Unstructured
An open-source document preprocessing and extraction tool that supports self-hosted deployment and can convert unstructured documents into structured data.
Timeline (last 90 days)
Unstructured is a preprocessing platform that deliberately does not perform field-level structured extraction, converts documents into arrays of elements for use in LLM pipelines, has 40+ connectors, and its commercial platform holds SOC 2 Type II, ISO 27001 and HIPAA certifications as well as FedRAMP High authorization.
Unstructured handles document preparation and parsing, ingesting raw files, PDFs, Office documents, images and scans, splitting them into elements, then chunking, optionally enriching, generating embeddings, and outputting to a destination.
Unstructured supports Databricks Delta Tables (including Volumes) as an output destination, producing a Delta Sync index that can be fed directly to Mosaic AI Vector Search.
Unstructured offers three processing strategies: a fast plain-text mode, a High Res mode combining document understanding with OCR, and a VLM mode that uses vision language models to handle handwriting and complex scans.
All Facts (4)
Unstructured is a preprocessing platform that deliberately does not perform field-level structured extraction, converts documents into arrays of elements for use in LLM pipelines, has 40+ connectors, and its commercial platform holds SOC 2 Type II, ISO 27001 and HIPAA certifications as well as FedRAMP High authorization.
50%UnverifiedUnstructured handles document preparation and parsing, ingesting raw files, PDFs, Office documents, images and scans, splitting them into elements, then chunking, optionally enriching, generating embeddings, and outputting to a destination.
50%UnverifiedUnstructured supports Databricks Delta Tables (including Volumes) as an output destination, producing a Delta Sync index that can be fed directly to Mosaic AI Vector Search.
50%UnverifiedUnstructured offers three processing strategies: a fast plain-text mode, a High Res mode combining document understanding with OCR, and a VLM mode that uses vision language models to handle handwriting and complex scans.
50%