Skip to content
Guide
General

Glossary

Definitions of key terms and concepts used across LlamaParse products and documentation.

Quick reference for terms used throughout the LlamaParse documentation.

An AI agent workflow that uses LlamaParse products (Parse, Extract, Index) as tools to process documents autonomously. Not to be confused with product-specific agents like the extraction agent. Currently in beta.

A Parse tier (10 credits/page) that uses AI agents for complex document understanding, handling tables, charts, and mixed layouts.

The highest Parse tier (45 credits/page) for the most complex documents requiring multi-pass agent processing.

A secret token used to authenticate requests to LlamaParse APIs. Managed at API Keys.

A deployment model where LlamaParse runs within your own cloud infrastructure for data sovereignty and compliance. Available on Enterprise plans — contact sales.

A segment of a parsed document, created during indexing. Chunks are embedded and stored for retrieval in RAG workflows.

A named category with a description used by Classify to sort documents. See Classify.

A service that classifies documents into user-defined categories. Formerly LlamaClassify. See Classify.

A Parse tier (3 credits/page) using LLM-based parsing. Optimized for speed and cost, best for text-heavy documents with minimal structure.

The billing unit for LlamaParse. All operations consume credits based on pages processed, mode, and model. See Pricing.

A destination where indexed data is stored (e.g., a managed vector store or external database like MongoDB or Pinecone).

An origin from which LlamaParse ingests files for indexing (e.g., S3, Google Drive, SharePoint, Confluence).

A numerical vector representation of text, generated during indexing for similarity search in RAG pipelines.

A service that extracts structured data from documents using user-defined schemas. Formerly LlamaExtract. See Extract.

A JSON schema defining the structured fields to extract from documents using Extract. See Schema Design.

The cheapest Parse tier (1 credit/page). Outputs spatial text only (not markdown). Best when you need raw text extraction without AI processing.

A searchable collection of parsed and embedded document chunks, used for RAG retrieval. See LlamaCloud Index.

An asynchronous processing task (parse, extract, classify, or split). Jobs can be polled for status or monitored via webhooks.

A processing mode that uses vision models to understand both text and visual content (images, charts, diagrams).

The process of extracting text from images or scanned documents. Used automatically when needed during parsing.

The top-level account unit in LlamaParse. Billing and members are scoped to an organization. Organizations contain one or more projects.

The base billing unit. For documents, one page equals one document page. For text files, one page equals approximately 600 tokens. For audio, billing is per minute.

The core document parsing service supporting 130+ file formats. Converts documents to markdown, text, or structured output. The parsing product within the LlamaParse platform. See Parse.

An indexing pipeline that defines how documents are ingested, parsed, chunked, embedded, and stored in an index.

A workspace within an organization for organizing files, jobs, indexes, and API keys.

A pattern where relevant document chunks are retrieved from an index and provided as context to an LLM for answering questions.

A spreadsheet-based interface for document processing. Formerly LlamaSheets. Currently in beta.

A service that splits documents into categorized sections based on content. See Split.

A parsing quality level in Parse v2 (Fast, Cost-effective, Agentic, Agentic Plus) that determines the processing approach and credit cost.

A database optimized for storing and searching embeddings. LlamaParse supports managed stores and external options (Pinecone, Qdrant, etc.).

An HTTP callback that notifies your application when a job completes or changes status. See Webhooks.