.

Notes

• Attempting to consume VecturaKit as a strict stable-version requirement currently fails resolution because upstream depends on swift-embeddings via an unstable requirement. This release reflects the working resolution path with current upstream state.

2.0.0

Changes since 1.1.2

• Fix paragraph oversized overlap behavior (#2)
• Audit fixes and improvements (#4)
• Add document count API and return IDs from indexing; use Vectura documentCount property
• Stabilize package dependencies (#6)
• Add deleteChunks(ids:) API and docs (#8)
• Fix WordChunker logic and chunking tests (#8)
• Fix markdown oversized section chunking and add regression test (#9)
• README updates (Star History chart, version bump to 2.0.0)

1.1.1

Code refactoring and improvements

• Refactored SemanticChunker into smaller, focused modules
• Fixed all SwiftLint violations including type body length and function parameter count
• Organized test suite into separate files by functionality
• Improved error handling in concurrent operations
• Enhanced code organization and maintainability

1.1.0

Full Changelog: v1.0.0...1.1.0

LumoKit v1.0.0

LumoKit v1.0.0 — What's New

Stable 1.0 release of LumoKit for on‑device Retrieval‑Augmented Generation (RAG).

Highlights

• Async initialization: public init(config:) is now async to align with VecturaKit’s async initializer; use try await LumoKit(config:).
• End‑to‑end ingestion: Parse documents with PicoDocs, chunk text, and index/search via VecturaKit.
• Strong defaults: Auto‑tuned search options with hybrid scoring support from VecturaKit.

Documentation

• Overhauled README with:
  • Table of Contents and API overview
  • Copy‑pasteable examples: single file indexing, batch folder ingestion, parse‑only, custom storage location, error handling
  • Practical tips and platform/Swift requirements

Breaking Changes

• LumoKit.init(config:) changed from sync to async. Update call sites to:

  let lumoKit = try await LumoKit(config: config)

Requirements

• Swift 6.2+
• iOS 18.0+ / macOS 15.0+

Upgrade Notes

• Replace any prior synchronous initializer usage with the async form.
• No other source‑level changes are required for existing APIs.

v0.2.1

Update Package.resolved

0.2.0

v0.2.0

Update Package.swift

0.1.0

Version 0.1.0

Initial release. Enjoy!