ChatDBG is now safe by default and won't let GPT call possibly harmful debugger commands. The `--unsafe` flag can be passed to allow any function call.
* Set the default model to gpt-4o. * Expose the `--instructions` command line option to supply different instructions than the (possibly mode-specific) default.
* Minor fix over v0.6. * Improved support for GDB. * Consistent formatting and outputs across all debuggers. * Upgrade minimum to Python 3.10 -- IPython requires >=3.10 by nicovank in https://github.com/plasma-umass/ChatDBG/pull/41
* Improved support for GDB. * Consistent formatting and outputs across all debuggers. * Upgrade minimum to Python 3.10 -- IPython requires >=3.10 by nicovank in https://github.com/plasma-umass/ChatDBG/pull/41
What's Changed * Many changes (see v0.4, which did not get pushed to PyPi due to an oversight). * Ignore any warnings LiteLLM might emit on import by nicovank in https://github.com/plasma-umass/ChatDBG/pull/38