1.3.34

What's Changed

• ci: add build step to prevent release failures by @GoBeromsu in #19
• fix: resolve critical best practice violations by @GoBeromsu in #20
• fix: extract magic number to named constant by @GoBeromsu in #24
• fix: use actual settings instead of constant in Tag.ts by @GoBeromsu in #23
• fix: remove unused OPENROUTER_STRUCTURE_OUTPUT (DRY) by @GoBeromsu in #22
• fix: remove duplicate Range interface (DRY violation) by @GoBeromsu in #21
• docs: improve project documentation by @GoBeromsu in #25
• refactor: generalize field value collection with Obsidian API by @GoBeromsu in #26
• fix: use model ID instead of name for selectedModel by @GoBeromsu in #27
• feat: replace model toggle with dropdown for better UX by @GoBeromsu in #28
• fix: resolve count undefined bug and improve count UX with slider by @GoBeromsu in #29

Full Changelog: 1.3.33...1.3.34

1.3.33

Full Changelog: 1.3.32...1.3.33

1.5.29

Bump version to 1.5.29

1.5.28

Bump version to 1.5.28

1.5.27

🔄 Visual Progress Indicator Update

Progress Visualization

• Real-time spinner indicator during API processing
• Clean, minimal progress display: ⠸ Tagging filename (frontmatter)
• Automatic progress management with function-based approach

Enhanced User Feedback

• Unified message styling across all notifications
• Reliability scores now displayed in success messages
• Minimal design with essential information only

1.5.25

🚀 What's New

UI Enhancements

• ✨ Enhanced button component animations for better user interaction
• 🔄 Improved internal UI logic for better performance and maintainability
• 🎯 Optimized component state management

Technical Improvements

• 🏗️ Refactored UI architecture for better maintainability
• 🔧 Streamlined internal component logic
• 🎨 Enhanced visual feedback through improved button animations

Under the Hood

• Optimized component rendering logic
• Improved state management within UI components
• Enhanced animation performance for button interactions

🐛 Bug Fixes

• Fixed potential UI state inconsistencies
• Improved component update lifecycle

1.5.22

🚀 Release Notes

✨ New Features

• Ollama Provider Added: Integrated native /api/chat endpoint with structured output

🔧 Improvements

• Centralised Provider Names: Used constants from providerPreset.json for type-safe access
• Dynamic Provider Sync: Removed hardcoded strings; names now sync with preset
• Extended Tests: Ollama responses now fully covered with mocks

🛠️ Technical Changes

• Introduced PROVIDER_NAMES for consistent naming
• Enabled JSON module import in TypeScript config
• Refactored provider logic for clarity

1.5.21

🔧 Provider Configuration Improvements

• DeepSeek Endpoint Support: Fixed endpoint compatibility issues with DeepSeek and similar providers
• Enhanced Validation: Improved API URL validation and error messages

🐛 Bug Fixes

• Resolved endpoint compatibility problems with various AI providers
• Fixed provider configuration loading issues

1.5.20

✅ What’s New

• Applied few-shot prompting for better accuracy
• Default prompt is cleaner and more intuitive
• Added full support for LM Studio

🧠 Output Example

{
  "output": ["example1", "example2"],
  "reliability": 0.92
}

🚀 Why it matters

• More consistent responses across models
• Easier for users to read and use the result
• LM Studio can now parse outputs without post-processing\

1.5.19

I’ve fixed the API format issue in Gemini, but there is still a critical limitation.

⚠️ Gemini Structured Output Warning

Gemini's response_schema does not enforce hard constraints. Even with a valid JSON schema (e.g., array<string>, number with bounds), Gemini may still return structurally or semantically incorrect outputs:

• Schema type names as literal values (e.g., "output": ["string"])
• Out-of-bound numbers (e.g., "reliability": 1.23 when max is 1)

Root Cause

Gemini treats response_schema as a guideline, not a validator. The model does not strictly adhere to:

• type, items, required
• minimum, maximum for numeric values
• Natural language rules in description