v2.1.2

v2.1.2 Release Notes:

Bug Fixes

• Resolved UI-level token limit calculation bug
• Maintains full functional compatibility with v2.1.1

Previous Version Features (v2.1.1)

For more details on the integration, check out the discussion on the Home Assistant Community forum

🔄 Major Architectural Changes

• Complete refactoring of token handling mechanism
• Elimination of custom token calculation approach
• Direct max_tokens parameter passing to LLM APIs

🎯 Key Technical Improvements

• Enhanced cross-provider compatibility
• Expanded support for large-context language models
• Robust and predictable token limit management
• Significant codebase simplification
• DeepSeek provider full integration

Provider Updates

DeepSeek — NEW Integration

DeepSeek is a cutting-edge AI provider specializing in advanced language models optimized for both conversational and reasoning tasks. This integration brings:

• High-performance model inference
• Cost-effective API endpoints
• Enterprise-grade reliability
• Flexible deployment options

Full Changelog: v2.1.1...v2.1.2

v2.1.1

v2.1.1 - Token Handling & DeepSeek Provider Integration

For more details on the integration, check out the discussion on the Home Assistant Community forum

🔄 Major Architectural Changes

• Complete refactoring of token handling mechanism
• Elimination of custom token calculation approach
• Direct max_tokens parameter passing to LLM APIs

🎯 Key Technical Improvements

• Enhanced cross-provider compatibility
• Expanded support for large-context language models
• Robust and predictable token limit management
• Significant codebase simplification
• DeepSeek provider full integration

🙏 Community Acknowledgments

Heartfelt gratitude to @estiens for identifying token handling complexities and providing comprehensive feedback that drove these critical improvements (#1).

Provider Updates

DeepSeek — NEW Integration

DeepSeek is a cutting-edge AI provider specializing in advanced language models optimized for both conversational and reasoning tasks. This integration brings:

• High-performance model inference
• Cost-effective API endpoints
• Enterprise-grade reliability
• Flexible deployment options

New Model Support

DeepSeek

• deepseek-chat (DeepSeek-V3) — NEW Model
  A state-of-the-art conversational AI model designed for natural, context-aware dialogues. Features include:

  • Enhanced context retention
  • Multi-turn conversation support
  • Emotion-aware responses
  • Multi-language capabilities

• deepseek-reasoner (DeepSeek-R1) — NEW Model
  A specialized reasoning engine optimized for:

  • Complex problem-solving
  • Logical inference tasks
  • Structured data analysis
  • Multi-step reasoning workflows

Full Changelog: v2.1.0...v2.1.1

v2.1.0

v2.1.0 Release Notes:

Fixed incorrect version display in the integration instance

Full Changelog: v2.0.9...v2.1.0

v2.0.9

Version increment

This is a routine version update without substantial modifications to the codebase.
Version bumped to v2.0.9 for maintenance purposes.

Full Changelog: v2.0.8...v2.0.9

v2.0.8

fix: Display only last Q&A in sensor state to prevent data truncation

• Show only the latest question and answer in sensor state
• Keep full conversation history in attributes
• Fix truncation issues in Home Assistant UI
• Maintain backwards compatibility
• No configuration changes required

Full Changelog: v2.0.7-beta...v2.0.8

v2.0.7-beta

Optimization and Performance Improvements

🚀 Key Enhancements

• JSON Handling: Optimized JSON file history management for improved efficiency
• Memory Management: Added comprehensive memory usage validation during initialization
• Disk Space Monitoring: Implemented free disk space verification to prevent potential storage-related issues
• Performance Optimization: Streamlined code structure, enhanced parallel request processing, and improved overall system responsiveness
• Token Estimation: Refined heuristic token counting method for more accurate conversation tracking

🔍 Technical Details

• Enhanced error handling and resource management
• Improved system resilience through proactive checks
• More precise token usage estimation

🛠 Recommended Update

Recommended for all users seeking improved system stability and performance.

Full Changelog: v2.0.5-beta...v2.0.7-beta

❌ v2.0.6-broken

❌ v2.0.6 — broken

v2.0.5-beta

🔄 Changes v2.0.5

🆕 New Features

• Conversation history now stored in files instead of Home Assistant states
• Added configurable history size limit (default: 100 entries)
• Added automatic cleanup of old history entries

💪 Improvements

• Reduced memory usage by moving history to file storage
• Better handling of large conversation histories
• Fixed async history updates
• Updated translations
• Fixed minor bugs
• Updated documentation

📝 Notes

• History is now stored in: .storage/ha_text_ai_history/
• No configuration changes required

Full Changelog: v2.0.4-beta...v2.0.5-beta

v2.0.4-beta

v2.0.4-beta Release Notes:

feat(localization): Expand multilingual support

• Added translations for:

  • Chinese (zh)
  • Serbian (sr)
  • Italian (it)
  • Hindi (hi)
  • Spanish (es)

• Fixed minor bugs

• Improved language coverage

Full Changelog: v2.0.3-beta...v2.0.4-beta

v2.0.3-beta

v2.0.3-beta Release Notes:

• Fixed sensor naming, now sensors are created correctly according to the Sensor Naming Convention.
• Added additional endpoint availability checks during operation.
• Improved Anthropic integration stability.
• Implemented more accurate token counting (heuristic in any case, but calculates more precisely).
• Enhanced max_tokens handling.
• Improved error descriptions and hints.
• Expanded sensor attribute list.
• Other bugfixes.

Important: If you previously created automations or sensors using special characters, please recreate them. Unfortunately, it's not possible to automatically reimport them with correct names.
ⓘ Sensors created according to the documentation description will continue to work correctly.

Full Changelog: v2.0.2-beta...v2.0.3-beta