cactus 0.2.7

Framework for running AI models on mobile devices

## 0.2.7

• Agents capable of phone tool use 8 Gemma3 270m support
• Separare model download as a separate function

0.2.4

• Bugfix
• Openai Compatible Messaging

0.2.0

• Changed visionUrl/visionFilename to mmprojUrl/mmprojFilename in CactusVLM.init()
• Updated to match actual implementation and align with other platforms

0.1.4

• added telemetry

0.1.3

• bug fix
• simplified APIs

0.1.2

• NEW: Conversation Management API - Optimized stateful chat with KV cache performance
  • Added generateResponse() method for simple text responses
  • Added continueConversation() method with detailed performance metrics
  • Added clearConversation() method to reset conversation state
  • Added isConversationActive() method to check conversation status
  • Added ConversationResult class with timing data (TTFT, total time, tokens/sec)
• PERFORMANCE: KV Cache Optimization - Consistent ~70ms TTFT across conversation turns
  • Eliminates conversation length impact on response time
  • 3-6x performance improvement for multi-turn conversations
  • Automatic conversation state management with token appending
• ENHANCED: Dual-Mode Architecture in Flutter service
  • Automatic switching between optimized text and multimodal modes
  • Real-time performance tracking with ValueNotifiers
  • Enhanced state management and cleanup
• IMPROVED: FFI Interface - Complete C FFI bindings for conversation management
  • Added cactus_generate_response_c, cactus_continue_conversation_c
  • Added cactus_clear_conversation_c, cactus_is_conversation_active_c
  • Added cactus_conversation_result_c_t struct with performance metrics
  • Proper memory management with cactus_free_conversation_result_members_c

0.1.0

• Major update with comprehensive API improvements
• Added multimodal completion support with completionSmart method
• Enhanced chat templating with advanced formatting options
• Improved conversation history management and mode switching
• Added tool calling framework for function execution
• Enhanced grammar generation system for structured output
• Added comprehensive error handling and exception types
• Updated example app with modern conversation flow
• Improved memory management and resource cleanup
• Added clear conversation functionality
• Fixed various linter warnings and code quality issues

0.0.3

• Fixed an issue with loading the native iOS library when using XCFrameworks.

0.0.2

• Fixed an issue with loading the native iOS library when using XCFrameworks.

0.0.1

• Initial release