Any document. Structured knowledge.
Purple8 DocIntel is a stateless microservice that turns 70+ enterprise document formats — PDFs, CAD files, BIM models, spreadsheets, and images — into structured entities and relationships committed directly into your Purple8 knowledge graph.
Format coverage
70+ formats. One API.
How it works
Parse → Extract → Commit
Seven-stage pipeline, fully async. Drop a file in — get knowledge graph nodes out.
Two-pass extraction
GLiNER NER for entities, LLM pass for relationships. Domain-tuned profiles for AEC, legal, finance, scientific.
Connector suite
Pull from SharePoint, Confluence, S3, or receive webhooks. Syncs continuously.
P8G-aware chunking
Derives optimal chunking strategy, embedding model suggestion, and doc type confidence — forwarded to Purple8 for indexing.
Zero data egress
Self-hosted OCR engine available. Nothing leaves your network unless you configure an external LLM provider.
Parallel extraction
Up to 8 concurrent LLM windows per document. Process large corpora fast.
Custom profiles
Define your own extraction ontology — entity types, relationship types, chunking rules — per document domain.
Add document intelligence to your stack.
Works standalone or alongside Purple8 Hyper Graph. Self-hosted, air-gap ready.