Regulatory RAG Corpus Methodology

How Quinn cites NI 43-101, S-K 1300, PIPEDA, and 10+ other regulatory regimes

What's in the corpus

Quintarth's RAG (Retrieval-Augmented Generation) corpus is a curated collection of ~415 chunks of public-domain regulatory text. Each chunk is a faithful summary (not a verbatim copy) of a specific provision, with a citation back to the authoritative source.

ClusterChunksCoverage
Canadian mineral disclosure~80NI 43-101 + CIM Definition Standards 2014 + provincial rules
US mineral disclosure~30S-K 1300 + SEC mining rules
Canadian privacy~18PIPEDA (federal) + Quebec Law 25 / Bill 64 + Alberta PIPA + BC PIPA
Securities (CA)~75NI 45-106, NI 51-102, NI 52-109, NI 58-101, OSC/AMF/BCSC/ASC
Securities (US)~85SEC Reg S-K, S-X, Form 10-K, 10-Q, 8-K, Form 4, Reg D, Reg S, Reg A+, NMS, FD, SHO, ATS
Cross-border / MJDS~12Multi-Jurisdictional Disclosure System, foreign issuer rules
Compliance / sanctions~25FINTRAC, FCPA, OFAC, Volcker, SOX, Basel III
Mining geology~30Deposit types, mining methods, processing, JORC, SAMREC, CRIRSCO
Markets / instruments~60VWAP, ATS, options, ETF mechanics, SPAC, perp DEX, RWA, MEV

How retrieval works

  1. User query → embedded via Ollama nomic-embed-text (768-dim, English-primary, runs locally on our VPS)
  2. Query vector → Qdrant cosine search → top-K chunks (typically K=5)
  3. Top chunks → packed into Quinn's prompt with their citation metadata
  4. Quinn's response cites each chunk by source (e.g. "per NI 43-101 §4.2") so users can verify

Update cadence

The corpus is curated by hand. New chunks are added in corpus_v{N}.py scripts; ingest is idempotent (deterministic UUIDs prevent duplicates). The current corpus is at v7 (2026-04-25), totaling 415 chunks across 13 jurisdictions.

What it's NOT

The RAG corpus is a research aid, not legal advice. Chunks are summaries of public-domain regulatory text and link back to canonical sources (Justice Canada, OSC, SEC, etc.). Quinn cites the source for every regulatory claim — please verify against the underlying statute or rule before taking any compliance action.