Instructor for PHP home page
Search...
⌘K
Issues
Github
Github
Search...
Navigation
Instructor
Polyglot
HTTP Client
Cookbook
Changelog
Community
Blog
Release Notes
Versions
V1.6.0
V1.5.0
V1.4.2
V1.4.1
V1.4.0
V1.3.0
V1.2.0
V1.1.0
V1.0.0
V1.0.0 RC22
V1.0.0 RC21
V1.0.0 RC20
V1.0.0 RC19
V1.0.0 RC18
V1.0.0 RC17
V1.0.0 RC16
V1.0.0 RC15
V1.0.0 RC14
V1.0.0 RC13
V1.0.0 RC12
V1.0.0 RC11
V1.0.0 RC10
V1.0.0 RC9
V1.0.0 RC8
V1.0.0 RC7
V1.0.0 RC6
V1.0.0 rc5
V1.0.0 rc4
V1.0.0 rc3
V1.0.0 rc2
V1.0.0 rc1
V0.17.12
V0.17.11
V0.17.10
V0.17.9
V0.17.8
V0.17.7
V0.17.6
V0.17.5
V0.17.4
V0.17.3
V0.17.1
V0.17.0
V0.16.0
V0.15.2
V0.15.1
V0.15.0
V0.14.7
V0.14.6
V0.14.5
V0.14.4
V0.14.3
V0.14.2
V0.14.1
V0.14.0
V0.13.0
V0.12.13
V0.12.12
V0.12.11
V0.12.10
V0.12.9
V0.12.8
V0.12.7
V0.12.6
V0.12.5
V0.12.4
V0.12.3
V0.12.2
V0.12.0
V0.8.0
Versions
On this page
Major Features
Chat System
Tool Use Enhancements
Architecture Improvements
Core Enhancements
Bug Fixes
Testing & Quality
Documentation
Release Notes
V1.6.0
Major Features
Chat Architecture
- Multi-participant chat mechanism with conversation coordination, and automatic summaries
ReAct Tool Use
- New ReAct (Reasoning and Acting) driver for advanced tool-based reasoning patterns alongside existing tool calling capabilities
Script System Refactoring
- Script functionality moved from templates to dedicated messages package with enhanced rendering capabilities
Chat System
Multi-Participant Chats
- Support for multiple LLM participants, human participants, and tool-equipped agents in conversation flows
Chat Coordination
- Built-in coordinators including LLM-based, tool-based, and round-robin participant selection
Conversation Summaries
- Automatic chat summarization with configurable buffering and context management
Chat State Management
- Comprehensive state tracking with steps, outcomes, and continuation criteria
Chat Events
- Complete event system for monitoring chat lifecycle, participant selection, and tool usage
Tool Use Enhancements
ReAct Driver
- New reasoning and acting pattern for more sophisticated tool use with decision tracking
Tool Calling Refactoring
- Enhanced tool calling driver with improved error handling and execution flow
Tool Execution Formatting
- Better formatting and display of tool execution results
Multiple Driver Support
- Clean separation between ReAct and tool calling approaches
Architecture Improvements
Messages Package
- Script system relocated from templates to messages package for better separation of concerns
Rendering System
- New arrow-pipe messages renderer and role-based rendering capabilities
Time Management
- Added clock interfaces for better testability and time-dependent operations
Continuation Criteria
- Enhanced continuation logic with token limits, execution time limits, and step limits
Core Enhancements
Structured Output
- Improved deserialization and streaming capabilities for structured data
HTTP Client
- Enhanced HTTP client handling with better error management
Documentation Testing
- Major Doctest architecture refactoring with improved code block extraction and validation
Cohere Integration
- Fixed max_tokens parameter handling for Cohere API
Bug Fixes
Property Info Adapter
- Enhanced compatibility with PropertyInfo v7 for better type introspection
Test Infrastructure
- Added comprehensive fake inference drivers and mock HTTP testing capabilities
Memory Management
- Improved memory handling in streaming and sequence processing
Testing & Quality
Test Coverage
- New test suites for Chat, ToolUse, and Doctest functionality
Mock Infrastructure
- Enhanced testing infrastructure with fake drivers and HTTP mocking
Static Analysis
- Updated PHPStan and Psalm configurations for better code quality
Documentation
Package Documentation
- Updated CHEATSHEET.md files across packages with comprehensive usage examples
Chat Documentation
- New dedicated documentation for chat and tool use patterns
Architecture Guides
- Enhanced OVERVIEW.md files explaining package internals and design decisions
Full Changelog
:
v1.5.0…v1.6.0
Versions
V1.5.0
Assistant
Responses are generated using AI and may contain mistakes.