[Built-in] Add Dos and Don'ts for AI implementation#893
Conversation
|
It'd be nice if the agent fetched this but i suspect they would rarely. (i didn't try though) Alternatively we could add evals to see if the dos and donts are followed.. in which case this reference is effective enough. |
|
In my experience, the agent was fetching the resource, especially when hinted to follow best practices. So based on this, I'd say let's merge and revisit if we get negative feedback?! |
| - **Aborting Tasks**: Use `AbortController` to allow users to stop long-running generations. Pass the `signal` to `prompt()` or `promptStreaming()`, not to `LanguageModel.create()`. | ||
| - **Security**: Use Permission Policies to control access in iframes: `<iframe src="..." allow="language-model"></iframe>`. | ||
| - **Design**: Review the [People + AI Guidebook](https://pair.withgoogle.com/guidebook/) to ensure responsible AI implementation. | ||
| - **Dos and Don'ts**: See this collection of [technical and UX considerations to help you avoid common pitfalls](https://developer.chrome.com/docs/ai/built-in-ai-dos-donts). |
There was a problem hiding this comment.
Similar to @paulirish's comment, my preference would be to inline any important and relevant points directly into this guide.
If that's not possible for whatever reason, we could link to the agent-optimized version of the doc: https://developer.chrome.com/docs/ai/built-in-ai-dos-donts.md.txt
3 participants
No description provided.