Getting started
Supported file formats
Learn which files Polaris can read and how to prepare them for better answers.
Polaris can use documents and files as knowledge for your chatbots. Answer quality depends on content that is clear, current and readable.
Accepted formats
Polaris accepts:
- TXT
- Markdown (
.md,.markdown) - HTML
- CSV
- XLSX
- PPTX
- JSON
- XML
- RTF
- DOCX
- PNG, JPG, JPEG and WebP images
PDFs
A PDF can be processed in different ways:
- Textual PDF: contains selectable text and can be processed directly.
- Scanned PDF: needs OCR to read text inside scanned pages.
- Visual PDF: can use Vision to describe images, screenshots, diagrams or charts when the plan includes it.
If your plan does not include OCR or Vision, Polaris can still upload the file and use the available textual content. Scanned or visual parts may stay out of the knowledge base.
JSON and XML
Polaris can convert JSON and XML into structured text. When it detects common sensitive keys, such as tokens, secrets or passwords, it redacts those values before using the content as knowledge.
Best practices
- Use descriptive titles for every file.
- Upload textual PDFs when possible.
- Split very long documents by topic.
- Avoid duplicate versions of the same document.
- Make sure tables and spreadsheets have clear headers.
- For images or screenshots, use sharp files with readable text.