Multi-source document ingestion, multimodal extraction, visual schema & Mermaid diagram transcription, token cost tracking, and Open Knowledge Format (OKF v0.2) packaging for the Quatrain ecosystem.
Features
OKF v0.2 Compliant: Native support for the # --- Open Knowledge Format v0.2 --- specification, including description (1-sentence summary), status, generated: { by, at, tokens }, and sources: [...].
Text & Multimodal Ingestion: Extracts digital text layers via pdf-parse and falls back automatically to multimodal Gemini vision for scanned or image-dense documents.
Visual Synthesis & Mermaid Transcription: Instructs Gemini to transcribe process flows, decision trees, and cycles into mermaid code blocks, and matrices/comparison charts into Markdown tables.
Accurate Token & Cost Tracking: Computes exact input/output/thinking token usage and USD costs per document.
Resilient Dedup Cache: SHA-256 hash tracking with persistent token accounting and automatic sync from existing Git knowledge repos.
Installation
yarnadd@quatrain/okf-ingest # or bunadd@quatrain/okf-ingest