Multimodal Documents

1. Plain-Language Definition

Multimodal capability means the system can ingest, understand, and extract text from complex documents that contain more than just plain, selectable text, such as scanned images.

2. Why It Exists

Real-world documents are messy. They contain charts, tables, embedded images, and are often scanned physical pages rather than digitally born PDFs.

3. What Breaks Without It

A pure text extractor will fail completely on a scanned PDF or an image, leaving you with an empty document and an AI that claims 'there is no text'.

4. How StarkLLM Implements It

StarkLLM utilizes Optical Character Recognition (OCR) pipelines and local Vision models to 'read' images and scanned pages, extracting the underlying text and descriptions before embedding.

5. User-Visible Benefit

Supported formats: PDF, DOCX, TXT, Markdown (.md), CSV, JSON, HTML, and images (PNG, JPG) processed through local OCR and vision models.

6. Practical Example

You upload a photo of a restaurant receipt. StarkLLM extracts the line items and prices, allowing you to ask 'How much did we spend on appetizers?'