When using tools with a Chat or AsyncChat class, it would be useful to have a publicly documented way of getting the LLM and tool responses at each step, rather than just when the whole call has finished. This would allow e.g. providing interim feedback to a user for longer running/multi-step calls.
This can currently be done by explicitly using the undocumented Chat._call method - so all that is needed is a public method around that and documentation.
When using tools with a Chat or AsyncChat class, it would be useful to have a publicly documented way of getting the LLM and tool responses at each step, rather than just when the whole call has finished. This would allow e.g. providing interim feedback to a user for longer running/multi-step calls.
This can currently be done by explicitly using the undocumented
Chat._callmethod - so all that is needed is a public method around that and documentation.