Add model selection within utterances. #169
Conversation
Enhances talon commands to allow specifying a model directly within the utterance. For example: 'four o mini explain this' will use gpt-4o-mini model. - Updated model.talon-list to include four o and four o mini models - Modified send_request and supporting functions to accept and use model parameter - Updated all Talon command files to pass the model parameter - Fixed parameter ordering to ensure optional params come after required ones - Deprecated openai_model setting in favor of model_default_model
for more information, see https://pre-commit.ci
|
Thanks for this PR, may be a couple days until I properly review/test locally just fyi; busy next few days; feel free to ping if I forget |
|
|
||
| def gpt_generate_shell(text_to_process: str) -> str: | ||
| def gpt_generate_shell(text_to_process: str, model: str) -> str: |
There was a problem hiding this comment.
Totally good w me either way but curious on the rationale for passing in model to each function instead of just calling the setting for it since it is global state anyways. I suppose this is better if we ever set up test w/ mocking?
There was a problem hiding this comment.
I have to pass in the model in order to support models provided within the utterance. With this change, anywhere that you can say "model" you can now also name a specific model to use. The global state will always be the same, but this parameter passes along the list value from the utterance.
|
I've addressed all of your comments. |
|
Looks good, thanks! |
Enhances talon commands to allow specifying a model directly within the utterance. For example: 'four o mini explain this' will use gpt-4o-mini model.
Note: I tested every command here. I fixed some issues with the "model find" command. I noticed that the Cursorless commands are broken, and there are some bugs in the beta "blend" commands. I compared to the baseline and these issues were already present. I can file separate bugs for those, since they seem to run a bit deeper.