Getting started
Vision for images and charts
How Polaris describes images, screenshots, diagrams and charts so the chatbot can answer.
Vision lets Polaris turn visual content into textual descriptions that the chatbot can use for answers.
What it can analyze
Vision can help with:
- images;
- screenshots;
- diagrams;
- charts;
- visual tables;
- visual PDF pages.
Plans
- Starter: Vision not included.
- Pro: Vision not included.
- Business: Vision included with monthly limits.
- Enterprise: Vision with custom limits.
Availability can also depend on workspace configuration.
Multimodal citations
When an answer uses visual evidence, Polaris can show a citation with more context.
Example:
Annual report.pdf · Page 3 · Chart · Vision
This means the answer used a visual description of a chart on page 3.
Current limitations
Vision is not image search. It does not yet let you upload an image as a question to search for similar images.
There are also intentional limits:
- no thumbnails in citations;
- no advanced visual preview;
- no separate visual index;
- visual content is converted to text so the chatbot can use it.
Best practices
- Use sharp images.
- Avoid blurry or very small screenshots.
- If a chart has important values, make sure they are readable.
- For PDFs with many visual pages, use shorter files when possible.