LlamaIndex Document Parsing Protocol
Draft standard (140 rules in v0.1) for how document parsers handle tables, forms and diagrams so agents get consistent structured input.
LlamaIndex’s Document Parsing Protocol (DPP) is a proposed “answer key” for document parsing: 140 rules in working draft 0.1 covering text, tables, forms, charts and diagrams, each with example pages and the output expected, so agents get consistent structured input regardless of parser. The rules grew out of what annotators followed when labeling pages for LlamaIndex’s ParseBench benchmark. Examples: an empty form field becomes [blank], charts become a table of printed values, and unreadable text is [illegible]. LlamaIndex invites issues and pull requests on GitHub (run-llama/document_parsing_protocol). Consistent parsing matters for RAG quality and for prompt-injection defenses that need to know what text a model actually sees.
Timeline
Every dated event in the wiki that involves LlamaIndex Document Parsing Protocol, newest first.
- LlamaIndex publishes the Document Parsing Protocol draft
Working draft 0.1: 140 rules for consistent document parsing for agents.
Connections
LlamaIndex Document Parsing Protocol and everything one link away. Drag to move, scroll to zoom, click a node to open it.Open in the full graph →
What links here
Company LlamaIndex
Sources
- LlamaIndex, “Introducing the Document Parsing Protocol”, Oct 28 2026