v1.1.4

Bug Fixes

• Fixed corrupted file path display in list command
• Added filtering for corrupted/invalid file paths
• Improved display sanitization to prevent terminal rendering issues

Full Changelog: v1.1.3...v1.1.4

v1.1.3

Bug Fixes and Improvements

• Fixed race condition when multiple files try to create database table simultaneously
• Fixed 'Directory not empty' errors when removing chunks from corrupted databases
• Added timeout handling for embedding operations (60s for single, 2min for batches)
• Improved error handling for corrupted LanceDB data files
• Added comprehensive logging and progress indicators during indexing
• Fixed hang during finalization step
• Better error recovery - indexing continues even if chunk removal fails
• Added detailed debug logging for VoyageAI and Ollama embedding providers

Full Changelog: v1.1.2...v1.1.3

v1.1.2

Changes

• Automated Homebrew formula updates on release
• Updated workflow to automatically update formula version and URL when new releases are created

Full Changelog: v1.1.1...v1.1.2

v1.1.1

Changes

• Added "search" alias to the "query" command
• Added "list" command to display all files and chunk counts in the database

Full Changelog: v1.1.0...v1.1.1

v1.1.0

What's New

Features

• Configurable minimum chunk size: Added minChunkSize option (default: 50 tokens) to prevent tiny chunk fragments. Configurable via config.yaml or --min-chunk-size CLI option.
• Automatic cleanup of deleted files: Index now automatically detects and removes files that have been deleted from the source directory during indexing runs.
• Homebrew support: QuickRAG can now be installed via Homebrew on macOS using brew install statico/quickrag/quickrag.

Improvements

• Better chunk overlap alignment: Improved overlap calculation to align to sentence and word boundaries, preventing mid-word chunk starts.
• Enhanced progress output: Indexing now shows information about deleted files being removed from the index.

Bug Fixes

• Fixed issue where tiny chunk fragments (like "ng." or "se.") were being created due to improper overlap handling.

Installation

Homebrew (macOS):

brew install statico/quickrag/quickrag

Manual download:
See releases page for pre-built binaries.

Configuration

Add to your ~/.config/quickrag/config.yaml:

chunking:
  minChunkSize: 50  # Minimum chunk size in tokens (default: 50)

Or use the CLI option:

quickrag index ./documents --min-chunk-size 50

Full Changelog: v1.0.2...v1.1.0

v1.0.2

Changes

• Add idempotent indexing: automatically skip already indexed files
• Add file tracking with modification time checking
• Add per-file transaction support (all chunks for a file are atomic)
• Improve UTF-8 sanitization for PDF documents with invalid characters
• Add --clear flag to index command for clearing existing index
• Restore batch X/Y progress reporting in indexer
• Update README with new features and usage documentation

Full Changelog: v1.0.1...v1.0.2

v1.0.1

Changes

• Fix embedding dimension mismatch between indexing and querying
• Improve dimension detection from existing databases
• Remove Homebrew support, use direct binary installation

v1.0.0

First stable release of QuickRAG

Features

• Multiple embedding providers (OpenAI, VoyageAI, Ollama)
• Configurable chunking algorithm with line number tracking
• Configuration management via ~/.config/quickrag/config.yaml
• LanceDB integration for fast vector search
• Comprehensive architecture documentation
• Homebrew support

Installation

brew tap statico/quickrag
brew install quickrag

Or build from source:

git clone https://github.com/statico/quickrag.git
cd quickrag
bun install

Full Changelog: https://github.com/statico/quickrag/commits/v1.0.0