Coverage Labelling
Transparent thin/contested/low coverage indicators
Labels indicating when answers rest on thin, contested, or insufficient evidence.
Read the full card →The knowledge hub
WikiCrux
A receipted knowledge hub for agents and teams: organisational knowledge that is versioned, auditable, and agent-ready, with every citation deref-able back to its source.
How it's built
A read-plane over CoreCrux; per-tenant corpora; deref-able wx:// citations. [Fable: expand.]
Why teams use it
Institutional knowledge agents can query. Every answer traceable. [Fable: expand.]
Corpus & quality capabilities
Coverage-Driven Sourcing
Pathway from low-coverage to ingestion requests
Automatically identifies coverage gaps and generates targeted ingestion requests.
Read the full card →
Ingestion Upvoting
Structured demand signal for corpus prioritization
Users vote on which sources should be prioritized for ingestion into the corpus.
Read the full card →
Data Quality Pipeline
Eight quality techniques with corpus-class gating
Multi-stage quality pipeline applying eight techniques before content enters the corpus.
Read the full card →
Semantic Deduplication
ANN search flagging near-duplicates and derivatives
Detects near-duplicate content using approximate nearest neighbour search.
Read the full card →Ingestion, explained
How does document ingestion actually work?
From raw file to searchable, receipted knowledge in minutes
CueCrux breaks documents into semantic chunks, generates embeddings, builds structural metadata, and creates cryptographic fingerprints for provenance.
Read the full card →
Can I upload scanned or image-based documents?
Yes, built-in OCR handles scans, photos, and faxes
CueCrux includes production-grade OCR that extracts text from scanned PDFs, photographed pages, and faxed documents with quality scoring.
Read the full card →
Does document quality affect answer accuracy?
Yes, and CueCrux tells you exactly how much
CueCrux scores every source chunk for quality and surfaces confidence indicators so you know when an answer rests on low-quality evidence.
Read the full card →