What it does
LlamaIndex describes itself as an applied AI lab working on models for document parsing and extraction. Its main product on the site is LlamaParse, a cloud service that turns complex documents such as PDFs, Office files, scans and handwritten notes into structured, LLM-ready output. LlamaParse covers several steps: parsing 50+ file types including tables and charts, schema-based extraction, splitting documents into sections, classifying them by natural-language rules, and an indexing pipeline for chunking and embedding aimed at retrieval (RAG). The site also lists LlamaIndex Workflows and the LlamaIndex libraries for building document agents. LiteParse is a separate open-source parser that runs locally on PDFs, Office documents and images without cloud calls or LLM tokens, and returns text with bounding boxes. LlamaParse can run in the vendor's cloud (Global or EU) or in your own VPC, and the site cites HIPAA, GDPR and SOC2 compliance.
Capabilities
- LlamaParse: layout-aware parsing for 50+ file types, including tables, charts and handwriting
- Schema-based extraction, document splitting and rule-based classification
- Indexing pipeline with chunking and embedding for retrieval
- LiteParse: open-source local parsing with bounding-box output
- Cloud or VPC deployment with HIPAA, GDPR and SOC2 compliance
When to use it
- Preparing messy PDFs and scans as context for an LLM or RAG pipeline
- Extracting structured fields from invoices, claims or financial documents
- Parsing documents locally without sending them to a cloud service
- Building document agents on the LlamaIndex libraries and Workflows
