:sparkles: feat: Add 'explain prompt' functionality to AI tools

C-Loftus / talon-ai-tools

Query LLMs and AI tools with voice commands

http://colton.place/talon-ai-tools/

MIT License

46 stars 17 forks source link

:sparkles: feat: Add 'explain prompt' functionality to AI tools #159

Closed jaresty closed 1 month ago

jaresty commented 1 month ago

Introduced new prompt 'explain prompt' to analyze and explain how a given prompt was understood to generate a response.
Replaced thread_to_string with chats_to_string to improve clarity in handling chat threads.
Added commands gptRequest and gptExchange to support retrieval of previous requests and exchanges.
Updated send_request to store the last request and response in GPTState for better context management.
Removed redundant GPTState.last_response assignment in gpt_query.

These changes enhance the ability to run 'explain prompt' and understand prompt-response mechanics.

jaresty commented 1 month ago

You can explain your last exchange with 'model explain prompt exchange'

jaresty commented 1 month ago

For what it's worth, the idea for this came from page four of this draft book from Michael Feathers (the author of Working Effectively With Legacy Code) https://www.linkedin.com/posts/michaelfeathers_use-waywords-ugcPost-7227367323878187008-RNZL?utm_source=share&utm_medium=member_desktop

C-Loftus commented 1 month ago

Sorry but I don't think I am going to merge this. I think it is a bit too specific. If the user wants to do this, I think it is better for them to just copy the info from the debug output and run a model command on it. I don't really like merging destinations that require storing / modifying additional state unless there is strong need for it.

We can reopen if there is need down the line.

jaresty commented 1 month ago

Can we discuss? I think there is value here beyond simply this use case. This is helpful for a variety of reasons and it's not so easy to do manually. Missing this tempts me to run off of a fork, since I can't implement it myself otherwise.