- Removed obsolete file upload/download functions and related constants from the Assistant struct, streamlining the codebase.
- Updated the connector type in load tests to use "gpt-3_5-turbo" for improved compatibility with current models.
- Introduced VisionCapableModels to support a range of models with vision capabilities, enhancing AI functionality.
- Enhanced the File struct to include additional fields for description, URL, and document IDs, improving data management for file handling.
These changes improve the overall structure and maintainability of the Neo API, paving the way for future enhancements in assistant capabilities.
- Removed unused assistant-related functions and types from the Neo API, including local chat and file handling, to streamline the codebase.
- Refactored the assistant initialization process, consolidating the loading and configuration of assistants, including Vision and Connector settings.
- Updated the DSL struct to enhance clarity and maintainability, ensuring better configuration management for assistants.
- Improved error handling and modularity in the assistant loading process, paving the way for future enhancements in AI functionalities.
- Updated the SetRAG function to include RAG settings, allowing for more detailed configuration during assistant initialization.
- Refactored the LoadBuiltIn function to streamline the saving process of assistants, improving error handling and code clarity.
- Modified loadPrompts and loadScript functions to return timestamps, enabling tracking of creation and update times for assistants.
- Enhanced the Assistant struct with created_at and updated_at fields, improving data management and traceability.
- Overall improvements to the assistant loading process, ensuring better organization and maintainability of the codebase.
- Introduced new methods for initializing RAG and store, improving modularity and clarity in the Load function.
- Refactored the assistant initialization process to streamline the loading of built-in assistants and set RAG configurations.
- Moved the defaultAssistant method from neo.go to load.go for better organization and accessibility.
- Enhanced error handling in store initialization to support multiple connector types, including Redis and Mongo.
- Updated the assistant struct to include RAG settings, improving the overall assistant management capabilities.
- Added a new endpoint to retrieve all assistant tags, improving data accessibility for clients.
- Updated the assistant list handling to support filtering by built-in status and assistant ID, enhancing the filtering capabilities.
- Introduced a method to load built-in assistants, streamlining the assistant initialization process.
- Enhanced the Assistant struct with new fields for path, built-in status, and sorting, improving data organization.
- Implemented validation and cloning methods for the Assistant struct, ensuring data integrity and ease of use.
- Updated tests to cover new functionalities, including validation, cloning, and tag retrieval, ensuring robust functionality across the assistant management operations.
- Removed Weaviate store implementation, streamlining the codebase and focusing on Mongo and Redis backends.
- Introduced GetAssistant method in both Mongo and Redis stores to retrieve a single assistant by ID, improving data access capabilities.
- Updated LoadStore function to utilize the new storage retrieval logic, enhancing the assistant loading process.
- Enhanced the Assistant struct to include a Script field for better management of assistant scripts.
- Improved tests to cover the new GetAssistant functionality, ensuring robust error handling and data retrieval across different scenarios.
- Updated the message handling logic to improve error reporting and content management, ensuring more robust communication with clients.
- Introduced a new ReadBase64 method in both Local and OpenAI assistant implementations to read files and return their base64 encoded content, enhancing file handling capabilities.
- Removed the deprecated JSON message handling code, streamlining the message processing structure.
- Refactored message struct to include new fields and methods for better data management and response handling.
- Improved the chat functionality to handle attachments and user messages more effectively, ensuring a smoother user experience.
- Updated SaveHistory method across Mongo, Redis, and Weaviate to include a context parameter, improving flexibility in message storage.
- Modified the Assistant struct in types.go to add new fields: Type, Avatar, and Flows, enhancing the representation of assistant attributes.
- Adjusted the Xun implementation to support the new context handling in SaveHistory, ensuring messages can now include contextual information.
- Updated tests to reflect changes in message structure and ensure proper functionality with the new context parameter.
- Introduced a new endpoint for downloading files, enhancing the API's capabilities.
- Implemented the handleDownload method to manage download requests, including validation for required parameters (sid and file_id).
- Updated the Download method in the assistant interface to retrieve files based on file_id, ensuring proper error handling and response structure.
- Enhanced file upload logic to include file extensions in generated IDs for better file management.
- Added necessary response headers for file downloads, improving user experience during file retrieval.
