Skip to main content
Corvic works natively across both structured and unstructured data. When you create a data source, you choose which type you’re bringing in, and Corvic routes it through the right ingestion path.
Unstructured data is information that doesn’t fit a tabular model — documents, spreadsheets, images, audio, and video. It’s parsed by Corvic’s extraction pipeline to pull out text, structure, and multimodal content. The supported formats span several categories:

​
Documents

​
Spreadsheets

​
Images

​
Audio

​
Video

​
How each type is processed

You don’t have to wire up ingestion by hand. Based on the data you bring in and the workflow you want, Corvic automatically builds the right processing pipeline for each format — selecting the appropriate engine for that modality:
  • Structured sources (Parquet, CSV, DSV, TSV) become tabular Corvic Tables, ready for transformation, querying, and embedding.
  • Documents are routed through OCR and parsing to extract text, tables, layout, and embedded images — see Multi-modal Knowledge Extraction and Document Digitization.
  • Images, audio, and video are handled by purpose-built extractors for each modality — for example, transcription for audio, frame and scene analysis for video, and vision models for images.
The same source can also be processed in different ways depending on the workflow you’re building. As your requirements change, Corvic adapts the pipeline so the data lands in the form your agents and workflows need.
Need data from the web instead of a file? Use a Web Search source to pull structured information directly from the internet.

Data Sources

Upload files, connect storage, and add web sources.

Multi-modal Knowledge Extraction

How unstructured data is parsed into usable outputs.

Extract Tables

Turn structured sources into clean tables.

Workflows

Process your data sources on the canvas.