Continuous monitoring, git crawling, and SHA-256 state tracking of external governance sources.
Canon continuously tracks changes to external repositories, statutory registries, and security framework portals (including OWASP LLM Top 10, NIST AI RMF, EU AI Act amendments, and SEC regulatory releases). It pulls raw source files, excludes localized noise, and computes deterministic SHA-256 state hashes.
- Step
- 01 / 05
- Engine
- Canon
- Stage
- Ingest
The Ingest worker operates as an automated web-crawler and AST pre-parser. It listens to git webhooks and regulatory RSS feeds, normalizing non-standard document formats into structured AST node schemas for downstream compilation.
Engineered Canon Ingest crawler, tracking 20+ statutory framework repositories with SHA-256 state matching and zero-copy diff calculation.
# Canon CLI Ingestion Command
$ canon ingest --source="https://github.com/OWASP/www-project-top-10-for-large-language-model-applications" \
--target-domain="OWASP_LLM" \
--compute-hash \
--out=".canon/drafts/owasp_llm_v2.json"
[OK] Fetched 10 rules. SHA-256: 4f8a91b2c... State: DRAFT_READYWhat this stage produces.
- 01
Repository Tracking: Automatically crawling main branches and git trees of official governance registries.
- 02
SHA-256 State Checking: Computing deterministic hashes of rule states to detect unauthorized upstream drift.
- 03
Translation Filtering: Excluding localized folders to focus on canonical English statutory source materials.
- 04
AST Node Normalization: Converting raw markdown/HTML into standardized AST AST_NODE structures.
What cannot be negotiated.
Deterministic Source Matching: Every ingested rule carries a cryptographic SHA-256 origin hash.
Zero-Copy Parsing: Source text is parsed in memory using Tree-sitter native AST queries.
Append-Only Draft Log: Ingested rule states cannot overwrite existing active production manifests without approval.