Changelog
View SourceAll notable changes to this project will be documented in this file.
[0.2.0] - 2025-01-28
Added
- Added
configure_http_client/3
function to provide a more flexible way to configure HTTP clients - Added type specifications for all public functions
Changed
- Breaking: Removed
adapter_config
fromstart_chat_with!/2
options. Configuration should now be done throughconfigure_adapter/2
- Improved error handling for HTTP client validation
- Enhanced error handling for better insight into validation failures
- Restructured the codebase for better maintainability
Fixed
- Fixed Dialyzer warnings across the codebase
- Fixed type specifications to be more accurate and complete
[0.1.3] - 2025-01-22
Added
- Added HTTP options configuration support for OpenAI adapter
- Added configurable timeouts for HTTP requests
Changed
- Improved HTTP client configuration handling
- Enhanced documentation and examples
[0.1.2] - 2025-01-22
Added
- Added temperature configuration option for OpenAI adapter
- Added proper HTTP timeout configurations
Fixed
- Fixed module loading issues with
Code.ensure_loaded?/1
- Fixed custom type handling in JSON Schema generation
Changed
- Improved error handling in custom type implementations
- Enhanced documentation for custom types
[0.1.1] - 2025-01-21
Added
- Added support for custom
llm_description/0
callback - Added new parser implementation for better field documentation validation
- Added comprehensive documentation and examples
- Added evaluation modules for testing and demonstration
Changed
- Simplified the parser implementation
- Improved documentation parsing and validation
- Enhanced error messages for missing field documentation
Fixed
- Fixed documentation validation for nested fields
- Fixed schema field parsing
All these changes follow semantic versioning principles, where:
- 0.2.0 includes breaking changes to the API
- 0.1.1 through 0.1.3 add new features and fixes in a backward-compatible way
[0.1.0] - 2025-01-20
Added
- Initial release of Mentor.
- Provides a high-level API to generate structured outputs based on various schemas, including raw maps, structs, and
Ecto
schemas. - Implements the
Mentor.LLM.Adapter
behaviour for integrating with different Large Language Models (LLMs). - Includes an adapter for OpenAI's language models.
- Supports defining and validating data structures using the
Mentor.Schema
protocol. - Offers basic debugging capabilities.
- Provides a high-level API to generate structured outputs based on various schemas, including raw maps, structs, and
Note: This changelog follows the principles of Keep a Changelog.