LLM cleanup #5462
LLM cleanup #5462
Conversation
| // any custom prompt from last user message | ||
| ...(input.user.system ? [input.user.system] : []), | ||
| ] | ||
| const system = [first, rest.join("\n")].filter((x) => x) |
There was a problem hiding this comment.
should system should only be multiple parts if header is present?
| }) | ||
|
|
||
| // TODO: we may wanna rename this tool so it works better on other shells | ||
| const getShell = lazy(() => { |
There was a problem hiding this comment.
idk how this got there but I think we can rm this
|
/review |
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
| input.small ? mergeDeep(ProviderTransform.smallOptions(input.model)) : mergeDeep({}), | ||
| mergeDeep(input.model.options), | ||
| mergeDeep(input.agent.options), | ||
| ), |
There was a problem hiding this comment.
Potential bug: pipe in remeda expects the first argument to be the initial value, not a transformation function. Currently mergeDeep(ProviderTransform.options(...)) is passed as the first argument, but mergeDeep returns a curried function. This should be:
| ), | |
| options: pipe( | |
| {}, | |
| mergeDeep(ProviderTransform.options(input.model, input.sessionID)), | |
| input.small ? mergeDeep(ProviderTransform.smallOptions(input.model)) : mergeDeep({}), | |
| mergeDeep(input.model.options), | |
| mergeDeep(input.agent.options), | |
| ), |
This matches the pattern used elsewhere in the codebase (e.g., the original prompt.ts code).
|
Code review completed. Found one issue - see inline comment on The rest of the code looks good and follows the style guide. The refactoring to centralize LLM operations is clean and well-structured. |
No description provided.