- Added a check to prevent empty strings from being added to the chat
- This ensures no messages with an empty 'content' field are inserted
- Enhances stability and error handling in chat functionality
- Added vscode_services.py to enable direct access to raw data from VSCode.
- Updated ide_services/services.py to use the new rpc_call function.
- Added vscode_services.py for specific VSCode service functionalities.
- Refactored chunks_call to utilize a list of tool calls
- Updated to_dict_content_and_call to handle multiple tool calls
- Enhanced chat_tools to display function call details
- Added new file tools_call.py containing functions and decorators
- Implements tool and function schemas for use in the application
- Includes a function to prompt the user to confirm function calls
.
- Added memory module for LLM API
- Implemented ChatMemory and FixSizeChatMemory classes
- Included methods for appending, requesting, and responding to memory
- Added new decorators for LLM functions, parameters, and confirmations.
- Introduced confirmation dialogs for calls and edits within LLM interactions.
- Streamlined integration of LLM tool schemas and response handling.
- Refactored language_prompt to handle Chinese characters
- Updated model_token_limit_error message formatting
- Modified function signature for chat_completion_stream
- Updated commit/command.yml to use "en" language code
- Updated commit/commit.py to handle model's maximum context length error
- Updated commit/zh/command.yml to use "zh" language code
- Updated libs/llm_api/openai.py to ignore tokens limit error
Closesdevchat-ai/devchat#189
- Make submit and cancel button labels optional in form and widgets
- Refactor chatmark header construction for conditional button labels
- Implement assertion for Radio widget's default_selected option
- Add trailing commas to button name parameters in widgets
- Reformat constructor definitions for consistency
- Clean up unnecessary whitespace in widget render logic
- Introduced the default_selected parameter in the Radio widget
- Updated Radio widget to mark default selected option visually
- Removed TODO comment for the default_selected implementation
- Form and TextEditor classes now accept submit_button_name and cancel_button_name
- Widget's chatmark header updated with customizable submit and cancel labels
- Adjusted initialization of Checkbox and Radio classes with button label parameters
- Simplify and clarify user prompts and output messages
- Remove unnecessary print statements and text
- Handle commit abort and errors more gracefully
- Improve error logging in OpenAI chat stream handler
- Update the signature of the chat_completion_stream function
- Remove unnecessary line breaks and indents
- Improve code readability and maintainability
- Reformatted multiline strings for clarity in commit.py
- Applied consistent quotation marks for string literals in commit.py
- Introduced newlines between rpc_call decorator functions in services.py
- Add stream_out parameter to chat_completion_no_stream function
- Print delta content when stream_out is True
- Ensure seamless streaming of chat completion response
- Removed unused functions `gpt_file_summary`, `gpt_file_group`, `get_file_summaries`, `get_file_summaries_and_groups`, `get_marked_files`, `generate_commit_message_base_file_summaries`
- Updated function `get_marked_files` to accept only `modified_files` and `staged_files`
- Updated function `generate_commit_message_base_diff` to provide a more descriptive start message for the commit workflow
