LlamaIndex

Freemium

基于 LlamaParse 的文档智能平台，支持解析、提取与工作流

什么是 LlamaIndex？

LlamaIndex is a document intelligence platform centered on LlamaParse, an agentic OCR and parsing product for complex documents. It is designed to turn PDFs, Office files, images, tables, charts, handwritten notes, and other unstructured inputs into structured, LLM-ready outputs.

The site describes a broader workflow stack around parsing, extraction, splitting, classification, indexing, and document-agent building. It also offers LiteParse, an open-source local parser for teams that want offline processing with bounding boxes and structured output.

LlamaIndex 能做什么？

Agentic document understanding

Turns complex documents into LLM-ready outputs through semantic understanding, with specialized handling for text, charts, tables, and other content types.

Auto-correction loops

Uses recursive checks to detect and fix errors automatically, which is aimed at improving pass-through on messy scans and multi-modal documents.

Structured extraction

Extracts structured data from unstructured content using schema-based, LLM-powered extraction agents without requiring training.

Flexible output formats

Supports parsing into multiple formats including Markdown, plain text, JSON, XLSX, HTML, tables, and annotated PDF, with output options that fit different downstream workflows.

Local parsing option

Provides local-only open-source parsing in LiteParse for PDFs, Office docs, and images, with bounding-box output and no cloud or LLM-token usage.

Document workflow platform

Includes parsing, extraction, splitting, classification, indexing, and workflow building for document-heavy teams.

使用场景

“Finance document analysis”

Parse contracts, filings, and research documents into structured outputs that analysts can review or feed into downstream workflows.

“KYC and AML processing”

Extract, verify, and route identity and transaction documents for KYC and AML workflows.

“Audit and back-office automation”

Automate handling of invoices, claims, and audit materials where consistent extraction and traceability matter.

“Multi-step document agents”

Build document agents that read complex sources and take actions such as routing, validation, logging, and notification.

常见问题

Is LlamaParse open source?

LlamaParse is a commercial product, not open source. The site says it includes 10,000 free credits per month for new users, while LlamaIndex and Workflows are the open-source projects in the ecosystem.

How does credit usage work for parsing?

The pricing page says parsing or extraction costs depend on the mode and options selected. Basic parsing can be as low as 1 credit per page, while layout-aware agentic parsing with LLMs or VLMs costs more for higher accuracy.

Can LlamaParse be deployed on premises?

Yes. The pricing page describes SaaS hosting on a secure cloud tenant, with an option for enterprise deployment in private VPCs across cloud providers.

What outputs and file types does it support?

The pricing page lists output options including Markdown, plain text, JSON, XLSX, HTML, tables, and annotated PDF, and it supports 130+ file formats.

Who is LlamaParse for?

The site positions LlamaParse for teams that need document parsing, extraction, indexing, and retrieval over complex documents, especially where layout, tables, charts, or handwritten notes matter.