* feat: add InvokeLLMWithStructuredOutput functionality
- Introduced a new method InvokeLLMWithStructuredOutput to the BackwardsInvocation interface for handling structured output requests.
- Added corresponding request and response types to support structured output.
- Implemented the method in both RealBackwardsInvocation and MockedDifyInvocation for testing purposes.
- Updated permission handling and task execution for the new structured output invocation type.
This enhancement allows for more flexible and detailed responses from the LLM, improving the overall functionality of the invocation system.
* refactor: enhance LLMResultChunkWithStructuredOutput structure
- Updated the LLMResultChunkWithStructuredOutput type to include additional fields: Model, SystemFingerprint, and Delta.
- Added comments to clarify the reasoning behind the structure and the use of type embedding for JSON marshaling.
This change improves the clarity and functionality of the LLMResultChunkWithStructuredOutput type, ensuring proper JSON serialization.
* refactor: streamline LLMResultChunk construction in InvokeLLMWithStructuredOutput
- Simplified the construction of LLMResultChunk and LLMResultChunkWithStructuredOutput by removing unnecessary type embedding.
- Enhanced readability and maintainability of the code while preserving functionality.
This change contributes to cleaner code and improved clarity in the handling of structured output responses.
* feat: enhance integration testing for specific plugins
- Added integration tests for the official agent, including test data for agent strategy invocation.
- Introduced JSON schema generation and validation utilities to ensure proper request formatting.
- Enhanced mock invocation handling in the plugin manager to support tool parameters.
- Added new test utilities for simulating OpenAI server responses and managing plugin runtime.
* fix: update RunOnce function to return response stream and enhance integration test
- Modified the RunOnce function to return a response stream instead of an error, allowing for better handling of streamed responses.
- Updated the integration test for the official agent to read from the response stream, ensuring proper validation of the agent strategy invocation.
- Introduced FetchApp method in BackwardsInvocation interface to retrieve application data.
- Implemented FetchAppRequest type for request payload with required AppId field.
- Updated RealBackwardsInvocation to handle FetchApp requests and return application data.
- Added mock implementation for FetchApp in the testing suite.
- Enhanced task handling in plugin daemon to support FetchApp invocation.
Yeuoly