Limitations
It is important to understand the limitations with AI in automations, specifically when extracting information from images and PDF documents. These limitations might change in the future; we recommend checking back to this page periodically.
Images
Here are some issues that generative AI models are known to struggle with when processing images:
- Reading an analog measurement device
- Extracting information from low-resolution, rotated, or poor quality images
- Extracting information from images that violate our safety policies
- Identifying people that aren't celebrities
- Counting objects
PDF documents
The following use cases for PDF document extraction might produce unexpected results:
- Counting objects
- Images within a PDF document may not be processed properly; small images relative to the size of the PDF are more likely to have processing issues
About credits and quotas with Gemini in AppSheet Solutions
When you use Gemini in AppSheet Solutions , you consume credits based on the number and complexity of the AI tasks that are run by users of the apps owned by creators in your organization. To learn more about how to get and monitor credits for your team or organization, see Manage credits for Gemini usage .
In addition, in order to ensure fairness and reduce spikes in resource use and availability, AppSheet limits how often an app can call Gemini. To that effect, your usage of AI in automations is restricted by a per-app per-min quota. You can monitor any quota violations for Gemini in AppSheet Solutions using the following:
If you require assistance with quotas, contact AppSheet Support .
