Add sample_images to enable browseruse to learn how to operate a plat… #2910
Conversation
|
|
There was a problem hiding this comment.
2 issues found across 3 files
Since this is your first cubic review, here's how it works:
- cubic automatically reviews your code and comments on bugs and improvements
- Teach cubic by replying to its comments. cubic learns from your replies and gets better over time
- Ask questions if you need clarification on any suggestion
React with 👍 or 👎 to teach cubic. You can also tag @cubic-dev-ai to give feedback, ask questions, or re-run the review.
fdb468a to
f68c086
Compare
Co-authored-by: cubic-dev-ai[bot] <191113872+cubic-dev-ai[bot]@users.noreply.github.com>
|
Awesome - can you add an example with a dummy image into |
|
|
||
| # Add sample images | ||
| content_parts.extend(self.sample_images) | ||
|
|
There was a problem hiding this comment.
Bug: Vision Mode Fails to Use Sample Images
sample_images are not included in the prompt when vision is enabled but no current screenshots exist. The logic for adding sample_images is incorrectly gated by the presence of self.screenshots, preventing them from serving as intended training examples.
The example supplementation has been completed, and the corresponding pull request is: #2949. Please kindly review and provide any suggestions for further optimization. |
…form.
When setting up task in browseruse, add a parameter for annotating images, enabling browseruse to recognize some interactive components on web pages.
Summary by cubic
Add optional sample_images to Agent and message flow to include annotated example images in the vision prompt, helping the agent recognize interactive UI components. When provided, these images are inserted before screenshots; behavior is unchanged if omitted.